searching/search_results_relevance_using_BERT.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predicting Search Engine Results Relevance using BERT\n",
    "In this notebook I'll show you how easy it is to fine tune BERT to predict the relevance of search results and beat a previous state of the art approach."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from IPython.display import Image\n",
    "Image(filename='myLateKaggleWinner.SearchRelevance.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "I0820 00:31:09.548614 139904012736320 file_utils.py:39] PyTorch version 1.4.0 available.\n",
      "I0820 00:31:11.283068 139904012736320 file_utils.py:55] TensorFlow version 2.0.0 available.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "device(type='cuda')"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline\n",
    "import warnings\n",
    "warnings.simplefilter(action='ignore', category=FutureWarning)\n",
    "\n",
    "import torch\n",
    "from torch.utils.data import DataLoader\n",
    "import argparse\n",
    "import time\n",
    "from sklearn.metrics import accuracy_score, classification_report\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "from dataloader import SearchRelevanceDataset\n",
    "from transformers import AdamW\n",
    "from transformers import BertTokenizer, BertForSequenceClassification\n",
    "from transformers import get_linear_schedule_with_warmup\n",
    "\n",
    "\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "device"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### To run this notebook, you'll need to agree to the Kaggle license terms and download the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# the following commented out lines were needed, along with manually editing the column name to convert to utf8 and to avoid a keyword collision\n",
    "# train_file ='kaggle_ecommerce_complete_Rank_Crowd_AGG_Descriptions_ALL_ROWS.csv'\n",
    "# train_file_path = f'../data/crowdflower-search-relevance/{train_file}'\n",
    "# train_df = pd.read_csv(train_file_path, encoding='iso-8859-1' )\n",
    "# train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_file ='kaggle_ecommerce_complete_Rank_Crowd_AGG_Descriptions_ALL_ROWS_utf8.csv'\n",
    "train_file_path = f'../data/crowdflower-search-relevance/{train_file}'\n",
    "# train_df.to_csv(train_file_path, encoding='utf-8', index=False)\n",
    "# train_df = pd.read_csv(train_file_path)\n",
    "# train_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>_unit_id</th>\n",
       "      <th>relevance_score</th>\n",
       "      <th>relevance:variance</th>\n",
       "      <th>product_image</th>\n",
       "      <th>product_link</th>\n",
       "      <th>product_price</th>\n",
       "      <th>product_title</th>\n",
       "      <th>query</th>\n",
       "      <th>rank</th>\n",
       "      <th>source</th>\n",
       "      <th>url</th>\n",
       "      <th>product_description</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>711158459</td>\n",
       "      <td>3.67</td>\n",
       "      <td>0.471</td>\n",
       "      <td>http:https://thumbs2.ebaystatic.com/d/l225/m/mzvzEUI...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...</td>\n",
       "      <td>$329.98</td>\n",
       "      <td>Sony PlayStation 4 (PS4) (Latest Model)- 500 G...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>1</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>The PlayStation 4 system opens the door to an ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>711158460</td>\n",
       "      <td>4.00</td>\n",
       "      <td>0.000</td>\n",
       "      <td>http:https://thumbs3.ebaystatic.com/d/l225/m/mJNDmSy...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-Lat...</td>\n",
       "      <td>$324.84</td>\n",
       "      <td>Sony PlayStation 4 (Latest Model)- 500 GB Jet ...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>2</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>The PlayStation 4 system opens the door to an ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>711158461</td>\n",
       "      <td>4.00</td>\n",
       "      <td>0.000</td>\n",
       "      <td>http:https://thumbs4.ebaystatic.com/d/l225/m/m10NZXA...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...</td>\n",
       "      <td>$324.83</td>\n",
       "      <td>Sony PlayStation 4 PS4 500 GB Jet Black Console</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>3</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>The PlayStation 4 system opens the door to an ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>711158462</td>\n",
       "      <td>3.67</td>\n",
       "      <td>0.471</td>\n",
       "      <td>http:https://thumbs2.ebaystatic.com/d/l225/m/mZZXTmA...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-500...</td>\n",
