anton/Reproduce Lesson 05 - NN from Scratch.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "75f11fd6-fb71-473e-8e14-30e4ced8f8f0",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os\n",
    "from pathlib import Path\n",
    "np.set_printoptions(linewidth=140)\n",
    "torch.set_printoptions(linewidth=140, sci_mode=False, edgeitems=7)\n",
    "pd.set_option('display.width', 140)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "63dee6ae-4fd7-4f76-9407-832364a3d63a",
   "metadata": {
    "tags": []
   },
   "source": [
    "# Load the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "9a92b5a2-7ef4-40a4-9245-4b3e8fff7f82",
   "metadata": {},
   "outputs": [],
   "source": [
    "iskaggle = os.environ.get('KAGGLE_KERNEL_RUN_TYPE', '')\n",
    "if iskaggle: path = Path('../input/titanic')\n",
    "else:\n",
    "    path = Path('titanic')\n",
    "    if not path.exists():\n",
    "        import zipfile,kaggle\n",
    "        kaggle.api.competition_download_cli(str(path))\n",
    "        zipfile.ZipFile(f'{path}.zip').extractall(path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "d6fb4d20-a1a1-4482-bb54-4fd3c1255cd1",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(path/'train.csv')\n",
    "df_test = pd.read_csv(path/'test.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "b8dc5ade-8bf1-49c1-8503-c8597980513e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>PassengerId</th>\n",
       "      <th>Survived</th>\n",
       "      <th>Pclass</th>\n",
       "      <th>Name</th>\n",
       "      <th>Sex</th>\n",
       "      <th>Age</th>\n",
       "      <th>SibSp</th>\n",
       "      <th>Parch</th>\n",
       "      <th>Ticket</th>\n",
       "      <th>Fare</th>\n",
       "      <th>Cabin</th>\n",
       "      <th>Embarked</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Braund, Mr. Owen Harris</td>\n",
       "      <td>male</td>\n",
       "      <td>22.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>A/5 21171</td>\n",
       "      <td>7.2500</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Cumings, Mrs. John Bradley (Florence Briggs Th...</td>\n",
       "      <td>female</td>\n",
       "      <td>38.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>PC 17599</td>\n",
       "      <td>71.2833</td>\n",
       "      <td>C85</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>Heikkinen, Miss. Laina</td>\n",
       "      <td>female</td>\n",
       "      <td>26.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>STON/O2. 3101282</td>\n",
       "      <td>7.9250</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Futrelle, Mrs. Jacques Heath (Lily May Peel)</td>\n",
       "      <td>female</td>\n",
       "      <td>35.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>113803</td>\n",
       "      <td>53.1000</td>\n",
       "      <td>C123</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Allen, Mr. William Henry</td>\n",
       "      <td>male</td>\n",
       "      <td>35.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>373450</td>\n",
       "      <td>8.0500</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>886</th>\n",
       "      <td>887</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>Montvila, Rev. Juozas</td>\n",
       "      <td>male</td>\n",
       "      <td>27.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>211536</td>\n",
       "      <td>13.0000</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>887</th>\n",
       "      <td>888</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Graham, Miss. Margaret Edith</td>\n",
       "      <td>female</td>\n",
       "      <td>19.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>112053</td>\n",
       "      <td>30.0000</td>\n",
       "      <td>B42</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>888</th>\n",
       "      <td>889</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Johnston, Miss. Catherine Helen \"Carrie\"</td>\n",
       "      <td>female</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>W./C. 6607</td>\n",
       "      <td>23.4500</td>\n",
       "      <td>NaN</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>889</th>\n",
       "      <td>890</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Behr, Mr. Karl Howell</td>\n",
       "      <td>male</td>\n",
       "      <td>26.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>111369</td>\n",
       "      <td>30.0000</td>\n",
       "      <td>C148</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>890</th>\n",
       "      <td>891</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Dooley, Mr. Patrick</td>\n",
       "      <td>male</td>\n",
       "      <td>32.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>370376</td>\n",
       "      <td>7.7500</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Q</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>891 rows × 12 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     PassengerId  Survived  Pclass                                               Name     Sex   Age  SibSp  Parch            Ticket  \\\n",
       "0              1         0       3                            Braund, Mr. Owen Harris    male  22.0      1      0         A/5 21171   \n",
       "1              2         1       1  Cumings, Mrs. John Bradley (Florence Briggs Th...  female  38.0      1      0          PC 17599   \n",
       "2              3         1       3                             Heikkinen, Miss. Laina  female  26.0      0      0  STON/O2. 3101282   \n",
       "3              4         1       1       Futrelle, Mrs. Jacques Heath (Lily May Peel)  female  35.0      1      0            113803   \n",
       "4              5         0       3                           Allen, Mr. William Henry    male  35.0      0      0            373450   \n",
       "..           ...       ...     ...                                                ...     ...   ...    ...    ...               ...   \n",
       "886          887         0       2                              Montvila, Rev. Juozas    male  27.0      0      0            211536   \n",
       "887          888         1       1                       Graham, Miss. Margaret Edith  female  19.0      0      0            112053   \n",
       "888          889         0       3           Johnston, Miss. Catherine Helen \"Carrie\"  female   NaN      1      2        W./C. 6607   \n",
       "889          890         1       1                              Behr, Mr. Karl Howell    male  26.0      0      0            111369   \n",
       "890          891         0       3                                Dooley, Mr. Patrick    male  32.0      0      0            370376   \n",
       "\n",
       "        Fare Cabin Embarked  \n",
       "0     7.2500   NaN        S  \n",
       "1    71.2833   C85        C  \n",
       "2     7.9250   NaN        S  \n",
       "3    53.1000  C123        S  \n",
       "4     8.0500   NaN        S  \n",
       "..       ...   ...      ...  \n",
       "886  13.0000   NaN        S  \n",
       "887  30.0000   B42        S  \n",
       "888  23.4500   NaN        S  \n",
       "889  30.0000  C148        C  \n",
       "890   7.7500   NaN        Q  \n",
       "\n",
       "[891 rows x 12 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e4d0a47a-edb1-4a98-aec7-48c3356259c5",
   "metadata": {},
   "source": [
    "# Wrangling\n",
    "\n",
    "Handle missing values, categorical values, etc."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "5e4f4973-e91d-47dc-8f45-dabd0a550da0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>PassengerId</th>\n",
       "      <th>Survived</th>\n",
       "      <th>Pclass</th>\n",
       "      <th>Age</th>\n",
       "      <th>SibSp</th>\n",
       "      <th>Parch</th>\n",
       "      <th>Fare</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>891.000000</td>\n",
       "      <td>891.000000</td>\n",
       "      <td>891.000000</td>\n",
       "      <td>714.000000</td>\n",
       "      <td>891.000000</td>\n",
       "      <td>891.000000</td>\n",
       "      <td>891.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>446.000000</td>\n",
       "      <td>0.383838</td>\n",
       "      <td>2.308642</td>\n",
       "      <td>29.699118</td>\n",
       "      <td>0.523008</td>\n",
       "      <td>0.381594</td>\n",
       "      <td>32.204208</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>257.353842</td>\n",
       "      <td>0.486592</td>\n",
       "      <td>0.836071</td>\n",
       "      <td>14.526497</td>\n",
       "      <td>1.102743</td>\n",
       "      <td>0.806057</td>\n",
       "      <td>49.693429</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.420000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>223.500000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>2.000000</td>\n",
       "      <td>20.125000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>7.910400</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>446.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>28.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>14.454200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>668.500000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>38.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>31.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>891.000000</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>3.000000</td>\n",
       "      <td>80.000000</td>\n",
       "      <td>8.000000</td>\n",
       "      <td>6.000000</td>\n",
       "      <td>512.329200</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       PassengerId    Survived      Pclass         Age       SibSp       Parch        Fare\n",
       "count   891.000000  891.000000  891.000000  714.000000  891.000000  891.000000  891.000000\n",
       "mean    446.000000    0.383838    2.308642   29.699118    0.523008    0.381594   32.204208\n",
       "std     257.353842    0.486592    0.836071   14.526497    1.102743    0.806057   49.693429\n",
       "min       1.000000    0.000000    1.000000    0.420000    0.000000    0.000000    0.000000\n",
       "25%     223.500000    0.000000    2.000000   20.125000    0.000000    0.000000    7.910400\n",
       "50%     446.000000    0.000000    3.000000   28.000000    0.000000    0.000000   14.454200\n",
       "75%     668.500000    1.000000    3.000000   38.000000    1.000000    0.000000   31.000000\n",
       "max     891.000000    1.000000    3.000000   80.000000    8.000000    6.000000  512.329200"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "8e53b0f1-8a29-4365-b06d-2da349eaf0c9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.Fare.hist()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2db21722-167f-4ab7-a05f-420a62b23161",
   "metadata": {},
   "source": [
    "Fare is a monetary value, best handled using log and then normalized."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "264873b3-b192-4bf2-a281-13ce7cb34f41",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Name</th>\n",
       "      <th>Sex</th>\n",
       "      <th>Ticket</th>\n",
       "      <th>Cabin</th>\n",
       "      <th>Embarked</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>891</td>\n",
       "      <td>891</td>\n",
       "      <td>891</td>\n",
       "      <td>204</td>\n",
       "      <td>889</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>unique</th>\n",
       "      <td>891</td>\n",
       "      <td>2</td>\n",
       "      <td>681</td>\n",
       "      <td>147</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>top</th>\n",
       "      <td>Braund, Mr. Owen Harris</td>\n",
       "      <td>male</td>\n",
       "      <td>347082</td>\n",
       "      <td>B96 B98</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>freq</th>\n",
       "      <td>1</td>\n",
       "      <td>577</td>\n",
       "      <td>7</td>\n",
       "      <td>4</td>\n",
       "      <td>644</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                           Name   Sex  Ticket    Cabin Embarked\n",
       "count                       891   891     891      204      889\n",
       "unique                      891     2     681      147        3\n",
       "top     Braund, Mr. Owen Harris  male  347082  B96 B98        S\n",
       "freq                          1   577       7        4      644"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe(include=object)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bdfccdd7-b179-448b-bdd2-f083c8c7f913",
   "metadata": {},
   "source": [
    "Sex and Embarked are categoricals with only 2 and 3 unique values. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "0f37fc32-b3d3-4b3f-9fe5-586ccaff3d00",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "PassengerId      0\n",
       "Survived         0\n",
       "Pclass           0\n",
       "Name             0\n",
       "Sex              0\n",
       "Age            177\n",
       "SibSp            0\n",
       "Parch            0\n",
       "Ticket           0\n",
       "Fare             0\n",
       "Cabin          687\n",
       "Embarked         2\n",
       "dtype: int64"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.isna().sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "443fa64b-029d-435e-b347-515e5a8ec43f",
   "metadata": {},
   "source": [
    "Handle missing values using the modes of the columns."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "77840792-bb0e-40de-a9d1-2d3d832b5c8d",
   "metadata": {},
   "outputs": [],
   "source": [
    "modes = df.mode().iloc[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "a1a2647b-0928-4363-9caf-95f7aad5aa38",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df.fillna(modes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "73412352-815f-4f49-97ff-524b6ac71aef",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.Fare.apply(np.log1p).hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "d0537628-6612-4ed9-bb2b-db53c72a1019",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.Fare = df.Fare.apply(np.log1p)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "ad053c86-4cd5-42b1-8479-6368fab16bbd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>PassengerId</th>\n",
       "      <th>Survived</th>\n",
       "      <th>Pclass</th>\n",
       "      <th>Name</th>\n",
       "      <th>Sex</th>\n",
       "      <th>Age</th>\n",
       "      <th>SibSp</th>\n",
       "      <th>Parch</th>\n",
       "      <th>Ticket</th>\n",
       "      <th>Fare</th>\n",
       "      <th>Cabin</th>\n",
       "      <th>Embarked</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Braund, Mr. Owen Harris</td>\n",
       "      <td>male</td>\n",
       "      <td>22.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>A/5 21171</td>\n",
       "      <td>2.110213</td>\n",
       "      <td>B96 B98</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Cumings, Mrs. John Bradley (Florence Briggs Th...</td>\n",
       "      <td>female</td>\n",
       "      <td>38.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>PC 17599</td>\n",
       "      <td>4.280593</td>\n",
       "      <td>C85</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>Heikkinen, Miss. Laina</td>\n",
       "      <td>female</td>\n",
       "      <td>26.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>STON/O2. 3101282</td>\n",
       "      <td>2.188856</td>\n",
       "      <td>B96 B98</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Futrelle, Mrs. Jacques Heath (Lily May Peel)</td>\n",
       "      <td>female</td>\n",
       "      <td>35.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>113803</td>\n",
       "      <td>3.990834</td>\n",
       "      <td>C123</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Allen, Mr. William Henry</td>\n",
       "      <td>male</td>\n",
       "      <td>35.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>373450</td>\n",
       "      <td>2.202765</td>\n",
       "      <td>B96 B98</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   PassengerId  Survived  Pclass                                               Name     Sex   Age  SibSp  Parch            Ticket  \\\n",
       "0            1         0       3                            Braund, Mr. Owen Harris    male  22.0      1      0         A/5 21171   \n",
       "1            2         1       1  Cumings, Mrs. John Bradley (Florence Briggs Th...  female  38.0      1      0          PC 17599   \n",
       "2            3         1       3                             Heikkinen, Miss. Laina  female  26.0      0      0  STON/O2. 3101282   \n",
       "3            4         1       1       Futrelle, Mrs. Jacques Heath (Lily May Peel)  female  35.0      1      0            113803   \n",
       "4            5         0       3                           Allen, Mr. William Henry    male  35.0      0      0            373450   \n",
       "\n",
       "       Fare    Cabin Embarked  \n",
       "0  2.110213  B96 B98        S  \n",
       "1  4.280593      C85        C  \n",
       "2  2.188856  B96 B98        S  \n",
       "3  3.990834     C123        S  \n",
       "4  2.202765  B96 B98        S  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "db1cbc30-bff1-44f1-b799-afc9377d69e8",
   "metadata": {},
   "outputs": [],
   "source": [
    "cats = ['Sex', 'Pclass', 'Embarked']\n",
    "df = pd.get_dummies(df, columns=cats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "3ddbe6b4-271b-4abc-933f-88c5a65e0665",
   "metadata": {},
   "outputs": [],
   "source": [
    "drop_cols = ['PassengerId', 'Name', 'Ticket', 'Cabin']\n",
    "df = df.drop(columns=drop_cols)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "3bd5d7ab-2476-4728-802f-42b760153f96",
   "metadata": {},
   "outputs": [],
   "source": [
    "# normalize\n",
    "maxes = df.max()\n",
    "df = df.div(maxes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "6e9f3fad-41cb-47a7-921a-7c0d2031e5d1",
   "metadata": {},
   "outputs": [],
   "source": [
    "# variable must be defined on train set:\n",
    "#   modes\n",
    "#   maxes\n",
    "\n",
    "def preprocess(df):\n",
    "    cats = ['Sex', 'Pclass', 'Embarked']\n",
    "    drop_cols = ['PassengerId', 'Name', 'Ticket', 'Cabin']\n",
    "    \n",
    "    df.Fare = df.Fare.apply(np.log1p)\n",
    "    \n",
    "    df = df.fillna(modes)\n",
    "    \n",
    "    df = pd.get_dummies(df, columns=cats)\n",
    "    \n",
    "    df = (df.drop(columns=drop_cols)\n",
    "            .div(maxes)\n",
    "         )\n",
    "    \n",
    "    \n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "5aea815e-15c4-4565-a4ea-05f75952b6f0",
   "metadata": {},
   "outputs": [],
   "source": [
    "comp = df.copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "5881adf5-3aef-4211-9561-01b94c6afc7a",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(path/'train.csv')\n",
    "df_test = pd.read_csv(path/'test.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "9cf41fa9-f53f-4f09-b2b1-1d3ef8e7b914",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Survived      891\n",
       "Age           891\n",
       "SibSp         891\n",
       "Parch         891\n",
       "Fare          891\n",
       "Sex_female    891\n",
       "Sex_male      891\n",
       "Pclass_1      891\n",
       "Pclass_2      891\n",
       "Pclass_3      891\n",
       "Embarked_C    891\n",
       "Embarked_Q    891\n",
       "Embarked_S    891\n",
       "dtype: int64"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = preprocess(df)\n",
    "\n",
    "(df == comp).sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "27bb112f-91d9-4a3c-8381-f502ef781d5d",
   "metadata": {},
   "source": [
    "train/valid split, and into tensors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "7659fd59-0223-458d-ac37-96ca8aef3097",
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.data.transforms import RandomSplitter\n",
    "from torch import tensor\n",
    "\n",
    "# see docs at https://docs.fast.ai/data.transforms.html\n",
    "train_idx, valid_idx = RandomSplitter(valid_pct=0.2, seed=42)(df)\n",
    "\n",