       "      <td>$350.00</td>\n",
       "      <td>Sony - PlayStation 4 500GB The Last of Us Rema...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>4</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>711158463</td>\n",
       "      <td>3.33</td>\n",
       "      <td>0.471</td>\n",
       "      <td>http:https://thumbs3.ebaystatic.com/d/l225/m/mzvzEUI...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...</td>\n",
       "      <td>$308.00\\nTrending at\\n$319.99</td>\n",
       "      <td>Sony PlayStation 4 (PS4) (Latest Model)- 500 G...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>5</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>The PlayStation 4 system opens the door to an ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    _unit_id  relevance_score  relevance:variance  \\\n",
       "0  711158459             3.67               0.471   \n",
       "1  711158460             4.00               0.000   \n",
       "2  711158461             4.00               0.000   \n",
       "3  711158462             3.67               0.471   \n",
       "4  711158463             3.33               0.471   \n",
       "\n",
       "                                       product_image  \\\n",
       "0  http:https://thumbs2.ebaystatic.com/d/l225/m/mzvzEUI...   \n",
       "1  http:https://thumbs3.ebaystatic.com/d/l225/m/mJNDmSy...   \n",
       "2  http:https://thumbs4.ebaystatic.com/d/l225/m/m10NZXA...   \n",
       "3  http:https://thumbs2.ebaystatic.com/d/l225/m/mZZXTmA...   \n",
       "4  http:https://thumbs3.ebaystatic.com/d/l225/m/mzvzEUI...   \n",
       "\n",
       "                                        product_link  \\\n",
       "0  http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...   \n",
       "1  http:https://www.ebay.com/itm/Sony-PlayStation-4-Lat...   \n",
       "2  http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...   \n",
       "3  http:https://www.ebay.com/itm/Sony-PlayStation-4-500...   \n",
       "4  http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...   \n",
       "\n",
       "                   product_price  \\\n",
       "0                       $329.98    \n",
       "1                       $324.84    \n",
       "2                       $324.83    \n",
       "3                       $350.00    \n",
       "4  $308.00\\nTrending at\\n$319.99   \n",
       "\n",
       "                                       product_title          query  rank  \\\n",
       "0  Sony PlayStation 4 (PS4) (Latest Model)- 500 G...  playstation 4     1   \n",
       "1  Sony PlayStation 4 (Latest Model)- 500 GB Jet ...  playstation 4     2   \n",
       "2    Sony PlayStation 4 PS4 500 GB Jet Black Console  playstation 4     3   \n",
       "3  Sony - PlayStation 4 500GB The Last of Us Rema...  playstation 4     4   \n",
       "4  Sony PlayStation 4 (PS4) (Latest Model)- 500 G...  playstation 4     5   \n",
       "\n",
       "  source                                                url  \\\n",
       "0   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "1   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "2   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "3   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "4   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "\n",
       "                                 product_description  \n",
       "0  The PlayStation 4 system opens the door to an ...  \n",
       "1  The PlayStation 4 system opens the door to an ...  \n",
       "2  The PlayStation 4 system opens the door to an ...  \n",
       "3                                                NaN  \n",
       "4  The PlayStation 4 system opens the door to an ...  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df = pd.read_csv(train_file_path)\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## After inspecting the data, we decide to only use values for which we have relevance scores.\n",
    "It is left to the reader as an exercise to determine if imputing the missing values would boost the classifier scores. So much low hanging fruit."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "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>_unit_id</th>\n",
       "      <th>relevance_score</th>\n",
       "      <th>relevance:variance</th>\n",
       "      <th>product_image</th>\n",
       "      <th>product_link</th>\n",
       "      <th>product_price</th>\n",
       "      <th>product_title</th>\n",
       "      <th>query</th>\n",
       "      <th>rank</th>\n",
       "      <th>source</th>\n",
       "      <th>url</th>\n",
       "      <th>product_description</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>20571</th>\n",
       "      <td>713184536</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>http:https://thumbs1.ebaystatic.com/d/l225/m/mhFJ074...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-Bun...</td>\n",
       "      <td>$430.00</td>\n",
       "      <td>Sony PlayStation 4  Bundle with 6 Games GTA5 A...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>26</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20572</th>\n",
       "      <td>713184537</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>http:https://thumbs3.ebaystatic.com/d/l225/m/mzvzEUI...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-Lat...</td>\n",
       "      <td>$350.00</td>\n",
       "      <td>Sony PlayStation 4 (Latest Model)- 500 GB Jet ...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>27</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20573</th>\n",
       "      <td>713184538</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>http:https://thumbs3.ebaystatic.com/d/l225/m/mZ_dYhk...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-Lat...</td>\n",
       "      <td>$299.99\\nTrending at\\n$319.99</td>\n",
       "      <td>Sony PlayStation 4 (Latest Model)- 500 GB Jet ...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>28</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>&lt;ul&gt;\\n\\t\\t&lt;li&gt;\\n\\t\\t\\tEnglish \\n\\t\\t\\t\\t&lt;/li&gt;\\...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20574</th>\n",
       "      <td>713184539</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>http:https://thumbs1.ebaystatic.com/d/l225/m/mlPL5eW...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-Lat...</td>\n",
       "      <td>$325.00</td>\n",
       "      <td>Sony PlayStation 4 (Latest Model)- 500 GB Jet ...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>29</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20575</th>\n",
       "      <td>713184540</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>http:https://thumbs3.ebaystatic.com/d/l225/m/mVjQFAh...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-500...</td>\n",
       "      <td>$273.99\\nTrending at\\n$319.99</td>\n",
       "      <td>Sony PlayStation 4 - 500 GB PS4 Jet Black --CO...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>30</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>&lt;ul&gt;\\n\\t\\t&lt;li&gt;\\n\\t\\t\\tEnglish \\n\\t\\t\\t\\t&lt;/li&gt;\\...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        _unit_id  relevance_score  relevance:variance  \\\n",
       "20571  713184536              NaN                 NaN   \n",
       "20572  713184537              NaN                 NaN   \n",
       "20573  713184538              NaN                 NaN   \n",
       "20574  713184539              NaN                 NaN   \n",
       "20575  713184540              NaN                 NaN   \n",
       "\n",
       "                                           product_image  \\\n",
       "20571  http:https://thumbs1.ebaystatic.com/d/l225/m/mhFJ074...   \n",
       "20572  http:https://thumbs3.ebaystatic.com/d/l225/m/mzvzEUI...   \n",
       "20573  http:https://thumbs3.ebaystatic.com/d/l225/m/mZ_dYhk...   \n",
       "20574  http:https://thumbs1.ebaystatic.com/d/l225/m/mlPL5eW...   \n",
       "20575  http:https://thumbs3.ebaystatic.com/d/l225/m/mVjQFAh...   \n",
       "\n",
       "                                            product_link  \\\n",
       "20571  http:https://www.ebay.com/itm/Sony-PlayStation-4-Bun...   \n",
       "20572  http:https://www.ebay.com/itm/Sony-PlayStation-4-Lat...   \n",
       "20573  http:https://www.ebay.com/itm/Sony-PlayStation-4-Lat...   \n",
       "20574  http:https://www.ebay.com/itm/Sony-PlayStation-4-Lat...   \n",
       "20575  http:https://www.ebay.com/itm/Sony-PlayStation-4-500...   \n",
       "\n",
       "                       product_price  \\\n",
       "20571                       $430.00    \n",
       "20572                       $350.00    \n",
       "20573  $299.99\\nTrending at\\n$319.99   \n",
       "20574                       $325.00    \n",
       "20575  $273.99\\nTrending at\\n$319.99   \n",
       "\n",
       "                                           product_title          query  rank  \\\n",
       "20571  Sony PlayStation 4  Bundle with 6 Games GTA5 A...  playstation 4    26   \n",
       "20572  Sony PlayStation 4 (Latest Model)- 500 GB Jet ...  playstation 4    27   \n",
       "20573  Sony PlayStation 4 (Latest Model)- 500 GB Jet ...  playstation 4    28   \n",
       "20574  Sony PlayStation 4 (Latest Model)- 500 GB Jet ...  playstation 4    29   \n",
       "20575  Sony PlayStation 4 - 500 GB PS4 Jet Black --CO...  playstation 4    30   \n",
       "\n",
       "      source                                                url  \\\n",