    "X_train, y_train = df.iloc[train_idx].drop(columns=['Survived']), df.iloc[train_idx]['Survived']\n",
    "X_valid, y_valid = df.iloc[valid_idx].drop(columns=['Survived']), df.iloc[valid_idx]['Survived']\n",
    "\n",
    "X_train = tensor(X_train.values, dtype=torch.float)\n",
    "X_valid = tensor(X_valid.values, dtype=torch.float)\n",
    "y_train = tensor(y_train.values, dtype=torch.float)\n",
    "y_valid = tensor(y_valid.values, dtype=torch.float)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "99ff2b92-31e2-42ee-93d2-0baa9995a60a",
   "metadata": {},
   "source": [
    "# Create linear learner"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f38bd3c9-f22c-47cf-9e30-7952bddfa556",
   "metadata": {},
   "source": [
    "Ok let's remember what we're even doing here. For a linear learner, we're going to need a set of randomly initialized weights equal in size to the length of the input. We're going to need a constant bias term (either a ones col, or just an extra param). We're going to need need a loss function, calculate gradients according to that loss, and a way of updating weights based on the gradients. Loop over epochs, print out loss and print out validation accuracy. Let's give it a whirl.\n",
    "\n",
    "(After doing all that, THEN I'll go back and see whether I missed anything.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "2743fadd-96f5-4bd2-bee9-8cc90a3ca181",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "12"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "8edfb10c-bf35-4761-ab7f-9a2274c9dd7d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([ 0.8738, -0.1948,  1.1477, -0.6449,  1.3456,  0.0802, -1.3830, -1.1722,  0.3663,  1.5820,  0.6060, -0.4627]),\n",
       " tensor([0.3514]))"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "weights = torch.randn(X_train.shape[1])\n",
    "bias = torch.randn(1)\n",
    "weights, bias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "9373712d-679b-439d-9a14-df0ca7fe338e",
   "metadata": {},
   "outputs": [],
   "source": [
    "def init_params():\n",
    "    weights = torch.randn(X_train.shape[1]).requires_grad_()\n",
    "    bias = torch.randn(1).requires_grad_()\n",
    "    return weights, bias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "41423512-6f0f-44fb-9417-58474fa912a8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([-0.7849, -0.8169,  1.3036, -1.8624, -1.0653, -0.0830, -1.2250, -0.0515, -0.5433,  1.7597, -1.4369, -0.9020], requires_grad=True),\n",
       " tensor([-1.9090], requires_grad=True))"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "weights, bias = init_params()\n",
    "weights, bias"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "97551ebe-be90-4e3f-9e47-ec26fa4ff131",
   "metadata": {},
   "source": [
    "for loss, we'll use mean absolute error. this is called the L1 loss. we can define this ourselves, or use the torch.nn.L1Loss function\n",
    "```loss = nn.L1Loss()```"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "c6f2e50d-1138-4b3c-ba58-ed0923421562",
   "metadata": {},
   "outputs": [],
   "source": [
    "def mean_abs_error(preds, target):\n",
    "    return torch.abs(target-preds).mean()\n",
    "\n",
    "loss_fn = mean_abs_error\n",
    "\n",
    "# import torch.nn as nn\n",
    "# loss_func = nn.L1Loss()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "644c184c-60d7-42a6-ad93-1d6b3fa5e341",
   "metadata": {},
   "source": [
    "so to compute the forward preds, we need to multiple all the inputs by the weights, plus the bias, and sum across rows. let's try this..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "37c2c07d-14a1-4c47-a514-bd426cd6f9ca",
   "metadata": {},
   "outputs": [],
   "source": [
    "preds = X_train@weights + bias\n",
    "preds = torch.sigmoid(preds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "d41982d5-b5b4-40ca-bba7-b90c33791aa8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(0.3860, grad_fn=<MeanBackward0>)"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "loss_fn(preds, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "3dafdd81-5f7a-419d-9710-7b7289b71207",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss = loss_fn(preds, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "f6a81519-a666-44ea-a394-fe0d635de700",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([ 0.0029,  0.0001, -0.0002,  0.0021, -0.0024,  0.0090, -0.0001,  0.0006,  0.0062,  0.0027,  0.0002,  0.0037]),\n",
       " tensor([0.0066]))"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "loss.backward()\n",
    "weights.grad, bias.grad"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "b6187383-df8f-4443-87a2-98b3ec793f16",
   "metadata": {},
   "outputs": [],
   "source": [
    "lr = 1e-5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "d7474309-989d-4114-9214-c2e54c4996e8",
   "metadata": {},
   "outputs": [],
   "source": [
    "params = [weights, bias]\n",
    "for p in params:\n",
    "    p.data -= p.grad * lr\n",
    "    p.grad.zero_()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "59b72d3a-6959-43ab-846c-f027226623b5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# how do we get the accuracy? \n",
    "preds_valid = X_valid@weights + bias\n",
    "preds_valid = torch.sigmoid(preds_valid)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "ec8c9410-7ce0-4d2a-b5d1-a149766ac595",
   "metadata": {},
   "outputs": [],
   "source": [
    "def forward(X, weights, biases):\n",
    "    preds = X@weights + biases\n",
    "    preds = torch.sigmoid(preds)\n",
    "    return preds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "id": "84603ae5-1aec-4805-9a5c-98270b65c425",
   "metadata": {},
   "outputs": [],
   "source": [
    "def acc_fn(preds, target):\n",
    "    acc = (target.bool() == (preds>0.5)).float().mean()\n",
    "    return acc\n",
    "\n",
    "acc = acc_fn(preds_valid, y_valid)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d0748a51-ab88-4868-a95e-75d07c509f78",
   "metadata": {},
   "source": [
    "Ok I think I have all the pieces - let's see if we can run this multiple times and converge."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "id": "ab8c30f4-a672-4559-a17a-ab98e004b7fc",
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_model(epochs=20, lr=60):\n",
    "    weights, biases = init_params()\n",
    "    \n",
    "    for n in range(epochs):\n",
    "        preds = forward(X_train, weights, biases)\n",
    "        loss = loss_fn(preds, y_train)\n",
    "        loss.backward()\n",
    "        for p in [weights, biases]:\n",
    "            p.data -= p.grad * lr\n",
    "            p.grad.zero_()\n",
    "            \n",
    "        with torch.no_grad():\n",
    "            preds_valid = forward(X_valid, weights, biases)\n",
    "            acc = acc_fn(preds_valid, y_valid)\n",
    "        \n",
    "        print(f'Epoch: {n}. Loss: {loss:.4f}. Valid acc: {acc:.4f}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "id": "a456dc48-21f4-4c8d-9a8b-ad1e3b68bb0f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 0. Loss: 0.6008. Valid acc: 0.7921\n",
      "Epoch: 1. Loss: 0.3191. Valid acc: 0.6404\n",
      "Epoch: 2. Loss: 0.3454. Valid acc: 0.8202\n",
      "Epoch: 3. Loss: 0.2327. Valid acc: 0.7921\n",
      "Epoch: 4. Loss: 0.2208. Valid acc: 0.8034\n",
      "Epoch: 5. Loss: 0.2093. Valid acc: 0.8371\n",
      "Epoch: 6. Loss: 0.2055. Valid acc: 0.8258\n",
      "Epoch: 7. Loss: 0.2045. Valid acc: 0.8258\n",
      "Epoch: 8. Loss: 0.2036. Valid acc: 0.8258\n",
      "Epoch: 9. Loss: 0.2029. Valid acc: 0.8258\n",
      "Epoch: 10. Loss: 0.2023. Valid acc: 0.8258\n",
      "Epoch: 11. Loss: 0.2017. Valid acc: 0.8258\n",
      "Epoch: 12. Loss: 0.2012. Valid acc: 0.8258\n",
      "Epoch: 13. Loss: 0.2008. Valid acc: 0.8258\n",
      "Epoch: 14. Loss: 0.2004. Valid acc: 0.8258\n",
      "Epoch: 15. Loss: 0.2001. Valid acc: 0.8258\n",
      "Epoch: 16. Loss: 0.1997. Valid acc: 0.8258\n",
      "Epoch: 17. Loss: 0.1995. Valid acc: 0.8258\n",
      "Epoch: 18. Loss: 0.1992. Valid acc: 0.8258\n",
      "Epoch: 19. Loss: 0.1989. Valid acc: 0.8258\n"
     ]
    }
   ],
   "source": [
    "train_model()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cd74ebec-8123-4f8c-ab06-d59511dc1942",
   "metadata": {},
   "source": [
    "Lessons learned: \n",
    "- value initialization really matters. initial learning rate of 1e-5 had me worried that the code wasn't working at all. learning rate of 50 ended up working best.\n",
    "- variable scoping. jesus. related to initial learning rate problem. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3c1060e0-2329-4040-840e-c4c208e52f10",
   "metadata": {},
   "source": [
    "# Create 2-layer neural network"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3f4e9b41-df2d-4e3c-beab-c7f3f5cf4c74",
   "metadata": {},
   "source": [
    "Alright so now we want to do two layers with a non-linear activation between. let's figure out how to structure this with code. not much is going to need to change..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "id": "c964f1c1-ab48-449d-b44b-69ff49fb533a",
   "metadata": {},
   "outputs": [],
   "source": [
    "n_hidden = 10\n",
    "w1 = torch.randn(X_train.shape[1], n_hidden).requires_grad_()\n",
    "w2 = torch.randn(n_hidden).requires_grad_()\n",
    "weights = [w1, w2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "id": "b97e9812-17c7-4dd9-8984-1a1720455abb",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "loss_fn = nn.L1Loss()\n",
    "\n",
    "def init_params():\n",
    "    n_hidden = 10\n",
    "    w1 = torch.randn(X_train.shape[1], n_hidden).requires_grad_()\n",
    "    w2 = torch.randn(n_hidden).requires_grad_()\n",
    "    weights = [w1, w2]\n",
    "    biases = [torch.randn(1)[0].requires_grad_(), torch.randn(1)[0].requires_grad_()]\n",
    "    return weights, biases\n",
    "\n",
    "def forward(X, params):\n",
    "    weights, biases = params\n",
    "    preds = X@weights[0] + biases[0]\n",
    "    preds = F.relu(preds)\n",
    "    preds = preds@weights[1] + biases[1]\n",
    "    preds = torch.sigmoid(preds)\n",
    "    return preds\n",
    "\n",
    "# no change\n",
    "def acc_fn(preds, target):\n",
    "    acc = (target.bool() == (preds>0.5)).float().mean()\n",
    "    return acc\n",
    "\n",
    "def train_model(epochs=20, lr=10):\n",
    "    params = init_params()\n",
    "    weights, biases = params\n",
    "    \n",
    "    for n in range(epochs):\n",
    "        preds = forward(X_train, params)\n",
    "        loss = loss_fn(preds, y_train)\n",
    "        loss.backward()\n",
    "        for p in weights+biases:\n",
    "            p.data -= p.grad * lr\n",
    "            p.grad.zero_()\n",
    "            \n",
    "        with torch.no_grad():\n",
    "            preds_valid = forward(X_valid, params)\n",
    "            acc = acc_fn(preds_valid, y_valid)\n",
    "        \n",
    "        print(f'Epoch: {n}. Loss: {loss:.4f}. Valid acc: {acc:.4f}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "id": "50c76f3b-52e2-44b1-992a-7a914b3719b6",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 0. Loss: 0.6109. Valid acc: 0.4101\n",
      "Epoch: 1. Loss: 0.5584. Valid acc: 0.7584\n",
      "Epoch: 2. Loss: 0.3520. Valid acc: 0.8202\n",
      "Epoch: 3. Loss: 0.2465. Valid acc: 0.8258\n",
      "Epoch: 4. Loss: 0.2173. Valid acc: 0.8202\n",
      "Epoch: 5. Loss: 0.2007. Valid acc: 0.8202\n",
      "Epoch: 6. Loss: 0.1971. Valid acc: 0.8258\n",
      "Epoch: 7. Loss: 0.1960. Valid acc: 0.8258\n",
      "Epoch: 8. Loss: 0.1952. Valid acc: 0.8258\n",
      "Epoch: 9. Loss: 0.1946. Valid acc: 0.8258\n",
      "Epoch: 10. Loss: 0.1940. Valid acc: 0.8258\n",
      "Epoch: 11. Loss: 0.1936. Valid acc: 0.8258\n",
      "Epoch: 12. Loss: 0.1933. Valid acc: 0.8258\n",
      "Epoch: 13. Loss: 0.1931. Valid acc: 0.8258\n",
      "Epoch: 14. Loss: 0.1929. Valid acc: 0.8258\n",
      "Epoch: 15. Loss: 0.1928. Valid acc: 0.8258\n",
      "Epoch: 16. Loss: 0.1926. Valid acc: 0.8258\n",
      "Epoch: 17. Loss: 0.1925. Valid acc: 0.8258\n",
      "Epoch: 18. Loss: 0.1924. Valid acc: 0.8258\n",
      "Epoch: 19. Loss: 0.1923. Valid acc: 0.8258\n"
     ]
    }
   ],
   "source": [
    "train_model()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4df83c6b-782f-4b6a-853b-76089e5a7f2d",
   "metadata": {},
   "source": [
    "# Create 3-layer neural network"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c89cc7cf-8168-4a21-8bd2-b1ae2b1df28a",
   "metadata": {},
   "source": [
    "let's add a hidden layer...\n",
    "\n",
    "gotta say, this one really sealed the deal as far as understanding what's going on with layers and parameter updates. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 206,
   "id": "72e1bc40-66a8-4537-9cf2-82770ba4334a",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch.nn.functional as F\n",
    "loss_fn = F.l1_loss\n",
    "\n",
    "def init_params():\n",
    "    n_hidden = 10\n",
    "    w1 = torch.randn(X_train.shape[1], n_hidden).requires_grad_()\n",
    "    w2 = torch.randn(n_hidden, n_hidden).requires_grad_()\n",
    "    w3 = torch.randn(n_hidden, 1).requires_grad_()\n",
    "    weights = [w1, w2, w3]\n",
    "    biases = [torch.randn(1)[0].requires_grad_() for _ in range(len(weights))]\n",
    "    return weights, biases\n",
    "\n",
    "def forward(X, params):\n",
    "    weights, biases = params\n",
    "    n = len(weights)\n",
    "    \n",
    "    preds = X@weights[0] + biases[0]\n",
    "    preds = F.relu(preds)\n",
    "    for i in range(1, n):\n",
    "        preds = preds@weights[i] + biases[i]\n",
    "        if i < n-1:\n",
    "            preds = F.relu(preds)\n",
    "        else: \n",
    "            preds = torch.sigmoid(preds)\n",
    "    return preds\n",
    "\n",
    "# no change\n",
    "def acc_fn(preds, target):\n",
    "    acc = (target.bool() == (preds>0.5)).float().mean()\n",
    "    return acc\n",
    "\n",
    "def train_model(epochs=20, lr=10):\n",
    "    params = init_params()\n",
    "    weights, biases = params\n",
    "    \n",
    "    for n in range(epochs):\n",
    "        preds = forward(X_train, params)\n",
    "        loss = loss_fn(preds, y_train[:,None])\n",
    "        loss.backward()\n",
    "        for p in weights+biases:\n",
    "            p.data -= p.grad * lr\n",
    "            p.grad.zero_()\n",
    "            \n",
    "        with torch.no_grad():\n",
    "            preds_valid = forward(X_valid, params)\n",
    "            acc = acc_fn(preds_valid, y_valid[:,None])\n",
    "        \n",
    "        print(f'Epoch: {n}. Loss: {loss:.4f}. Valid acc: {acc:.4f}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 214,
   "id": "77a6df89-998d-4e7c-b6a3-122daf7a250f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 0. Loss: 0.3856. Valid acc: 0.7584\n",
      "Epoch: 1. Loss: 0.2782. Valid acc: 0.7753\n",
      "Epoch: 2. Loss: 0.2186. Valid acc: 0.6910\n",
      "Epoch: 3. Loss: 0.3193. Valid acc: 0.7360\n",
      "Epoch: 4. Loss: 0.2739. Valid acc: 0.7079\n",
      "Epoch: 5. Loss: 0.2927. Valid acc: 0.7921\n",
      "Epoch: 6. Loss: 0.2129. Valid acc: 0.7921\n",
      "Epoch: 7. Loss: 0.2126. Valid acc: 0.7978\n",
      "Epoch: 8. Loss: 0.2113. Valid acc: 0.8034\n",
      "Epoch: 9. Loss: 0.2056. Valid acc: 0.8090\n",
      "Epoch: 10. Loss: 0.2032. Valid acc: 0.8146\n",
      "Epoch: 11. Loss: 0.1988. Valid acc: 0.8202\n",
      "Epoch: 12. Loss: 0.1943. Valid acc: 0.8202\n",
      "Epoch: 13. Loss: 0.1936. Valid acc: 0.8202\n",
      "Epoch: 14. Loss: 0.1932. Valid acc: 0.8315\n",
      "Epoch: 15. Loss: 0.1929. Valid acc: 0.8371\n",
      "Epoch: 16. Loss: 0.1931. Valid acc: 0.8371\n",
      "Epoch: 17. Loss: 0.1927. Valid acc: 0.8371\n",
      "Epoch: 18. Loss: 0.1924. Valid acc: 0.8315\n",
      "Epoch: 19. Loss: 0.1913. Valid acc: 0.8258\n"
     ]
    }
   ],
   "source": [
    "train_model()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d7da0bad-2803-4882-ae92-5b2ef2c18720",
   "metadata": {},
   "source": [
    "# Using a Framework"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "da133e96-5627-440a-b91e-2701a9621d1f",
   "metadata": {},
   "source": [
    "Here we'll use fastai and PyTorch to simplify many of those steps above."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "473a5560-a2cb-4f15-9a14-ad00b309e39d",
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.tabular.all import *\n",
    "\n",
    "set_seed(42)\n",
    "\n",
    "df = pd.read_csv(path/'train.csv')\n",
    "df_test = pd.read_csv(path/'test.csv')\n",
    "\n",
    "splits = RandomSplitter(seed=42)(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "b8886ac9-64e8-471d-ada3-ab83b4d53070",
   "metadata": {},
   "outputs": [],
   "source": [
    "# variable must be defined on train set:\n",
    "#   modes\n",
    "#   maxes\n",
    "\n",
    "def preprocess(df):\n",
    "    df.Fare = df.Fare.fillna(0)\n",
    "    df['LogFare'] = df.Fare.apply(np.log1p)\n",
    "    \n",
    "    df['Family'] = df.SibSp+df.Parch\n",
    "    df['Alone'] = df.Family==0\n",
    "    df['Deck'] = (df.Cabin.str[0]\n",
    "                    .map(dict(A='ABC',\n",
    "                              B='ABC',\n",
    "                              C='ABC',\n",
    "                              D='DE',\n",
    "                              E='DE',\n",
    "                              F='FG',\n",
    "                              G='FG'))\n",
    "                 )\n",
    "    df['Title'] = (df.Name\n",
    "                     .str.split(', ', expand=True)[1]\n",
    "                     .str.split('.', expand=True)[0]\n",
    "                     .map(dict(Mr='Mr',\n",
    "                               Miss='Miss',\n",
    "                               Mrs='Mrs',\n",
    "                               Master='Master'))\n",
    "                  )\n",
    "    \n",
    "    # no longer need this, as will be handled by dataloader\n",
    "#     df = df.fillna(modes)\n",
    "    \n",
    "#     cats = ['Sex', 'Pclass', 'Embarked']\n",
    "#     drop_cols = ['PassengerId', 'Name', 'Ticket', 'Cabin']\n",
    "    \n",
    "#     df = pd.get_dummies(df, columns=cats)\n",
    "    \n",
    "#     df = (df.drop(columns=drop_cols)\n",
    "#             .div(maxes)\n",
    "#          )\n",
    "    \n",
    "    \n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "0cc78c54-f234-4ac9-b31a-3b9ba0801096",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['PassengerId', 'Survived', 'Pclass', 'Name', 'Sex', 'Age', 'SibSp', 'Parch', 'Ticket', 'Fare', 'Cabin', 'Embarked', 'Family',\n",
       "       'Alone', 'Deck', 'Title', 'LogFare'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = preprocess(df)\n",
    "df.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "443dd622-f0a9-4b3e-a164-382ee577cba2",
   "metadata": {},
   "outputs": [],
   "source": [
    "# see https://docs.fast.ai/tabular.core.html#tabularpandas\n",
    "dls = TabularPandas(df,\n",
    "                    procs=[Categorify, FillMissing, Normalize], \n",
    "                    cat_names=['Sex', 'Pclass', 'Embarked', 'Deck', 'Title'], \n",
    "                    cont_names=['Age', 'SibSp', 'Parch', 'LogFare', 'Family', 'Alone'],\n",
    "                    y_names='Survived',\n",
    "                    y_block=CategoryBlock(),\n",
    "                    splits=splits,\n",
    "                    do_setup=True,\n",
    "                    device=None,\n",
    "                    inplace=False,\n",
    "                    reduce_memory=True).dataloaders(path='.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "58d1f498-1324-46b6-8d65-50e973da7e58",
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = tabular_learner(dls,\n",
    "                        metrics=accuracy,\n",