       "20571   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "20572   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "20573   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "20574   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "20575   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "\n",
       "                                     product_description  \n",
       "20571                                                NaN  \n",
       "20572                                                NaN  \n",
       "20573  <ul>\\n\\t\\t<li>\\n\\t\\t\\tEnglish \\n\\t\\t\\t\\t</li>\\...  \n",
       "20574                                                NaN  \n",
       "20575  <ul>\\n\\t\\t<li>\\n\\t\\t\\tEnglish \\n\\t\\t\\t\\t</li>\\...  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.query('relevance_score < 0 | relevance_score.isnull()', engine='python').head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### In another notebook we might predict this nan values and cross validate, and maybe even manually label, etc. There are all sorts of improvements that can be done, but this notebook focuses on how big results can be gained with just fine tuning BERT in a coarse manner."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "32671"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(train_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "20571"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df.dropna(subset=['relevance_score'], inplace=True)\n",
    "len(train_df)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## We'll also collapse the relevance scores into four categories"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>_unit_id</th>\n",
       "      <th>relevance_score</th>\n",
       "      <th>relevance:variance</th>\n",
       "      <th>product_image</th>\n",
       "      <th>product_link</th>\n",
       "      <th>product_price</th>\n",
       "      <th>product_title</th>\n",
       "      <th>query</th>\n",
       "      <th>rank</th>\n",
       "      <th>source</th>\n",
       "      <th>url</th>\n",
       "      <th>product_description</th>\n",
       "      <th>median_relevance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>711158459</td>\n",
       "      <td>3.67</td>\n",
       "      <td>0.471</td>\n",
       "      <td>http:https://thumbs2.ebaystatic.com/d/l225/m/mzvzEUI...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...</td>\n",
       "      <td>$329.98</td>\n",
       "      <td>Sony PlayStation 4 (PS4) (Latest Model)- 500 G...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>1</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>The PlayStation 4 system opens the door to an ...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>711158460</td>\n",
       "      <td>4.00</td>\n",
       "      <td>0.000</td>\n",
       "      <td>http:https://thumbs3.ebaystatic.com/d/l225/m/mJNDmSy...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-Lat...</td>\n",
       "      <td>$324.84</td>\n",
       "      <td>Sony PlayStation 4 (Latest Model)- 500 GB Jet ...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>2</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>The PlayStation 4 system opens the door to an ...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>711158461</td>\n",
       "      <td>4.00</td>\n",
       "      <td>0.000</td>\n",
       "      <td>http:https://thumbs4.ebaystatic.com/d/l225/m/m10NZXA...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...</td>\n",
       "      <td>$324.83</td>\n",
       "      <td>Sony PlayStation 4 PS4 500 GB Jet Black Console</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>3</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>The PlayStation 4 system opens the door to an ...</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>711158462</td>\n",
       "      <td>3.67</td>\n",
       "      <td>0.471</td>\n",
       "      <td>http:https://thumbs2.ebaystatic.com/d/l225/m/mZZXTmA...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-500...</td>\n",
       "      <td>$350.00</td>\n",
       "      <td>Sony - PlayStation 4 500GB The Last of Us Rema...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>4</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>711158463</td>\n",
       "      <td>3.33</td>\n",
       "      <td>0.471</td>\n",
       "      <td>http:https://thumbs3.ebaystatic.com/d/l225/m/mzvzEUI...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...</td>\n",
       "      <td>$308.00\\nTrending at\\n$319.99</td>\n",
       "      <td>Sony PlayStation 4 (PS4) (Latest Model)- 500 G...</td>\n",
       "      <td>playstation 4</td>\n",
       "      <td>5</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>The PlayStation 4 system opens the door to an ...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    _unit_id  relevance_score  relevance:variance  \\\n",