    "                        layers=[10,10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "7bd02950-7244-40f4-af65-4bf2b346027c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "SuggestedLRs(slide=0.0831763744354248, valley=0.015848932787775993)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAkAAAAG1CAYAAAARLUsBAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/NK7nSAAAACXBIWXMAAA9hAAAPYQGoP6dpAABpWklEQVR4nO3deVxU5f4H8M+ZgWHfEUR2xQXEDVATxT1cCrdKS9PMrCxvpWalP6tbtlhmRhtmLnkrU8qlvDdc0FRQzAWXXBAXQJBVRPZ95vz+QKZGUFlmOAPzeb9e51XzzHPOfB9A+fqsgiiKIoiIiIgMiEzqAIiIiIhaGhMgIiIiMjhMgIiIiMjgMAEiIiIig8MEiIiIiAwOEyAiIiIyOEyAiIiIyOAwASIiIiKDYyR1APpIpVIhIyMDVlZWEARB6nCIiIioAURRRFFRETp06ACZ7N59PEyA6pGRkQF3d3epwyAiIqImSEtLg5ub2z3rMAGqh5WVFYCaL6C1tbXE0RAREVFDFBYWwt3dXf17/F6YANWjdtjL2tqaCRAREVEr05DpK5wETURERAaHCRAREREZHA6BERERaZFSqURVVZXUYbRZCoXiviu8GoIJEBERkRaIooisrCzk5+dLHUqbJpPJ4O3tDYVC0aznMAEiIiLSgtrkx8nJCebm5txHTgdq9+nLzMyEh4dHs77GTICIiIiaSalUqpMfBwcHqcNp09q1a4eMjAxUV1fD2Ni4yc/hJGgiIqJmqp3zY25uLnEkbV/t0JdSqWzWc5gAERERaQmHvXRPW19jJkBERERkcJgAERERkcFhAkRERKQvVEogORY4u6Xmv6rmzXNpjpkzZ2LChAnq10OHDsW8efPueY+XlxfCw8N1Gpe2cBUYERGRPriwA9j1BlCY8XeZdQdg9MeA3zjp4rpt27ZtzVp1pW/YA9SCzmcUYPZ/TuDVn89IHQoREemTCzuAn2doJj8AUJhZU35hhzRx/YO9vX2DTllvLZgAtaBqpYi9Cdk4fCVX6lCIiEhfqJQ1PT8Q63nzdtmuRTobDtuyZQt69OgBMzMzODg4YOTIkSgpKalT784hsJycHISFhcHMzAze3t7YuHFjnXsKCgrw3HPPwcnJCdbW1hg+fDjOnNGPTgAOgbUgF1tTAEB2UTmqlCoYy5l/EhEZvGtxdXt+NIhAYXpNPe8QrX50ZmYmnnjiCSxfvhwTJ05EUVERYmNjIYr1JWOaZs6cibS0NPzxxx9QKBR4+eWXkZOT83fUooiHHnoI9vb2iIqKgo2NDVavXo0RI0bg0qVLsLe312pbGosJUAtytDCBQi5DpVKF7MJyuNlxwywiIoNXnK3deo2QmZmJ6upqTJo0CZ6engCAHj163Pe+S5cuYefOnfjzzz/Rv39/AMC6devg6+urrrN//36cPXsWOTk5MDExAQCsWLECv/76K7Zs2YLnnntO6+1pDHZBtCCZTEB7m5peoIz8comjISIivWDprN16jdCrVy+MGDECPXr0wGOPPYY1a9bg1q1b970vISEBRkZGCAoKUpd169YNtra26tfx8fEoLi6Gg4MDLC0t1VdycjKuXr2q9bY0FnuAWlgHW1Ok5pUis6BM6lCIiEgfeAbXrPYqzET984CEmvc9g7X+0XK5HNHR0YiLi8OePXvw5ZdfYsmSJTh69Og976sdIrvXrswqlQouLi44cOBAnff+mShJhT1ALayDjRkAID2fCRAREQGQyWuWugMA7kwobr8e/VFNPR0QBAEDBw7Eu+++i1OnTkGhUGD79u33vMfX1xfV1dU4ceKEuiwxMRH5+fnq1wEBAcjKyoKRkRF8fHw0LkdHR520pTGYALWwDrY1CVAmh8CIiKiW3zhg8veAtYtmuXWHmnId7QN09OhRfPjhhzhx4gRSU1Oxbds23LhxQ2MuT326du2K0aNH49lnn8XRo0cRHx+P2bNnw8zMTF1n5MiRGDBgACZMmIDdu3cjJSUFcXFxePPNNzUSJ6lwCKyF1a4Ey2APEBER/ZPfOKDbQzWrvYqza+b8eAbrrOcHAKytrRETE4Pw8HAUFhbC09MTn376KcaMGYPIyMh73vvdd99h9uzZGDJkCJydnfH+++/jrbfeUr8vCAKioqKwZMkSzJo1Czdu3ED79u0xePBgODtrfz5TYwliQ9a6GZjCwkLY2NigoKAA1tbWWn32/sQcPP3dcfi6WGPnK9pdzkhERNIoLy9HcnIyvL29YWpqKnU4bdq9vtaN+f3NIbAWVjsHiD1ARERE0mEC1MI63B4CKyirQklFtcTREBERGSYmQC3MytQYViY1U6+4FJ6IiEgaTIAkULsSjJshEhERSYMJkAS4EoyIiEhaTIAkoO4BKmAPEBERkRSYAEmggw17gIiIiKTEBEgC6t2gOQmaiIhIEkyAJOBiw0nQRETUNnh5eSE8PFz9WhAE/Prrr5LF01CSJ0ARERHq3RwDAwMRGxt717ozZ86EIAh1ru7du9dbf/PmzRAEARMmTNBR9E3javv3ZojciJuIiGopVUoczzqOqKQoHM86DqVKKXVIbZakZ4FFRkZi3rx5iIiIwMCBA7F69WqMGTMGFy5cgIeHR536n3/+OT766CP16+rqavTq1QuPPfZYnbrXrl3DwoULERKif8dNONuYAAAqqlXIK6mEg6WJxBEREZHU9l7bi4+OfYTs0mx1mbO5Mxb1W4SRniMljKxtkrQHaOXKlXjmmWcwe/Zs+Pr6Ijw8HO7u7li1alW99W1sbNC+fXv1deLECdy6dQtPP/20Rj2lUolp06bh3XffRceOHVuiKY1iYiRHO6uapCeTK8GIiAze3mt7seDAAo3kBwBySnOw4MAC7L22Vyefu3r1ari6ukKlUmmUjxs3Dk899RSuXr2K8ePHw9nZGZaWlujbty/27m1cLOnp6ZgyZQrs7Ozg4OCA8ePHIyUlBQAQExMDY2NjZGVladzz6quvYvDgwc1q2/1IlgBVVlYiPj4eoaGhGuWhoaGIi4tr0DPWrVuHkSNHwtPTU6N86dKlaNeuHZ555pkGPaeiogKFhYUal67VrgRL50owIiKDplQp8dGxjyCi7pSI2rKPj32sk+Gwxx57DLm5udi/f7+67NatW9i9ezemTZuG4uJijB07Fnv37sWpU6cwatQohIWFITU1tUHPLy0txbBhw2BpaYmYmBgcOnQIlpaWGD16NCorKzF48GB07NgRP/zwg/qe6upq/Pjjj3U6N7RNsgQoNzcXSqUSzs7OGuXOzs51MsH6ZGZmYufOnZg9e7ZG+eHDh7Fu3TqsWbOmwbEsW7YMNjY26svd3b3B9zaVeiUYEyAiIoN2MudknZ6ffxIhIqs0CydzTmr9s+3t7TF69Gj89NNP6rJffvkF9vb2GDFiBHr16oXnn38ePXr0QOfOnfH++++jY8eO2LFjR4Oev3nzZshkMqxduxY9evSAr68vvvvuO6SmpuLAgQMAgGeeeQbfffed+p7ff/8dpaWlmDx5slbbeifJJ0ELgqDxWhTFOmX12bBhA2xtbTUmOBcVFeHJJ5/EmjVr4Ojo2OAYFi9ejIKCAvWVlpbW4HubSr0SjENgREQG7UbpDa3Wa6xp06Zh69atqKioAABs3LgRjz/+OORyOUpKSvD666/Dz88Ptra2sLS0xMWLFxvcAxQfH48rV67AysoKlpaWsLS0hL29PcrLy3H16lUANQucrly5gj///BMAsH79ekyePBkWFhY6aW8tySZBOzo6Qi6X1+ntycnJqdMrdCdRFLF+/XpMnz4dCoVCXX716lWkpKQgLCxMXVY7rmlkZITExER06tSpzvNMTExgYtKyE5E78DgMIiIC0M68nVbrNVZYWBhUKhV+//139O3bF7GxsVi5ciUA4LXXXsPu3buxYsUK+Pj4wMzMDI8++igqKysb9GyVSoXAwEBs3Lixznvt2tW0x8nJCWFhYfjuu+/QsWNHREVFqXuHdEmyBEihUCAwMBDR0dGYOHGiujw6Ohrjx4+/570HDx7ElStX6szx6datG86ePatR9uabb6KoqAiff/55iwxtNVSHfyyFJyIiwxXgFABnc2fklObUOw9IgABnc2cEOAXo5PPNzMwwadIkbNy4EVeuXEGXLl0QGBgIAIiNjcXMmTPVv6eLi4vVE5gbIiAgAJGRkXBycoK1tfVd682ePRuPP/443Nzc0KlTJwwcOLBZbWoISYfAFixYgLVr12L9+vVISEjA/PnzkZqaijlz5gCoGZqaMWNGnfvWrVuH/v37w9/fX6Pc1NQU/v7+GpetrS2srKzg7++v0Vsktb93g+YQGBGRIZPL5FjUbxGAmmTnn2pfv9HvDchlcp3FMG3aNPz+++9Yv349nnzySXW5j48Ptm3bhtOnT+PMmTOYOnVqnRVj93uuo6Mjxo8fj9jYWCQnJ+PgwYN45ZVXcP36dXW9UaNGwcbGBu+//77OJz/XkjQBmjJlCsLDw7F06VL07t0bMTExiIqKUq/qyszMrDPOWFBQgK1btzZ4hZe+qh0Cyy4sR7Wy4T9MRETU9oz0HImVQ1fCydxJo9zZ3Bkrh67U+T5Aw4cPh729PRITEzF16lR1+WeffQY7OzsEBwcjLCwMo0aNQkBAw3uizM3NERMTAw8PD0yaNAm+vr6YNWsWysrKNHqEZDIZZs6cCaVSWW/Hhy4IIrcirqOwsBA2NjYoKCi4Z5ddc6hUIrq+tRNVShGHFw1X7w5NREStT3l5OZKTk9UnGzSVUqXEyZyTuFF6A+3M2yHAKUCnPT/65Nlnn0V2dvZ9V5jd62vdmN/fku4EbchkMgEuNmZIzStFRn4ZEyAiIoJcJkff9n2lDqNFFRQU4Pjx49i4cSN+++23FvtcJkAScrExVSdAREREhmj8+PE4duwYnn/+eTz44IMt9rlMgCT096GonAhNRESGqSWWvNdH8o0QDZnL7YnQmQXsASIiImpJTIAkxL2AiIiIpMEESEIdbDgERkTUlnBhte5p62vMBEhC6h4gDoEREbVqxsbGAGpOPyfdqj2GQy5v3vYAnAQtodo5QPmlVSitrIa5gt8OIqLWSC6Xw9bWFjk5OQBqNgBsyMHe1DgqlQo3btyAubk5jIya9zuTv3ElZG1qDCsTIxRVVCMjvxw+TpZSh0RERE3Uvn17AFAnQaQbMpkMHh4ezU4wmQBJzMXWFEXZxcgsKGMCRETUigmCABcXFzg5OaGqqkrqcNoshUIBmaz5M3iYAEmsg60ZLmUXcyUYEVEbIZfLmz0/hXSPk6Al5sKVYERERC2OCZDEXG9PhGYPEBERUcthAiSx2h6gdCZARERELYYJkMQ6O9dMfD6fUcgNtIiIiFoIEyCJdWtvDRMjGQrKqpCcWyJ1OERERAaBCZDEFEYy9HC1AQCcTM2XNhgiIiIDwQRIDwR42gEATqXekjgSIiIiw8AESA/0cbcFAJxiDxAREVGLYAKkB/p41PQAXcwqRGlltcTREBERtX1MgPRAextTuNiYQiUCf10vkDocIiKiNo8JkJ7o42ELgMNgRERELYEJkJ7o486J0ERERC2FCZCeUPcApeVzQ0QiIiIdYwKkJ/xdbWAkE3CjqILHYhAREekYEyA9YWosR/cO1gA4D4iIiEjXmADpkdrl8EyAiIiIdIsJkB6pnQd0khOhiYiIdIoJkB6pXQl2IaMQFdVKiaMhIiJqu5gA6RF3ezM4WChQqVThfEah1OEQERG1WZInQBEREfD29oapqSkCAwMRGxt717ozZ86EIAh1ru7du6vrrFmzBiEhIbCzs4OdnR1GjhyJY8eOtURTmk0QBG6ISERE1AIkTYAiIyMxb948LFmyBKdOnUJISAjGjBmD1NTUeut//vnnyMzMVF9paWmwt7fHY489pq5z4MABPPHEE9i/fz+OHDkCDw8PhIaGIj09vaWa1Sx/T4TmPCAiIiJdEUQJd93r378/AgICsGrVKnWZr68vJkyYgGXLlt33/l9//RWTJk1CcnIyPD09662jVCphZ2eHr776CjNmzGhQXIWFhbCxsUFBQQGsra0b1hgtibuai6lrjsLV1gyHFw1v0c8mIiJqzRrz+1uyHqDKykrEx8cjNDRUozw0NBRxcXENesa6deswcuTIuyY/AFBaWoqqqirY29vftU5FRQUKCws1Lqn0dLOFTADS88uQU1guWRxERERtmWQJUG5uLpRKJZydnTXKnZ2dkZWVdd/7MzMzsXPnTsyePfue9RYtWgRXV1eMHDnyrnWWLVsGGxsb9eXu7t6wRuiApYkRujhbAag5FoOIiIi0T/JJ0IIgaLwWRbFOWX02bNgAW1tbTJgw4a51li9fjk2bNmHbtm0wNTW9a73FixejoKBAfaWlpTU4fl2onQd0IiVP0jiIiIjaKskSIEdHR8jl8jq9PTk5OXV6he4kiiLWr1+P6dOnQ6FQ1FtnxYoV+PDDD7Fnzx707Nnzns8zMTGBtbW1xiWl4E4OAICNR1Nx7WaJpLEQERG1RZIlQAqFAoGBgYiOjtYoj46ORnBw8D3vPXjwIK5cuYJnnnmm3vc/+eQTvPfee9i1axeCgoK0FnNLeaiHC/p726O0UomFv5yBUsXT4YmIiLRJ0iGwBQsWYO3atVi/fj0SEhIwf/58pKamYs6cOQBqhqbqW7m1bt069O/fH/7+/nXeW758Od58802sX78eXl5eyMrKQlZWFoqLi3XeHm2RyQSseKwXLBRyHE+5hbWxSVKHRERE1KZImgBNmTIF4eHhWLp0KXr37o2YmBhERUWpV3VlZmbW2ROooKAAW7duvWvvT0REBCorK/Hoo4/CxcVFfa1YsULn7dEmd3tz/DusZoPHT/dcwsUs7gxNRESkLZLuA6SvpNwH6J9EUcSz35/A3oQc+LlY49e5A6EwknzeOhERkV5qFfsA0f0JgoAPJ/WAnbkxLmQW4ot9l9XvKVUicgrLkXqzVMIIiYiIWif2ANVDX3qAau08m4kXNp6ETAC6d7BBTlE5bhRVoHZu9LyRnTFvZBdpgyQiIpIYe4DamDE9XDCxjytUInA2vQDZhTXJT+12SeF7L+Ono/Wfn0ZERER1GUkdADXMskk9MKybE8yN5XC2NoWTtQkcLBT4fN9lfPnHFbz561k4WZlgpN+991AiIiIiDoHVS9+GwO5FFEW8vuUv/BJ/HabGMvz07AMIuL2TNBERkSHhEJgBqZ0oPbRrO5RXqfDMhuO4eqP17HlEREQkBSZAbYCxXIaIaQHo5WaDW6VVeGr9MdwsrpA6LCIiIr3FBKiNMFcYYd3MvvByMMf1W2X4ZHei1CERERHpLSZAbYijpQk+ndwLABB5Ig0XMpq/e3S1UgUVzyIjIqI2hglQGxPoaY+HerpAFIH3f7+A5sxxT8srxeDl+zH+68OoqFZqMUoiIiJpMQFqgxaN7gaFXIa4qzfxx8WcJj2jpKIaz35/AhkF5TibXoC1sclajpKIiEg6TIDaIHd7czw9yAsA8EFUAqqUqkbdr1KJWPDzaVzMKoLJ7bPHvvrjCtLzy7QdKhERkSSYALVRc4f5wMFCgaQbJdj457VG3Ru+7zJ2n8+GQi7DT8/2R18vO5RVKfHB7xd0FC0REVHLYgLURlmbGmP+gzXng4Xvu4yC0qoG3ff7X5nqQ1ffn+iPQE97LB3vD7lMQNTZLMRevqGzmImIiFoKE6A27PG+7ujibIn80ip8+cfl+9Y/l16AV385DQB4ZpA3Jge5AwB8Xawx/QFPAMC/d5xHZXXjhtSIiIj0DROgNsxILsOSh/wAAP85koK0vNK71q1WqvDCxniUV6kwuEs7LB7TTeP9+Q92gaOlCZJulGDdIf2dEK1SiVy2T0RE98UEqI0b0qUdgjs5oEopYtOxu58Yvz/xBtLyyuBgocCXj/eBkVzzR8PGzFidFH35x2VkFpRBqRKRX1qJ1JuluJBRiPIqaZfKn07Lx8jPDqL/sn34/kgKe6qIiOiueBq8AXjyAU/EXb2JLfHXseDBLnWSGwCIPJ4GAHgk0A025sb1PmdSgCs2HUvFiWu3MGT5AVTesbpsSJd22PB0XwiCoP1G3INSJeKbg1fxWfQlVN/u/Xn7t/NYG5uMV0O7IKxnB8hkLRsTERHpN/YAGYCRvs6wt1Agp6gCBxLrTmLOKSzH/sSa/YImB7nd9TmCIGDpeH+YGMk0kh9zhRwyATh46QZiLudqvwH3kJFfhqlr/sQnuxNRrRLxUE8XvDuuO9pZmSA1rxSvbD6Nh748hJhLnLxNRER/Yw+QAVAYyTCpjyvWHkrG5uNpGOnnrPH+tlPpUKpEBHjYwsfJ6p7P8utgjUNvDEdxRTWsTY1gbWYMY7kM7//vAtYeSsbHOy8ixMexRXpc9iVkY8HPZ1BQVgVzhRzvjuuORwPdIAgCHgtyw3eHU/DNgatIyCzEjPXH8EiAG95+2O+uPVz6oFqpwpLt55BbXIF/h3WHh4O51CEREbVJ7AEyEFP61qzo2p+Yg5zCcnW5KIr4+fbwV+2qr/tpZ2UCb0cLOFiawPj2cNrcYT6wMjHChcxC/PevDC1HX9fFrEK8sPEkCsqq0MvNBr+/HILHgtzVw2/mCiPMHeaDmNeHYWawFwQB2HryOkZ+dhB7zmfpPL6m+mRPIiJPpGHfxRyM/SIWv51Ob9T9zTn6pDWpVqqwLyEbhy7n4lJ2EfJLKw2m7USkHewBMhCdna0Q4GGLk6n52HoyHS8M7QQAOHHtFpJyS2CukOPhXh2a/Hw7CwXmDO2ET3Yn4pPdiRjt3x4mRnJtha+hvEqJVzadRmW1CkO7tsOaGUHqRKy+uN4Z1x1hvTrg9S1ncPVGCZ77IR5hvTrgnTA/OFiaNPhzlSoRKTdLkHSjBEYyAWYKOcxvX7bmCjg24ln1iTqbidUHkwAAXZwtcSm7GK9sPo1Dl3PxzrjusDC5+x/X4opqfBuThO8OJaO3hy0+eqQnXG3NmhWPvhJFEa9t+QvbT2kmhwojGRwsFJAJAlSiCJUoQqkCZAJgZ66Ag6UC9hY13ydna1MEeNiil7stTI01f07zSipxIDEHBxJvwMJEjvkju8DJ2rQlm0hELUAQ+c+mOgoLC2FjY4OCggJYW1tLHY7W/Hw8Da9v/Qvejhb449UhEAQBr/1yBr/EX8djgW745LFezXp+WaUSQz7Zj5yiCvw7zA9PD/TWUuSa3tlxHhviUuBoaYJd80IanHiUVykRvvcyvo25CpUImBrLMLGPK54K9kK39prfZ1EUcfVGMY5cvYnzGYVIyCxEYnYRyqvuvrJsfO8OePvhxiVVtS5nF2HC14dRUqnE84M74rVRXfHFH1fw1R+XoRKBjo4WeHd8d/TxsIPlPxKhymoVfjp6DV/+cQU3SyrV5VamRnhvvD/G9+7Q4pPSdW3lnkR88ccVyGUCOrWzQE5RBfIbuNHnnRRyGXq526Cvlz0sTIzwx8UcnEy9hX/+rWhvocDHj/TEg3cMHROR/mnM728mQPVoqwlQSUU1+n2wFyWVSkQ+9wC6u9qg3wd7UVqpxC9zBqCvl32zP+Ono6n4v+1nYW+hwMHXhsLK9O/5NgVlVTienIdBnR3r/Kv7n86k5ePX0+mY1t+jzpyk/Rdz8PSG4wCADU/3xdCuTo2O8UxaPt789RzOpheoywZ0dMCMAZ6oqFYh9nIuDl/JRdY/hgprmRrL4ONkCQECSiurUVapRGmVEgVlVRBFwM7cGG8+5IdJAa4NTjyKyqsw/qvDSMotQXAnB3w/q596pd6fSTcxb/NpjVg8Hczh52INb0cL/PevDKTl1ZzR5u1ogecHd8Tm42k4nZYPAHi4pwven+APW3OF+n6lSkRpZbXG9+ZOoihiz4VsnErNx5whHTXul9IvJ9Lw2pa/AAAfP9IDU/p6AKhJbm8UVSCvpBKCAMgEQf1fpUrErdJK5JVU4mZxJW6WVCAltxTHUvJwo6ii3s/xdbHG8G7t8MfFG0jILAQAPPmAB5aM9YOZQjc9m0TUfEyAmqmtJkAAsGjrX9h8PA2T+riin7c9Fm07i46OFth3u0eouaqVKoR+FoOk3BK8PNwHC0K7IqugHOsOJeGno6koqVRiXK8O+OKJPvXeX1pZjeErDiKrsBxGMgGzQzri5RE+MFcY4UZRBcZ8HoPc4ko8PdAL/w7r3uQ4RVHE8ZRb2BCXjN3ns6GsZ/NEhZEMQZ526ONhC18Xa/i6WMPLwQLyeiZ4n0nLx6JtZ9W/LEM6O+KDCT3uO4lZpRIx58d47LmQjQ42pvjvS4Pq9CDdKqnEh1EJiL1cf1LWzsoE80Z2xuQgdxjLZahWqhBx4Co+33cZSpUIZ2sT+LlYI6eoAjlFFbhZXAGVCPR0s8GTD3girGcH9S91URQRczkXn+5JxF/XaxLEkb5OWDMjSPKepMNXcvHU+mOoVomYO6wTXhvV7f433YMoiki5WYrjyXk4mpyHkopqDOrsiOHdnNDh9vBhRbUSK3YnYk1szeafPk6WeOthP3R0tICztSkURpxGSaRPmAA1U1tOgE6l3sLEiDiYGsvg7WiJhMxCvDG6m3pOkDbsOpeJOT+ehJmxHGN7uGDHmXRUKTV/zDY9+wAGdHKoc+/K6Ev4Yt9lKIxk6o0MXW3N8NbDfth8PBUHEm+gW3sr/Dp34D17kRojI78MP/55Db+dzoCtuTEGdXbEIB9H9PWyb9RnVClVWBObhM/3XkZFtQqmxjJ8Oz0Ig7u0u+s9X/1xGSv2XIJCLsMvcwagl7vtPT8jr6QSCZk1Q3KXsovQqZ0lpg/whLmi7vygM2n5mB95Gkm5Jfd8prWpER4NdEf/jvZYdygZx5LzANRsb1ClVKFKKWr0tmhLQVkVPt97GbnFFXC3N4O7nTnc7c3hbmcOJ2sTja99YlYRHl0Vh6KKaozr1QHhU3q36N5OsZdv4NWfzyDnHz1GggA4Wpqgg40p+nnbY9Ygb7jYtM15V0StBROgZmrLCZAoihgVHoNL2cUAALlMwJHFw+Fkpb1JnqIoYmJEnHoYBgD6edljztCO2JeQg41HU9HV2Qq/vzxIY1PG9PwyDF9xABXVKkRMC4CxXIZ3dpxHen6Zuo6JkQz/fWkQujjfe7m+lJJzS7Bo6184mpwHC4Uckc8PgL+rTZ16m46lYvG2swCAjyb1wOP9tJtgADXzsv77VwZEUYSTlSnaWZnAycoEEIBtJ9Ox8eg19RBaLYWRDNMf8MQLQzthS/x1fLTzIiwUcux8ZbDWluVfySnGs9+fQPI9kjNzhRz2FjUTlzPyy5BbXIl+Xvb4YXY/nU2wv5e8kkp88HsCjqfkIaugvM5GoMZyARP7uGLOkE7o2M6yxeMjIiZAzdaWEyAAWHcoGe/97wKAmk0S1z4VpPXPOJOWjxc3noRfB2vMGdIJgZ52AGqGc4Z9egD5pVV1Jkq/vOkUdpzJQD9ve0Q+9wAEQUBZpRJf77+C1TFXUaUUsXR8d8wY4KX1eLWtslqFpzccw+ErN+FoaYLtLwbD3f7v5OG30+mYF3kaogg8P7gjFo/1lSROlUrEwcs3sPHPazidlo8H/drj5RE+6p4MpUrEE9/+iWMpeejrZYfNzw2oMwQoiiLyS6tq5tzIBMgEATIBMDWS19tL88fFbLyy6TSKKqrhamuGaQ94IDO/HGm3SpGaV4rrt8rqPcako6MFtr4QDDsL6ecjiaKImyWVyMwvR8rNEvz45zUcvd1zJgjAGP/26Olmi5KKahRXVKO4vBrVKhGPBboh2MdR4uiJ2i4mQM3U1hOgvJJK9P9wL6qUIr6dHojQ7u1b9PM3Hr2GJdvPwcrUCPsXDoWjpQnir+XhkVVHIAjAf/81qE6PSerNUqTdKkVwJwfJ56I0VFF5FSav/hMJmYXo6GiBLS8Ew95CgegL2ZjzYzyUKhFPPuCB98b763Wb0vJKMTo8BiWVSiwa0w1zhtQMl6pUIv77VwbC916utyfH0sQID3R0wCAfBwzq7IhO7Syx6uBVfLI7EaJY0ysY8WRAnVV8oiiisLwat0oqcbOkErdKKlFYXoVBnR212lOpbfHXbmHVgSvYm5Bz1zrGcgGrpgXW2YyUiLSDCVAztfUECKjpgUi6UYKXR3Sud1KvLilVIsZ/fQjn0gsxOcgNH03qiYkRh3HmegGmBLnj40d7tmg8upRdWI5JEXFIzy9DgIctXhzqgxd/OonKahUm9XHFisd6tYpzyiKPp+KNrWehkMvw69yBSM0rxWfRl5CYXdTgZ1ibGqGwvBoAMLW/B94J694mJxEnZhVh49FrKK6ohqWJESxNjGBhYoRTqfnYm5ANY7mA1dMDMbwbkyAibWMC1EyGkABJrbbHBwBmBnthQ1wKLE1qeoTaWTVvQ0F9cyWnCI+sOoKCsr/3qhnV3RlfTw2o92BafSSKIp79/gT2JuRAIf/7LDgrUyM8P7gjngr2gpmxHCoR6k0Ik26UqLcUOJaSh8pqFYxkAt4Z1x1PPuApcYtaXrVShVciT+P3vzKhkMuwekYghjVhGwciurvG/P6W/G/fiIgIeHt7w9TUFIGBgYiNjb1r3ZkzZ0IQhDpX9+6ay6G3bt0KPz8/mJiYwM/PD9u3b9d1M6iRAj3t8UhAzcGrG+JSAAD/Gu7T5pIfAPBxssLap4LUvR0hnR3xxRN9Wk3yA9QchLtsUk84WChQqVTBXCHHv4b54NDrw/Gv4Z1hZWoMI7kMCiMZTI3lMFcYwd/VBi8M7YQfZ/fHX/8OxU/P9kfUKyEGmfwAgJFchvApvTHGvz0qlSo8/0M8DvKQXiLJSNoDFBkZienTpyMiIgIDBw7E6tWrsXbtWly4cAEeHnVXxBQUFKCs7O8VK9XV1ejVqxdeeuklvPPOOwCAI0eOICQkBO+99x4mTpyI7du34+2338ahQ4fQv3//BsXFHqCWkVNUjuErDqK4ohoe9uaIXjBYktU9LeVESs1+M08P9Kp32XprkJBZiMNXcjGhj2uzj/4wVFVKFf7100nsPp8NEyMZvnyiT4vPwyNqq1rNEFj//v0REBCAVatWqct8fX0xYcIELFu27L73//rrr5g0aRKSk5Ph6Vnzr8opU6agsLAQO3fuVNcbPXo07OzssGnTpgbFxQSo5WyNv46Pd13Eysm9MagzV8eQYaisVmHuTycRfSEbADBroDfeGNO1Tf8DgKgltIohsMrKSsTHxyM0NFSjPDQ0FHFxcQ16xrp16zBy5Eh18gPU9ADd+cxRo0bd85kVFRUoLCzUuKhlPBLohmNLRjL5IYOiMJLh66kBmHV7G4j1h5Px6KojSLnPppVEpD2SJUC5ublQKpVwdtZcCeHs7IysrKz73p+ZmYmdO3di9uzZGuVZWVmNfuayZctgY2Ojvtzd3RvREiKixlMYyfB2mB/WzgiCrbkxzqYX4KEvYrH91HWUVlaD61OIdEvyiQh37n8iimKD9kTZsGEDbG1tMWHChGY/c/HixViwYIH6dWFhIZMgImoRI/2csfOVELyy6TSOpeRhfuQZAGdgLBdgY6aAjZkRXO3MsWh0N/h14JA8kbZI1gPk6OgIuVxep2cmJyenTg/OnURRxPr16zF9+nQoFJq7wrZv377RzzQxMYG1tbXGRUTUUlxszPDTs/3xyojOML99MG2VUkRucQWu3ihBzKUbmPLtEfU5bUTUfJIlQAqFAoGBgYiOjtYoj46ORnBw8D3vPXjwIK5cuYJnnnmmznsDBgyo88w9e/bc95lERFIykssw/8EuOP/uKJx/dxTiFg1H1Msh+Gl2f/TzskdReTWmrzuKfQnZUodK1CZIOgS2YMECTJ8+HUFBQRgwYAC+/fZbpKamYs6cOQBqhqbS09Px/fffa9y3bt069O/fH/7+/nWe+corr2Dw4MH4+OOPMX78ePz222/Yu3cvDh061CJtIiJqDkEQYHF79+gOtjVnsgV42uFfP53E3oQcPPdDPJY/0hOPBLpJHClR6ybpTmxTpkxBeHg4li5dit69eyMmJgZRUVHqVV2ZmZlITU3VuKegoABbt26tt/cHAIKDg7F582Z899136NmzJzZs2IDIyMgG7wFERKRvTI3lWPVkICYFuEKpEvHqL2ew7lCy1GERtWo8CqMe3AeIiPSRSiXi/d8TsP5wTfLzRD8P/DvMD6bG3D+ICGgl+wAREVHjyGQC3nrYF2+M7gZBADYdS8Ujq+KQerNU6tCIWh0mQERErYggCHhhaCd8P6sf7C0UOJ9RiIe+jFXvKk1EDcMEiIioFQrp3A7/e2kQAjxsUVRejWe/P4EPoxJQWlktdWhErQITICKiVqqDrRk2PzdAfaTGtzFJGPLJAWw8eg1VSpXE0RHpN06CrgcnQRNRa7P7fBbe+98FXL9VBgDo6GiBhaO6Yox/+wbtrk/UFrSa0+D1FRMgImqNKqqV+OloKr784wrySioBAK62ZnC0VMDazBhWpkawNjWGj5MlJvZxhYOlicQRE2kXE6BmYgJERK1ZUXkV1sQmY21sEkorlfXWMZYLCPVrjyf6eSC4kwNkMvYSUevHBKiZmAARUVuQX1qJxKwiFJVXo7C8CkXl1bhVWon9F3Nw5nqBup67vRnmjejC3aWp1WMC1ExMgIiorTufUYDNx9Lw66l0FFXUrBxbOyMII/3ufRg1kT7jRohERHRP3TvY4L0J/ji2ZCSe6OcOAJj/82mk5JZIHBlRy2ACRERkwMwUcrw7zh+BnnYoKq/G8z/Ecy8hMghMgIiIDJzCSIaIaQFoZ2WCxOwiLNp6FpwdQW0dEyAiIoKztSm+nhoAI5mAHWcy8N3hFKlDItIpJkBERAQA6Odtj/8b6wsA+DAqAUeTbkocEZHuMAEiIiK1pwd6YXzvDqhWiXjqu2PYEn9d6pCIdIIJEBERqQmCgGWTeiCksyPKq1RY+MsZvL7lDMrusqEiUWvFBIiIiDSYK4yw4el+WPBgFwgC8POJ65jw9WFcvVEsdWhEWsMEiIiI6pDLBLw8ojM2PtMfjpY1q8PCvjyE306nSx0akVYwASIiorsK9nFE1CuD8EBHe5RWKvHK5tN489ezqKjmkBi1bkyAiIjonpysTPHjM/3xr2E+AIAf/0zFo6uOIPVmqcSRETUdEyAiIrovI7kMC0d1xXdP94WduTHOphfgoS9jsed8ltShETUJEyAiImqwYV2d8PvLIejjYYui8mo890M8vjucLHVYRI3GBIiIiBqlg60ZIp8bgJnBXgCAD35PwOm0fEljImosJkBERNRoCiMZ/h3mh4d6uqBaJeLlTadQVF4ldVhEDcYEiIiImkQQBHw4sQfc7MyQmleKJdvP8RBVajWYABERUZPZmBnj88f7QH77EFUenUGtBRMgIiJqlkBPOyx4sAsA4O3fznPHaGoVmAAREVGzzRnSCcGdHFBWpcRLP53iRomk95gAERFRs8llAj6b0hv2FgpcyCzE1/uvSh0S0T0xASIiIq1wtjbF0vHdAQDrDyXjVkmlxBER3Z3kCVBERAS8vb1hamqKwMBAxMbG3rN+RUUFlixZAk9PT5iYmKBTp05Yv369Rp3w8HB07doVZmZmcHd3x/z581FeXq7LZhAREYCx/i7wc7FGcUU11sQmSR0O0V1JmgBFRkZi3rx5WLJkCU6dOoWQkBCMGTMGqampd71n8uTJ2LdvH9atW4fExERs2rQJ3bp1U7+/ceNGLFq0CP/+97+RkJCAdevWITIyEosXL26JJhERGTSZTMC8kZ0BABviUpDHXiDSU4Io4aYN/fv3R0BAAFatWqUu8/X1xYQJE7Bs2bI69Xft2oXHH38cSUlJsLe3r/eZ//rXv5CQkIB9+/apy1599VUcO3bsvr1LtQoLC2FjY4OCggJYW1s3slVERIZNFEWEfXUI59IL8fyQjlg8xlfqkMhANOb3t2Q9QJWVlYiPj0doaKhGeWhoKOLi4uq9Z8eOHQgKCsLy5cvh6uqKLl26YOHChSgrK1PXGTRoEOLj43Hs2DEAQFJSEqKiovDQQw/dNZaKigoUFhZqXERE1DSCIGD+yJpl8d/HJWNv8mFEJUXheNZxKFVcHUb6wUiqD87NzYVSqYSzs7NGubOzM7Ky6j9dOCkpCYcOHYKpqSm2b9+O3NxcvPjii8jLy1PPA3r88cdx48YNDBo0CKIoorq6Gi+88AIWLVp011iWLVuGd999V3uNIyIycMO7OcHHKwmZRpGYH1OgLnc2d8aifosw0nOkhNER6cEkaEEQNF6LolinrJZKpYIgCNi4cSP69euHsWPHYuXKldiwYYO6F+jAgQP44IMPEBERgZMnT2Lbtm343//+h/fee++uMSxevBgFBQXqKy0tTXsNJCIyQPtS9yHb7FsIRgUa5TmlOVhwYAH2XtsrUWRENSTrAXJ0dIRcLq/T25OTk1OnV6iWi4sLXF1dYWNjoy7z9fWFKIq4fv06OnfujLfeegvTp0/H7NmzAQA9evRASUkJnnvuOSxZsgQyWd2cz8TEBCYmJlpsHRGR4VKqlPjo2EcAgDv/PStChAABHx/7GMPch0Euk0sQIZGEPUAKhQKBgYGIjo7WKI+OjkZwcHC99wwcOBAZGRkoLv57m/VLly5BJpPBzc0NAFBaWlonyZHL5RBFkYf0ERG1gJM5J5Fdmn3X90WIyCrNwsmcky0YFZEmSYfAFixYgLVr12L9+vVISEjA/PnzkZqaijlz5gCoGZqaMWOGuv7UqVPh4OCAp59+GhcuXEBMTAxee+01zJo1C2ZmZgCAsLAwrFq1Cps3b0ZycjKio6Px1ltvYdy4cZDL+S8NIiJdu1F6Q6v1iHRBsiEwAJgyZQpu3ryJpUuXIjMzE/7+/oiKioKnpycAIDMzU2NPIEtLS0RHR+Oll15CUFAQHBwcMHnyZLz//vvqOm+++SYEQcCbb76J9PR0tGvXDmFhYfjggw9avH1ERIaonXk7rdYj0gVJ9wHSV9wHiIio6ZQqJUZtHYWc0hyIqP9XTHvz9tj1yC7OASKtahX7ABERUdskl8mxqF/N1iMC7lzpW/PfN/q9weSHJMUEiIiItG6k50isHLoSTuZOGuVitQ3Krj+JysLuEkVGVINDYPXgEBgRkXYoVUqczDmJG6U30M68HQ6escBX+5PhYKHAnvmD4WDJLUhIexrz+1vSSdBERNS2yWVy9G3fV/26l6MKexNycTGrCG//dh5fTwuQMDoyZE0aAktLS8P169fVr48dO4Z58+bh22+/1VpgRETU9iiMZFjxWC/IZQJ+P5uJ/RdzpA6JDFSTEqCpU6di//79AICsrCw8+OCDOHbsGP7v//4PS5cu1WqARETUtvi72mDWQC8AwFf7r3CTWpJEkxKgc+fOoV+/fgCAn3/+Gf7+/oiLi8NPP/2EDRs2aDM+IiJqg54d3BEKIxnir93CseQ8qcMhA9SkBKiqqkp9dtbevXsxbtw4AEC3bt2QmZmpveiIiKhNcrIyxWOBNUcYfX3gqsTRkCFqUgLUvXt3fPPNN4iNjUV0dDRGjx4NAMjIyICDg4NWAyQiorbp+cGdIJcJiLl0A+fSC+5/A5EWNSkB+vjjj7F69WoMHToUTzzxBHr16gUA2LFjh3pojIiI6F48HMwR1tMFABBx4IrE0ZChafI+QEqlEoWFhbCzs1OXpaSkwNzcHE5OTve4U/9xHyAiopaRmFWEUeExEARg74Ih6NTOUuqQqBXT+VEYZWVlqKioUCc/165dQ3h4OBITE1t98kNERC2na3srjPR1higC33AuELWgJiVA48ePx/fffw8AyM/PR//+/fHpp59iwoQJWLVqlVYDJCKitu3FYZ0AANtPpSM9v0ziaMhQNCkBOnnyJEJCQgAAW7ZsgbOzM65du4bvv/8eX3zxhVYDJCKiti3Aww4DOjqgWiViTUyS1OGQgWhSAlRaWgorKysAwJ49ezBp0iTIZDI88MADuHbtmlYDJCKitq+2F2jz8VTkFldIHA0ZgiYlQD4+Pvj111+RlpaG3bt3IzQ0FACQk5PDScNERNRog3wc0cvNBuVVKkTs51wg0r0mJUBvv/02Fi5cCC8vL/Tr1w8DBgwAUNMb1KdPH60GSEREbZ8gCHg1tCsA4Mc/r3EuEOlckxKgRx99FKmpqThx4gR2796tLh8xYgQ+++wzrQVHRESGI6SzIwZ0dEClUoXP916SOhxq45qUAAFA+/bt0adPH2RkZCA9PR0A0K9fP3Tr1k1rwRERkeEQBAGvja7pBdoSfx1XcookjojasiYlQCqVCkuXLoWNjQ08PT3h4eEBW1tbvPfee1CpVNqOkYiIDESAhx0e9HOGSgQ+3cNeINIdo6bctGTJEqxbtw4fffQRBg4cCFEUcfjwYbzzzjsoLy/HBx98oO04iYjIQLw2qiv2JmRj57ksnEnLRy93W6lDojaoSUdhdOjQAd988436FPhav/32G1588UX1kFhrxaMwiIikteDn09h2Mh2DfBzx4+z+UodDrYTOj8LIy8urd65Pt27dkJeX15RHEhERqc0f2QXGcgGHruTi8JVcqcOhNqhJCVCvXr3w1Vdf1Sn/6quv0LNnz2YHRUREhs3d3hzT+nsCAJbvuogmnttNdFdNmgO0fPlyPPTQQ9i7dy8GDBgAQRAQFxeHtLQ0REVFaTtGIiIyQHOH+eDnE2k4c70Au89nY7R/e6lDojakST1AQ4YMwaVLlzBx4kTk5+cjLy8PkyZNwvnz5/Hdd99pO0YiIjJA7axM8MwgbwDAp3sSoVSxF4i0p0mToO/mzJkzCAgIgFKp1NYjJcFJ0ERE+qGgrAqDl+9HQVkVVjzWC48GukkdEukxnU+CJiIiagk2ZsZ4YWjNQamfRV9CRXXr/gc26Q8mQEREpNeeGuAFJysTpOeXYfOxNKnDoTaCCRAREek1M4UcL4/oDAD48o8rKK2sljgiagsalQBNmjTpntf8+fMbHUBERAS8vb1hamqKwMBAxMbG3rN+RUUFlixZAk9PT5iYmKBTp05Yv369Rp38/HzMnTsXLi4uMDU1ha+vL1enERG1YpOD3OFhb47c4gp8dzhF6nCoDWjUMngbG5v7vj9jxowGPy8yMhLz5s1DREQEBg4ciNWrV2PMmDG4cOECPDw86r1n8uTJyM7Oxrp16+Dj44OcnBxUV//9r4HKyko8+OCDcHJywpYtW+Dm5oa0tDRYWVk1OC4iItIvCiMZFjzYBfMiT+Obg1cxrb8HbM0VUodFrZhWV4E1Vv/+/REQEIBVq1apy3x9fTFhwgQsW7asTv1du3bh8ccfR1JSEuzt7et95jfffINPPvkEFy9ehLGxcZPi4iowIiL9o1KJGPtFLC5mFWHOkE5YNKbuiQRk2FrFKrDKykrEx8cjNDRUozw0NBRxcXH13rNjxw4EBQVh+fLlcHV1RZcuXbBw4UKUlZVp1BkwYADmzp0LZ2dn+Pv748MPP7zn0vyKigoUFhZqXEREpF9kMgGvhnYFAGyIS8atkkqJI6LWTLIEKDc3F0qlEs7Ozhrlzs7OyMrKqveepKQkHDp0COfOncP27dsRHh6OLVu2YO7cuRp1tmzZAqVSiaioKLz55pv49NNP73lC/bJly2BjY6O+3N3dtdNIIiLSqpG+TvBzsUZ5lQo/n+CKMGo6yVeBCYKg8VoUxTpltVQqFQRBwMaNG9GvXz+MHTsWK1euxIYNG9S9QCqVCk5OTvj2228RGBiIxx9/HEuWLNEYZrvT4sWLUVBQoL7S0viHiohIHwmCgKeCa84I++HPa9wdmppMsgTI0dERcrm8Tm9PTk5OnV6hWi4uLnB1ddWYjO3r6wtRFHH9+nV1nS5dukAul2vUycrKQmVl/d2lJiYmsLa21riIiEg/jevlChszY1y/VYb9F3OkDodaKckSIIVCgcDAQERHR2uUR0dHIzg4uN57Bg4ciIyMDBQXF6vLLl26BJlMBjc3N3WdK1euQKVSadRxcXGBQsEVA0RErZ2ZQo4pfWumKvznSIq0wVCrJekQ2IIFC7B27VqsX78eCQkJmD9/PlJTUzFnzhwANUNT/1xWP3XqVDg4OODpp5/GhQsXEBMTg9deew2zZs2CmZkZAOCFF17AzZs38corr+DSpUv4/fff8eGHH2rMEyIiotbtyf6eEAQg9nIurt4ovv8NRHeQNAGaMmUKwsPDsXTpUvTu3RsxMTGIioqCp2fN+G5mZiZSU1PV9S0tLREdHY38/HwEBQVh2rRpCAsLwxdffKGu4+7ujj179uD48ePo2bMnXn75ZbzyyitYtGhRi7ePiIh0w8PBHMO7OgEAfjhyTeJoqDWSdB8gfcV9gIiI9N/BSzfw1PpjsDIxwpH/GwFLk0bt7UttUKvYB4iIiKg5Qnwc4e1ogaKKamw/lS51ONTKMAEiIqJWSSYTMP2