       "0  711158459             3.67               0.471   \n",
       "1  711158460             4.00               0.000   \n",
       "2  711158461             4.00               0.000   \n",
       "3  711158462             3.67               0.471   \n",
       "4  711158463             3.33               0.471   \n",
       "\n",
       "                                       product_image  \\\n",
       "0  http:https://thumbs2.ebaystatic.com/d/l225/m/mzvzEUI...   \n",
       "1  http:https://thumbs3.ebaystatic.com/d/l225/m/mJNDmSy...   \n",
       "2  http:https://thumbs4.ebaystatic.com/d/l225/m/m10NZXA...   \n",
       "3  http:https://thumbs2.ebaystatic.com/d/l225/m/mZZXTmA...   \n",
       "4  http:https://thumbs3.ebaystatic.com/d/l225/m/mzvzEUI...   \n",
       "\n",
       "                                        product_link  \\\n",
       "0  http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...   \n",
       "1  http:https://www.ebay.com/itm/Sony-PlayStation-4-Lat...   \n",
       "2  http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...   \n",
       "3  http:https://www.ebay.com/itm/Sony-PlayStation-4-500...   \n",
       "4  http:https://www.ebay.com/itm/Sony-PlayStation-4-PS4...   \n",
       "\n",
       "                   product_price  \\\n",
       "0                       $329.98    \n",
       "1                       $324.84    \n",
       "2                       $324.83    \n",
       "3                       $350.00    \n",
       "4  $308.00\\nTrending at\\n$319.99   \n",
       "\n",
       "                                       product_title          query  rank  \\\n",
       "0  Sony PlayStation 4 (PS4) (Latest Model)- 500 G...  playstation 4     1   \n",
       "1  Sony PlayStation 4 (Latest Model)- 500 GB Jet ...  playstation 4     2   \n",
       "2    Sony PlayStation 4 PS4 500 GB Jet Black Console  playstation 4     3   \n",
       "3  Sony - PlayStation 4 500GB The Last of Us Rema...  playstation 4     4   \n",
       "4  Sony PlayStation 4 (PS4) (Latest Model)- 500 G...  playstation 4     5   \n",
       "\n",
       "  source                                                url  \\\n",
       "0   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "1   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "2   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "3   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "4   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "\n",
       "                                 product_description  median_relevance  \n",
       "0  The PlayStation 4 system opens the door to an ...                 4  \n",
       "1  The PlayStation 4 system opens the door to an ...                 4  \n",
       "2  The PlayStation 4 system opens the door to an ...                 4  \n",
       "3                                                NaN                 4  \n",
       "4  The PlayStation 4 system opens the door to an ...                 3  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df['median_relevance'] = [int(round((tmp) )) for tmp in train_df['relevance_score'].tolist()]\n",
    "train_df.head()                                    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(20571, 20571)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ids = train_df['_unit_id'].tolist()\n",
    "len(ids), len(set(ids))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(20571, 18894)"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "titles = train_df['product_title'].tolist()\n",
    "len(titles), len(set(titles))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(set( train_df['median_relevance'].tolist()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "I0820 00:31:19.779527 139904012736320 tokenization_utils.py:1015] loading file https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-vocab.txt from cache at /home/todd/.cache/torch/transformers/26bc1ad6c0ac742e9b52263248f6d0f00068293b33709fae12320c0e35ccfbbb.542ce4285a40d23a559526243235df47c5f75c197f04f37d1a0c124c32c9a084\n",
      "I0820 00:31:20.293649 139904012736320 configuration_utils.py:285] loading configuration file https://s3.amazonaws.com/models.huggingface.co/bert/bert-base-uncased-config.json from cache at /home/todd/.cache/torch/transformers/4dad0251492946e18ac39290fcfe91b89d370fee250efe9521476438fe8ca185.7156163d5fdc189c3016baca0775ffce230789d7fa2a42ef516483e4ca884517\n",
      "I0820 00:31:20.294543 139904012736320 configuration_utils.py:321] Model config BertConfig {\n",
      "  \"architectures\": [\n",
      "    \"BertForMaskedLM\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 768,\n",
      "  \"id2label\": {\n",
      "    \"0\": \"LABEL_0\",\n",
      "    \"1\": \"LABEL_1\",\n",
      "    \"2\": \"LABEL_2\",\n",
      "    \"3\": \"LABEL_3\"\n",