BmikT38elgAMa1BhMgIiIqNV6JNAN5go5LucU48+kPKnDoQa4nF2Epf+9gO8lXsHHBIiIiFotGzNjTOzjCgD4ev8VlFRU3+cOklpSbgnWH07GrxIPWzIBIiKiVu2pYC/IBODQlVwM+eQANh1L5Q7Reqy0siZJNVdIO2mdCRAREbVqXZytsGZGEDwdzJFbXIHF285i7OexOHjphtShUT1KK2sOJzdTyO9TU7eYABERUas3wtcZ0fOH4M2HfGFjZozE7CI8tf4Y3tlxXurQ6A5ltxMgCyZAREREzacwkmF2SEccfG0onhnkDaDmqIzk3BKJI6N/Kqmo7QHiEBgREZHW2Jor8NbDfhjWtR1EEfjucLLUIdE/lFbVzAFiDxAREZEOzA7pCAD45cR15JdWShwN1Sq93QNkzgSIiIhI+4I7OcDXxRplVUpsPJp6/xuoRfw9CZpDYERERFonCAKeDbk9FyguBZXVKokjIgAoqx0CM2EPEBERkU483LMDnK1NkFNUgR1nMqQOh/CPSdDGTICIiIh0QmEkw1PBXgCAtbFJPC9MD6iXwZtwCIyIiEhnpvbzgJmxHBezinD4yk2pwzF4Jbd3guZGiERERDpka67A5CA3AMCa2CSJo6HaHiBzDoERERHp1qxB3hAE4OClG7iUXSR1OAatlENgRERELcPTwQKhfs4AuDGi1DgERkRE1IJqJ0PvPJeFaiWXxEtBFMV/nAXGHiAiIiKd6+dlD1tzY+SXVuHEtVtSh2OQKpUqVKtqVuKxB4iIiKgFGMllGN7NCQAQfSFb4mgMU23vD8CjMIiIiFrMg74184D2JmRzTyAJ1E6AVshlMJZLm4IwASIiIoMxuEs7KOQyXLtZiss5xVKHY3BK9WQCNMAEiIiIDIiFiRGCfRwAcBhMCuol8EyAiIiIWtaDt5fDMwFqeepzwJgAERERtayRt+cBnU7LR05hucTRGJa/T4KXdgk8wASIiIgMjLO1KXq52QAA9l3MkTgaw1I7BCb1SfAAEyAiIjJAHAaTRuntITCpl8ADTICIiMgAjbydAB26kqtemUS6V/u1NucQGBERUcvr6mwFd3szVFarEHMpV+pwDEaJnpwED+hBAhQREQFvb2+YmpoiMDAQsbGx96xfUVGBJUuWwNPTEyYmJujUqRPWr19fb93NmzdDEARMmDBBB5ETEVFrJQiCejI0h8FaTpmenAQPAJJGEBkZiXnz5iEiIgIDBw7E6tWrMWbMGFy4cAEeHh713jN58mRkZ2dj3bp18PHxQU5ODqqr63ZfXrt2DQsXLkRISIium0FERK3Qg37O+O5wCv64mA2lSoRcJkgdUpunngStB3OAJE2AVq5ciWeeeQazZ88GAISHh2P37t1YtWoVli1bVqf+rl27cPDgQSQlJcHe3h4A4OXlVaeeUqnEtGnT8O677yI2Nhb5+fm6bAYREbVCfb3sYWNmjFulVYi/dgv9vO2lDqnNU88BMuQhsMrKSsTHxyM0NFSjPDQ0FHFxcfXes2PHDgQFBWH58uVwdXVFly5dsHDhQpSVlWnUW7p0Kdq1a4dnnnmmQbFUVFSgsLBQ4yIiorbNWC7DsK7tAAC/nEiTOBrDUNsDpA+ToCWLIDc3F0qlEs7Ozhrlzs7OyMrKqveepKQkHDp0CKampti+fTtyc3Px4osvIi8vTz0P6PDhw1i3bh1Onz7d4FiWLVuGd999t8ltISKi1umxIHf8ejoDv8RfRw83G8wY4CV1SG2augdID4bAJJ8ELQiaY66iKNYpq6VSqSAIAjZu3Ih+/fph7NixWLlyJTZs2ICysjIUFRXhySefxJo1a+Do6NjgGBYvXoyCggL1lZbGfwkQERmCgT6OeG1UVwDAOzvOY19CNqBSAsmxwNktNf9VKSWOsu1Q9wDpQQIkWQ+Qo6Mj5HJ5nd6enJycOr1CtVxcXODq6gobGxt1ma+vL0RRxPXr11FSUoKUlBSEhYWp31epVAAAIyMjJCYmolOnTnWea2JiAhMTE200i4iIWpkXh3ZC6s1SRJ5Iw2+bvkGI1SYoSjL/rmDdARj9MeA3Trog24i/EyDph8Ak6wFSKBQIDAxEdHS0Rnl0dDSCg4PrvWfgwIHIyMhAcXGxuuzSpUuQyWRwc3NDt27dcPbsWZw+fVp9jRs3DsOGDcPp06fh7u6u0zYREVHrIwgC3p/oj/muFxEurITRP5MfACjMBH6eAVzYIU2AbQiHwG5bsGAB1q5di/Xr1yMhIQHz589Hamoq5syZA6BmaGrGjBnq+lOnToWDgwOefvppXLhwATExMXjttdcwa9YsmJmZwdTUFP7+/hqXra0trKys4O/vD4VCIVVTiYhIjxkLIl6qXAtBqO8Xo1jzn12LOBzWTBwCu23KlCm4efMmli5diszMTPj7+yMqKgqenp4AgMzMTKSmpqrrW1paIjo6Gi+99BKCgoLg4OCAyZMn4/3335eqCURE1BZci4OsKOMeFUSgMB24Fgd4c3+5ptKnITBBFEVR6iD0TWFhIWxsbFBQUABra2upwyEiIl07uwXY2oCtUx5ZB/R4VPfxtFHd3tqJ8ioVYl8fBnd7c60/vzG/vyVfBUZERCQ5y/oX3zS5HtWhVIkor6pZmKQPQ2BMgIiIiDyDa1Z7of5tWEQAorVrTT1qkrKqv+dP6cMQGBMgIiIimbxmqTuAO5MglQiIIrDd+aWaetQktSvABAEwNZY+/ZA+AiIiIn3gNw6Y/D1g7aJRXGbWHi9UzcOCsx747XS6RMG1fqUVtydAG8vvuuFxS5K+D4qIiEhf+I0Duj1Us9qrOBuwdIaFZzB8oi9j9/6rWHXgKsb16qAXv8BbG306BwxgDxAREZEmmbxmqXuPR2v+K5PjuZBOMDOW42JWEY4k3ZQ6wlaprEp/NkEEmAARERHdl425MR4NdAMArD+UIm0wrVRJhf7sAQQwASIiImqQmQO9AAD7LmYjJbdE2mBaIX3aBRpgAkRERNQgndpZYljXdhBFYENcitThtDr6dA4YwASIiIiowWYN8gYA/HIiDYXlVRJH07qwB4iIiKiVGuTjiC7OliipVOLn42lSh9OqlOnROWAAEyAiIqIGEwQBswbW9AJ9dzgF1UqVxBG1HiUcAiMiImq9JvRxhZ25MdLzy7A3IVvqcFqNMg6BERERtV6mxnJM6+8JAFh3KFniaFqP2h4gMw6BERERtU7TB3jCWC7geMotnL1eIHU4rULtJGgL9gARERG1Ts7WphjtX3NmGM8Haxj1WWBMgIiIiFqvh3vWJEA7z2VBFEWJo9F/pVVcBUZERNTqDenSDuYKOdLzy3CGw2D3VcZVYERERK2fqbEcw7s5AQB2ns2UOBr9V3sWmBkTICIiotZtbI+aYbCoc5kcBruPsttDYBYmHAIjIiJq1YZ1dYKZsRxpeWU4n1EodTh6raTi9jJ4Y/YAERERtWpmCjmGdWsHAIjiMNg91W6EyB4gIiKiNmDM7eXwUWc5DHY3oij+YxUYe4CIiIhaveHdnGBiJEPKzVIkZBZJHY5eqqhWQamqSQ45CZqIiKgNsDAxwtCuNcNgO89xGKw+tcNfAGDOOUBERERtQ+1qsN85DFav2nPAFEYyGMn1I/XQjyiIiIhaseHdnKCQy5B0owSXc4qlDkfvlOnZOWAAEyAiIqJmszI1xuAujgC4Gqw+tQeh6ssxGAATICIiIq2oHQbbeTZL4kj0T+0QmL5MgAaYABEREWnFCF9nGMsFJGYX4Vw6zwb7Jw6BERERtVE2ZsYY1b09AOCdHeehUnEydK2SSv06BwzQgwQoIiIC3t7eMDU1RWBgIGJjY+9Zv6KiAkuWLIGnpydMTEzQqVMnrF+/Xv3+mjVrEBISAjs7O9jZ2WHkyJE4duyYrptBRESE/xvrC3OFHCeu3cLPJ9KkDkdv1J4Eb8E5QDUiIyMxb948LFmyBKdOnUJISAjGjBmD1NTUu94zefJk7Nu3D+vWrUNiYiI2bdqEbt26qd8/cOAAnnjiCezfvx9HjhyBh4cHQkNDkZ6e3hJNIiIiA9bB1gwLHuwCAFi28yJyiyskjkg/lOphD5AgSrhhQf/+/REQEIBVq1apy3x9fTFhwgQsW7asTv1du3bh8ccfR1JSEuzt7Rv0GUqlEnZ2dvjqq68wY8aMBt1TWFgIGxsbFBQUwNraumGNISIiAlCtVGHcV4dxIbMQkwJcsXJyb6lDktzX+6/gk92JmBLkjo8f7amzz2nM72/JeoAqKysRHx+P0NBQjfLQ0FDExcXVe8+OHTsQFBSE5cuXw9XVFV26dMHChQtRVlZ2188pLS1FVVXVPROmiooKFBYWalxERERNYSSX4cNJPSAIwLaT6Yi7mit1SJIr5Sqwv+Xm5kKpVMLZ2Vmj3NnZGVlZ9S8hTEpKwqFDh3Du3Dls374d4eHh2LJlC+bOnXvXz1m0aBFcXV0xcuTIu9ZZtmwZbGxs1Je7u3vTGkVERASgt7stnuzvCQB4c/s5VFQr73NH21ZSoV8HoQJ6MAlaEASN16Io1imrpVKpIAgCNm7ciH79+mHs2LFYuXIlNmzYUG8v0PLly7Fp0yZs27YNpqamd41h8eLFKCgoUF9paZy4RkREzfPa6K5oZ2WCpNwSfHMgSepwJKVeBm/CSdBwdHSEXC6v09uTk5NTp1eolouLC1xdXWFjY6Mu8/X1hSiKuH79ukbdFStW4MMPP8SePXvQs+e9xxtNTExgbW2tcRERETWHtakx3nrYDwDw9YEruH6rVOKIpFNadXsStJ4chApImAApFAoEBgYiOjpaozw6OhrBwcH13jNw4EBkZGSguPjvc1YuXboEmUwGNzc3ddknn3yC9957D7t27UJQUJBuGkBERHQfYT1dMKCjAyqrVYg4cFXqcCRTWnF7GbwJEyAAwIIFC7B27VqsX78eCQkJmD9/PlJTUzFnzhwANUNT/1y5NXXqVDg4OODpp5/GhQsXEBMTg9deew2zZs2CmZkZgJphrzfffBPr16+Hl5cXsrKykJWVpZE0ERERtQRBEDD/9rL4X06kGWwv0N/L4DkEBgCYMmUKwsPDsXTpUvTu3RsxMTGIioqCp2fNxLHMzEyNPYEsLS0RHR2N/Px8BAUFYdq0aQgLC8MXX3yhrhMREYHKyko8+uijcHFxUV8rVqxo8fYRERH187bHQB8HVClFg+0Fql0FZq5HQ2CS7gOkr7gPEBERadOx5DxMXn0ExnIB+xcOhZududQhtagHVx7E5Zxi/PRsfwR3ctTZ57SKfYCIiIgMhaH3AtUOgZlzCIyIiMiwvDLCcOcClarPAtOfITAmQERERC2gsb1AucUVmLXhOML3XkJrn62ij2eBMQEiIiJqIQ3tBRJFEa/9cgZ/XMxB+N7L+PHPay0VotYpVSIqqlUAOARGRERkkBraC/T9kWvYn3hD/frd/15A/LW8lghR62qHvwAehUFERGSwanuBNh9LxQ/19OxczCrEB1EJAIB/h/nhoR4uqFaJmPPjSeQUlrdorNpQewyGTABMjPQn7dCfSIiIiAxAP297TH/AEyoReOvXc1i+66J6jk95lRKvbDqNymoVhnVth5nBXlj+aE90cbbEjaIKvLDxJCpvDye1FiW154ApjO561qcUmAARERG1sKXju2PB7R2iIw5cxYKfz6CyWoVlUQlIzC6Co6UJPnmsFwRBgIWJEVZPD4KVqRHir93Ce/+7IHH0jVM7BKZPE6ABJkBEREQtThAEvDyiMz55tCeMZAK2n0rH+K8P4z9HaobEVjzWE46WJur63o4WCJ/SGwDww5/XsOlYan2P1Ut/7wHEBIiIiIgAPBbkjnUz+8JCIUdCZiEAYNZAbwzt6lSn7ghfZ8wb2RkAsHjbWWw4nNyisTaVPm6CCDABIiIiktSQLu0Q+fwAeDtaILiTA14f3fWudV8e3hkzg70AAO/89wJWRuv/HkFlteeA6VkPkH6lY0RERAbI39UGf7w6BADuOVFYJhPw7zA/OFgo8Gn0JXyx7zLySyvxTlh3yGT6M8H4n0oqbvcAmehXysEeICIiIj0gCEKDVkkJgoCXRnTGexP8IQg1ewbNizytt6vDSqtuJ0B6dBI8wASIiIioVZr+gCe+eLwPjOUCdpzJwOtbzujlcFhphX4OgTEBIiIiaqXCenXAmhlBkMsE/Ho6A5uPp0kdUh3qSdAmTICIiIhIS4Z2dcLro2omTv97x3lcyCiUOCJNZVVcBUZEREQ68GxIRwzv5oTKahXm/nQSxRXV97+phZRwCIyIiIh0QSYT8OljvdDBxhTJuSX4v21n9WY+UBk3QiQiIiJdsbNQ4MupATCS1UyK/klPdosuUe8DxCEwIiIi0oFATzv1Rorv/vcC4q/dkjgiHoVBRERELWD2oI4YcXs+0BPf/okf/rwm6XAYh8CIiIhI52QyAeGP98aDfs6oVKrw1q/n8Mrm0+rJyC2thGeBERERUUuwMjXGt9MD8X9ju0F+e07Q+K8P43J2UYvHoq9ngTEBIiIiaoMEQcBzgzth83MPwNnaBFdyijHuq8M4ldqy84LYA0REREQtrq+XPX5/OQT9ve1RVqXEF/sut+jncw4QERERScLR0gQfP9ITAHDg0g2k5Ja0yOeKoohSDoERERGRVLwcLTC0azuIIvDjn9da5DMrqlVQ3V6AZm7CITAiIiKSwFMDvAAAP59IU/fM6FLtHkAAYGbMHiAiIiKSwJAu7eBhb47C8mr8djpD559Xu/Te1FgGuUzQ+ec1BhMgIiIiAyGTCZgxwBMA8J+4FJ1vkKivJ8EDepAARUREwNvbG6ampggMDERsbOw961dUVGDJkiXw9PSEiYkJOnXqhPXr12vU2bp1K/z8/GBiYgI/Pz9s375dl00gIiJqNR4LdIepsQwXs4pwPEW3S+KTbhQDAKxNmQBpiIyMxLx587BkyRKcOnUKISEhGDNmDFJT736A2+TJk7Fv3z6sW7cOiYmJ2LRpE7p166Z+/8iRI5gyZQqmT5+OM2fOYPr06Zg8eTKOHj3aEk0iIiLSazbmxpjYxxUA8J8jKTr9rLWxyQCAMT1cdPo5TSGIEh4Q0r9/fwQEBGDVqlXqMl9fX0yYMAHLli2rU3/Xrl14/PHHkZSUBHt7+3qfOWXKFBQWFmLnzp3qstGjR8POzg6bNm1qUFyFhYWwsbFBQUEBrK2tG9kqIiIi/XYhoxBjv4iFkUzA4UXD4WxtqvXPiL+Wh0dWHYFCLsOhN4bBSQefcafG/P6WrAeosrIS8fHxCA0N1SgPDQ1FXFxcvffs2LEDQUFBWL58OVxdXdGlSxcsXLgQZWVl6jpHjhyp88xRo0bd9ZlAzbBaYWGhxkVERNRW+XWwRj8ve1SrRPx09O6jLs2x+mASAGBiH9cWSX4aS7IEKDc3F0qlEs7Ozhrlzs7OyMrKqveepKQkHDp0COfOncP27dsRHh6OLVu2YO7cueo6WVlZjXomACxbtgw2Njbqy93dvRktIyIi0n8zgmsmQ/90LBWV1SqtPvvqjWJEJ2QDAJ4d7K3VZ2uL5JOgBUFzWZwoinXKaqlUKgiCgI0bN6Jfv34YO3YsVq5ciQ0bNmj0AjXmmQCwePFiFBQUqK+0tLRmtIiIiEj/jereHk5WJrhRVIFtJ69r9dlrY5MgisBIX2f4OFlp9dnaIlkC5OjoCLlcXqdnJicnp04PTi0XFxe4urrCxsZGXebr6wtRFHH9es03r3379o16JgCYmJjA2tpa4yIiImrLjOUyPDe4IwDgk92JKCir0spzbxRVYOvJdADA80M6auWZuiBZAqRQKBAYGIjo6GiN8ujoaAQHB9d7z8CBA5GRkYHi4mJ12aVLlyCTyeDm5gYAGDBgQJ1n7tmz567PJCIiMlQzBnihUzsL3CypRPjeS1p55n/iUlBZrUIfD1sEedpp5Zm6IOkQ2IIFC7B27VqsX78eCQkJmD9/PlJTUzFnzhwANUNTM2bMUNefOnUqHBwc8PTTT+PChQuIiYnBa6+9hlmzZsHMzAwA8Morr2DPnj34+OOPcfHiRXz88cfYu3cv5s2bJ0UTiYiI9JbCSIZ3xnUHAHx/5BoSs4qa9bySimr8cPucsecHd7rn9BOpSZoATZkyBeHh4Vi6dCl69+6NmJgYREVFwdOzZmJWZmamxp5AlpaWiI6ORn5+PoKCgjBt2jSEhYXhiy++UNcJDg7G5s2b8d1336Fnz57YsGEDIiMj0b9//xZvHxERkb4L6dwOo7o7Q6kS8c6O883aHTryeBoKyqrg7WiBB/3uPvVEH0i6D5C+4j5ARERkSNLySjFy5UFUVKvw9dQAPNSz8RsXVitVGPLJAaTnl+GDif6Y1t9TB5HeW6vYB4iIiIj0g7u9OeYM6QQA+OD3C006Kf50Wj7S88tgY2aMRwLctB2i1jEBIiIiIrwwtBNcbc2QUVCOVQeuNvr+2Mu5AIBBnR1haizXdnhaxwSIiIiIYGosx1sP+wIAVsckIS2vtFH3x16+AQAI8XHUemy6wASIiIiIANRsjjjQxwGV1Sp8uiexwfcVllfhzPUCADU9QK0BEyAiIiICUHOSwuIxNb1Av57OwPmMggbdd+TqTShVIjo6WsDNzlyXIWoNEyAiIiJS83e1wbheHQAAy3c1rBfo0D/m/7QWTICIiIhIw8LQrjCWCzh46QbiruTet37t/J9BrWT+D8AEiIiIiO7g4WCu3sfno10X77k5YlpeKVJulkIuEzCgk0NLhdhsTICIiIiojn8N94GFQo6/rhcg6mzWXesdut1D1MfdFlamxi0VXrMxASIiIqI6HC1N8Kz6tPiLqFKq6q3XGuf/AEyAiIiI6C5mh3SEo6UCKTdLEXk8rc77SpWo7gEKYQJEREREbYGliRFeGt4ZABC+9zIKSqs03j+XXoCCsipYmRihl5utBBE2HRMgIiIiuqsn+nnA29ECucUVeCXyFJSqvydE1/b+DOjkACN560opWle0RERE1KIURjJ8+UQfmBjJcCDxBj6LvqR+T338RSsb/gKYABEREdF9+Lva4ONHegIAvtp/BbvOZaKkohrx124BAEI6t5MyvCZhAkRERET3NaGPK54Z5A0AePXnM9h49BqqlCLc7Mzg6dA6jr/4JyZARERE1CCLx3TDAx3tUVKpxIdRFwHUDH8JgiBxZI3HBIiIiIgaxEguw9dTA9DBxlRdNsin9Q1/AUyAiIiIqBEcLE2wenoQTIxkMFfIMdCn9Rx/8U9GUgdARERErUsPNxv8/nIIlCoRtuYKqcNpEiZARERE1Gg+TpZSh9AsHAIjIiIig8MEiIiIiAwOEyAiIiIyOEyAiIiIyOAwASIiIiKDwwSIiIiIDA4TICIiIjI4TICIiIjI4DABIiIiIoMjeQIUEREBb29vmJqaIjAwELGxsXete+DAAQiCUOe6ePGiRr3w8HB07doVZmZmcHd3x/z581FeXq7rphAREVErIelRGJGRkZg3bx4iIiIwcOBArF69GmPGjMGFCxfg4eFx1/sSExNhbW2tft2u3d8n0W7cuBGLFi3C+vXrERwcjEuXLmHmzJkAgM8++0xnbSEiIqLWQ9IEaOXKlXjmmWcwe/ZsADU9N7t378aqVauwbNmyu97n5OQEW1vbet87cuQIBg4ciKlTpwIAvLy88MQTT+DYsWNaj5+IiIhaJ8mGwCorKxEfH4/Q0FCN8tDQUMTFxd3z3j59+sDFxQUjRozA/v37Nd4bNGgQ4uPj1QlPUlISoqKi8NBDD931eRUVFSgsLNS4iIiIqO2SrAcoNzcXSqUSzs7OGuXOzs7Iysqq9x4XFxd8++23CAwMREVFBX744QeMGDECBw4cwODBgwEAjz/+OG7cuIFBgwZBFEVUV1fjhRdewKJFi+4ay7Jly/Duu+/WKWciRERE1HrU/t4WRfH+lUWJpKeniwDEuLg4jfL3339f7Nq1a4Of8/DDD4thYWHq1/v37xednZ3FNWvWiH/99Ze4bds20d3dXVy6dOldn1FeXi4WFBSorwsXLogAePHixYsXL16t8EpLS7tv/iBZD5CjoyPkcnmd3p6cnJw6vUL38sADD+DHH39Uv37rrbcwffp09byiHj16oKSkBM899xyWLFkCmazuqJ+JiQlMTEzUry0tLZGWlgYrKysIggAA6Nu3L44fP65x351l93pd+/+FhYVwd3dHWlqaxkTupqgvpqbWvdv792vjnWW6bnND2tLQeg1tc31ld2tz7et9+/ZJ0ub71dVWm+983Rp/vg2xzXcrb+jPN/9M8890c7TEn+ljx46hqKgIHTp0uO9nSJYAKRQKBAYGIjo6GhMnTlSXR0dHY/z48Q1+zqlTp+Di4qJ+XVpaWifJkcvlEEWxYV1iAGQyGdzc3Oo8485v/p1l93p953vW1tbN/mGqL6am1r3b+/dr451lum7zvWJtbL2Gtrm+snu185+vW7rN96urrTbf+bo1/nwbYpvvVt7QdvLPNP9MN0dL/Jm2sbGBjY1Ngz5D0lVgCxYswPTp0xEUFIQBAwbg22+/RWpqKubMmQMAWLx4MdLT0/H9998DqFkl5uXlhe7du6OyshI//vgjtm7diq1bt6qfGRYWhpUrV6JPnz7o378/rly5grfeegvjxo2DXC5vcqxz5869b9m9Xtd3f3M15pn3q3u39+/XxjvLdN3mxjxXW22ur+xe7dTn77W22nzna31u893eN8Q23628oe3kn2nt4Z/p5tVt6J/pe2rwZBsd+frrr0VPT09RoVCIAQEB4sGDB9XvPfXUU+KQIUPUrz/++GOxU6dOoqmpqWhnZycOGjRI/P333zWeV1VVJb7zzjvqeu7u7uKLL74o3rp1q4VadG8FBQUiALGgoEDqUFoM22w4DLHdbLNhMMQ2i2LbbrfkCZChKS8vF//973+L5eXlUofSYthmw2GI7WabDYMhtlkU23a7BVFs4MQYIiIiojZC8rPAiIiIiFoaEyAiIiIyOEyAiIiIyOAwASIiIiKDwwSIiIiIDA4TID2VmJiI3r17qy8zMzP8+uuvUoelc8nJyRg2bBj8/PzUx5gYAiMjI/X3uvYYF0NQWloKT09PLFy4UOpQdK6oqAh9+/ZF79690aNHD6xZs0bqkFpEWloahg4dCj8/P/Ts2RO//PKL1CG1iIkTJ8LOzg6PPvqo1KHozP/+9z907doVnTt3xtq1a6UOp9G4DL4VKC4uhpeXF65duwYLCwupw9GpIUOG4P3330dISAjy8vJgbW0NIyNJNyxvEY6OjsjNzZU6jBa3ZMkSXL58GR4eHlixYoXU4eiUUqlERUUFzM3NUVpaCn9/fxw/fhwODg5Sh6ZTmZmZyM7ORu/evZGTk4OAgAAkJia2+b/L9u/fj+LiYvznP//Bli1bpA5H66qrq+Hn54f9+/fD2toaAQEBOHr0KOzt7aUOrcHYA9QK7NixAyNGjGjzf2GcP38exsbGCAkJAQDY29sbRPJjqC5fvoyLFy9i7NixUofSIuRyOczNzQEA5eXlUCqVDT6fsDVzcXFB7969AQBOTk6wt7dHXl6etEG1gGHDhsHKykrqMHTm2LFj6N69O1xdXWFlZYWxY8di9+7dUofVKEyAmigmJgZhYWHo0KEDBEGod3gqIiIC3t7eMDU1RWBgIGJjY5v0WT///DOmTJnSzIibT9dtvnz5MiwtLTFu3DgEBATgww8/1GL0TdcS3+vCwkIEBgZi0KBBOHjwoJYib7qWaPPChQuxbNkyLUXcfC3R5vz8fPTq1Qtubm54/fXX4ejoqKXom64l/y47ceIEVCoV3N3dmxl187Rkm/VVc78GGRkZcHV1Vb92c3NDenp6S4SuNUyAmqikpAS9evXCV199Ve/7kZGRmDdvHpYsWYJTp04hJCQEY8aMQWpqqrpOYGAg/P3961wZGRnqOoWFhTh8+LBe/CtZ122uqqpCbGwsvv76axw5cgTR0dGIjo5uqebdVUt8r1NSUhAfH49vvvkGM2bMQGFhYYu07W503ebffvsNXbp0QZcuXVqqSffVEt9nW1tbnDlzBsnJyfjpp5+QnZ3dIm27l5b6u+zmzZuYMWMGvv32W5236X5aqs36rLlfg/p6LwVB0GnMWiflORxtBQBx+/btGmX9+vUT58yZo1HWrVs3cdGiRY169vfffy9OmzatuSFqnS7aHBcXJ44aNUr9evny5eLy5cubHas26fJ7XWv06NHi8ePHmxqi1umizYsWLRLd3NxET09P0cHBQbS2thbfffddbYXcbC3xfZ4zZ474888/NzVEndBVu8vLy8WQkBDx+++/10aYWqXL7/X+/fvFRx55pLkh6lxTvgaHDx8WJ0yYoH7v5ZdfFjdu3KjzWLWJPUA6UFlZifj4eISGhmqUh4aGIi4urlHP0pfhr/vRRpv79u2L7Oxs3Lp1CyqVCjExMfD19dVFuFqjjXbfunULFRUVAIDr16/jwoUL6Nixo9Zj1RZttHnZsmVIS0tDSkoKVqxYgWeffRZvv/22LsLVCm20OTs7W92zV1hYiJiYGHTt2lXrsWqTNtotiiJmzpyJ4cOHY/r06boIU6u0+fd3a9WQr0G/fv1w7tw5pKeno6ioCFFRURg1apQU4TYZZ5jqQG5uLpRKJZydnTXKnZ2dkZWV1eDnFBQU4NixY9i6dau2Q9Q6bbTZyMgIH374IQYPHgxRFBEaGoqHH35YF+FqjTbanZCQgOeffx4ymQyCIODzzz/X65UU2vr5bk200ebr16/jmWeegSiKEEUR//rXv9CzZ09dhKs12mj34cOHERkZiZ49e6rnmfzwww/o0aOHtsPVCm39fI8aNQonT55ESUkJ3NzcsH37dvTt21fb4epEQ74GRkZG+PTTTzFs2DCoVCq8/vrrrW5FIxMgHbpzPFQUxUaNkdrY2OjFHIHGaG6bx4wZgzFjxmg7LJ1rTruDg4Nx9uxZXYSlU839XteaOXOmliLSvea0OTAwEKdPn9ZBVLrXnHYPGjQIKpVKF2HpVHN/vlvbiqj63O9rMG7cOIwbN66lw9IaDoHpgKOjI+RyeZ1/LeTk5NTJqNsKQ2wzYJjtZpv/1pbbDBhmuw2xzXcylK8BEyAdUCgUCAwMrLOCKTo6GsHBwRJFpVuG2GbAMNvNNv+tLbcZMMx2G2Kb72QoXwMOgTVRcXExrly5on6dnJyM06dPw97eHh4eHliwYAGmT5+OoKAgDBgwAN9++y1SU1MxZ84cCaNuHkNsM2CY7WabDaPNgGG22xDbfCd+DcBl8E21f/9+EUCd66mnnlLX+frrr0VPT09RoVCIAQEB4sGDB6ULWAsMsc2iaJjtZpsNo82iaJjtNsQ234lfA1HkWWBERERkcDgHiIiIiAwOEyAiIiIyOEyAiIiIyOAwASIiIiKDwwSIiIiIDA4TICIiIjI4TICIiIjI4DABIiIiIoPDBIiI2hwvLy+Eh4dLHQYR6TEmQETUJDNnzsSECROkDqNex48fx3PPPafzz/Hy8oIgCBAEAWZmZujWrRs++eQTNHaDfSZsRC2Ph6ESUatRVVUFY2Pj+9Zr165dC0RTY+nSpXj22WdRXl6OvXv34oUXXoC1tTWef/75FouBiBqPPUBEpBMXLlzA2LFjYWlpCWdnZ0yfPh25ubnq93ft2oVBgwbB1tYWDg4OePjhh3H16lX1+ykpKRAEAT///DOGDh0KU1NT/Pjjj+qepxUrVsDFxQUODg6YO3cuqqqq1Pfe2aMiCALWrl2LiRMnwtzcHJ07d8aOHTs04t2xYwc6d+4MMzMzDBs2DP/5z38gCALy8/Pv2U4rKyu0b98eXl5emD17Nnr27Ik9e/ao37969SrGjx8PZ2dnWFpaom/fvti7d6/6/aFDh+LatWuYP3++ujepVlxcHAYPHgwzMzO4u7vj5ZdfRklJSYO/B0R0d0yAiEjrMjMzMWTIEPTu3RsnTpzArl27kJ2djcmTJ6vrlJSUYMGCBTh+/Dj27dsHmUyGiRMnQqVSaTzrjTfewMsvv4yEhASMGjUKALB//35cvXoV+/fvx3/+8x9s2LABGzZsuGdM7777LiZPnoy//voLY8eOxbRp05CXlwegJtl69NFHMWHCBJw+fRrPP/88lixZ0qg2i6KIAwcOICEhQaOXqri4GGPHjsXevXtx6tQpjBo1CmFhYUhNTQUAbNu2DW5ubli6dCkyMzORmZkJADh79ixGjRqFSZMm4a+//kJkZCQOHTqEf/3rX42Ki4juQtrD6ImotXrqqafE8ePH1/veW2+9JYaGhmqUpaWliQDExMTEeu/JyckRAYhnz54VRVEUk5OTRQBieHh4nc/19PQUq6ur1WWPPfaYOGXKFPVrT09P8bPPPlO/BiC++eab6tfFxcWiIAjizp07RVEUxTfeeEP09/fX+JwlS5aIAMRbt27V/wW4/TkKhUK0sLAQjY2NRQCiqampePjw4bveI4qi6OfnJ3755Zd3jVcURXH69Onic889p1EWGxsrymQysays7J7PJ6L7Yw8QEWldfHw89u/fD0tLS/XVrVs3AFAPc129ehVTp05Fx44dYW1tDW9vbwBQ94zUCgoKqvP87t27Qy6Xq1+7uLggJyfnnjH17NlT/f8WFhawsrJS35OYmIi+fftq1O/Xr1+D2vraa6/h9OnTOHjwIIYNG4YlS5YgODhY/X5JSQlef/11+Pn5wdbWFpaWlrh48WKddt4pPj4eGzZs0Pgajho1CiqVCsnJyQ2KjYjujpOgiUjrVCoVwsLC8PHHH9d5z8XFBQAQFhYGd3d3rFmzBh06dIBKpYK/vz8qKys16ltYWNR5xp0ToQVBqDN01ph7RFHUmHtTW9YQjo6O8PHxgY+PD7Zu3QofHx888MADGDlyJICaBGn37t1YsWIFfHx8YGZmhkcffbROO++kUqnw/PPP4+WXX67znoeHR4NiI6K7YwJERFoXEBCArVu3wsvLC0ZGdf+auXnzJhISErB69WqEhIQAAA4dOtTSYap169YNUVFRGmUnTpxo9HPs7Ozw0ksvYeHChTh16hQEQUBsbCxmzpyJiRMnAqiZE5SSkqJxn0KhgFKp1CgLCAjA+fPn4ePj0+g4iOj+OARGRE1WUFCA06dPa1ypqamYO3cu8vLy8MQTT+DYsWNISkrCnj17MGvWLCiVStjZ2cHBwQHffvstrly5gj/++AMLFiyQrB3PP/88Ll68iDfeeAOXLl3Czz//rJ5UfWfP0P3MnTsXiYmJ2Lp1KwDAx8cH27Ztw+nTp3HmzBlMnTq1Tm+Vl5cXYmJikJ6erl4p98Ybb+DIkSOYO3cuTp8+jcuXL2PHjh146aWXmt9gImICRERNd+DAAfTp00fjevvtt9GhQwccPnwYSqUSo0aNgr+/P1555RXY2NhAJpNBJpNh8+bNiI+Ph7+/P+bPn49PPvlEsnZ4e3tjy5Yt2LZtG3r27IlVq1apV4GZmJg06lnt2rXD9OnT8c4770ClUuGzzz6DnZ0dgoODERYWhlGjRiEgIEDjnqVLlyIlJQWdOnVS72HUs2dPHDx4EJcvX0ZISAj69OmDt956Sz2ESETNI4gNHegmIjIgH3zwAb755hukpaVJHQoR6QDnABERAYiIiEDfvn3h4OCAw4cP45NPPuGeO0RtGBMgIiIAly9fxvvvv4+8vDx4eHjg1VdfxeLFi6UOi4h0hENgREREZHA4CZqIiIgMDhMgIiIiMjhMgIiIiMjgMAEiIiIig8MEiIiIiAwOEyAiIiIyOEyAiIiIyOAwASIiIiKDwwSIiIiIDM7/A2JFlOEIrPTRAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.lr_find(suggest_funcs=(slide, valley))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "853bee4f-1b0a-418f-bc23-6aa25c15ea10",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.601420</td>\n",
       "      <td>0.646064</td>\n",
       "      <td>0.595506</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.525531</td>\n",
       "      <td>0.494071</td>\n",
       "      <td>0.775281</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.477161</td>\n",
       "      <td>0.434220</td>\n",
       "      <td>0.792135</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.450097</td>\n",
       "      <td>0.427103</td>\n",
       "      <td>0.814607</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.438997</td>\n",
       "      <td>0.412734</td>\n",
       "      <td>0.842697</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.426286</td>\n",
       "      <td>0.421427</td>\n",
       "      <td>0.803371</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.416050</td>\n",
       "      <td>0.439325</td>\n",
       "      <td>0.837079</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.410603</td>\n",
       "      <td>0.415310</td>\n",
       "      <td>0.786517</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.404997</td>\n",
       "      <td>0.424563</td>\n",
       "      <td>0.825843</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.399146</td>\n",
       "      <td>0.444472</td>\n",
       "      <td>0.837079</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.395762</td>\n",
       "      <td>0.433492</td>\n",
       "      <td>0.837079</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.391731</td>\n",
       "      <td>0.438659</td>\n",
       "      <td>0.825843</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.387056</td>\n",
       "      <td>0.441008</td>\n",
       "      <td>0.831461</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.382388</td>\n",
       "      <td>0.437289</td>\n",
       "      <td>0.831461</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.377014</td>\n",
       "      <td>0.452733</td>\n",
       "      <td>0.831461</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.376351</td>\n",
       "      <td>0.463995</td>\n",
       "      <td>0.831461</td>\n",
       "      <td>00:00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit(16, lr=0.03)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "10ce1db2-7a38-4339-b93f-3db690093f56",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "<style>\n",
       "    /* Turns off some styling */\n",
       "    progress {\n",
       "        /* gets rid of default border in Firefox and Opera. */\n",
       "        border: none;\n",
       "        /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "        background-size: auto;\n",
       "    }\n",
       "    progress:not([value]), progress:not([value])::-webkit-progress-bar {\n",
       "        background: repeating-linear-gradient(45deg, #7e7e7e, #7e7e7e 10px, #5c5c5c 10px, #5c5c5c 20px);\n",
       "    }\n",
       "    .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "        background: #F44336;\n",
       "    }\n",
       "</style>\n"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_test = preprocess(df_test)\n",
    "dl_test = learn.dls.test_dl(df_test)\n",
    "\n",
    "preds, _ = learn.get_preds(dl=dl_test)\n",
    "df_test['Survived'] = (preds[:,1]>0.5).int()\n",
    "\n",
    "submission = df_test[['PassengerId', 'Survived']]\n",
    "submission.to_csv('submission.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "f31d0612-803e-4e4d-9b83-165064b4f886",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "PassengerId,Survived\n",
      "892,0\n",
      "893,1\n",
      "894,0\n",
      "895,0\n",
      "896,1\n",
      "897,0\n",
      "898,1\n",
      "899,1\n",
      "900,1\n"
     ]
    }
   ],
   "source": [
    "!head submission.csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "0f315540-335f-4caa-883d-78df8b17098f",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0a1d835d-e726-4736-b229-01a26abc11f0",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "environment": {
   "kernel": "python3",
   "name": "pytorch-gpu.1-12.m99",
   "type": "gcloud",
   "uri": "gcr.io/deeplearning-platform-release/pytorch-gpu.1-12:m99"
  },
  "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": "3.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}