      "  },\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 3072,\n",
      "  \"label2id\": {\n",
      "    \"LABEL_0\": 0,\n",
      "    \"LABEL_1\": 1,\n",
      "    \"LABEL_2\": 2,\n",
      "    \"LABEL_3\": 3\n",
      "  },\n",
      "  \"layer_norm_eps\": 1e-12,\n",
      "  \"max_position_embeddings\": 512,\n",
      "  \"model_type\": \"bert\",\n",
      "  \"num_attention_heads\": 12,\n",
      "  \"num_hidden_layers\": 12,\n",
      "  \"pad_token_id\": 0,\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"vocab_size\": 30522\n",
      "}\n",
      "\n",
      "I0820 00:31:20.511566 139904012736320 modeling_utils.py:650] loading weights file https://cdn.huggingface.co/bert-base-uncased-pytorch_model.bin from cache at /home/todd/.cache/torch/transformers/f2ee78bdd635b758cc0a12352586868bef80e47401abe4c4fcc3832421e7338b.36ca03ab34a1a5d5fa7bc3d03d55c4fa650fed07220e2eeebc06ce58d0e9a157\n",
      "I0820 00:31:21.983448 139904012736320 modeling_utils.py:741] Weights of BertForSequenceClassification not initialized from pretrained model: ['classifier.weight', 'classifier.bias']\n",
      "I0820 00:31:21.984080 139904012736320 modeling_utils.py:747] Weights from pretrained model not used in BertForSequenceClassification: ['cls.predictions.bias', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.weight', 'cls.seq_relationship.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.predictions.transform.LayerNorm.bias']\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "device(type='cuda')"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')\n",
    "model = BertForSequenceClassification.from_pretrained('bert-base-uncased', num_labels=4)\n",
    "model.to(device)\n",
    "device"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Let's take a look at the class distribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4    11065\n",
      "3     5411\n",
      "2     2879\n",
      "1     1216\n",
      "Name: median_relevance, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "print(train_df['median_relevance'].value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "4    0.537893\n",
       "3    0.263040\n",
       "2    0.139954\n",
       "1    0.059112\n",
       "Name: median_relevance, dtype: float64"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_df['median_relevance'].value_counts() / len(train_df)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## So our baseline to beat is 53%, which we get if we predict the most common class\n",
    "\n",
    "## Let's create dataframes for training and validation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "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>_unit_id</th>\n",
       "      <th>relevance_score</th>\n",
       "      <th>relevance:variance</th>\n",
       "      <th>product_image</th>\n",
       "      <th>product_link</th>\n",
       "      <th>product_price</th>\n",
       "      <th>product_title</th>\n",
       "      <th>query</th>\n",
       "      <th>rank</th>\n",
       "      <th>source</th>\n",
       "      <th>url</th>\n",
       "      <th>product_description</th>\n",
       "      <th>median_relevance</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>711158495</td>\n",
       "      <td>4.00</td>\n",
       "      <td>0.000</td>\n",
       "      <td>http:https://thumbs1.ebaystatic.com/d/l225/m/mFuQxKo...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Lancome-High-Resolutio...</td>\n",
       "      <td>$23.50</td>\n",
       "      <td>Lancome High Resolution Eye Cream .95 Oz Refil...</td>\n",
       "      <td>eye cream</td>\n",
       "      <td>12</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46</th>\n",
       "      <td>711158505</td>\n",
       "      <td>3.67</td>\n",
       "      <td>0.471</td>\n",
       "      <td>http:https://thumbs2.ebaystatic.com/d/l225/m/m5EfWbi...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Jeunesse-Instantly-Age...</td>\n",
       "      <td>$65.99</td>\n",
       "      <td>Jeunesse Instantly Ageless -- New Box Of 50 Sa...</td>\n",
       "      <td>eye cream</td>\n",
       "      <td>22</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>59</th>\n",
       "      <td>711158518</td>\n",
       "      <td>3.67</td>\n",
       "      <td>0.471</td>\n",
       "      <td>http:https://thumbs2.ebaystatic.com/d/l225/m/mUPdyhb...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Netgear-N600-300-Mbps-...</td>\n",
       "      <td>$27.00</td>\n",
       "      <td>Netgear N600 300 Mbps 4-Port 10/100 Wireless N...</td>\n",
       "      <td>routers</td>\n",
       "      <td>10</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>81</th>\n",
       "      <td>711158540</td>\n",
       "      <td>4.00</td>\n",
       "      <td>0.000</td>\n",
       "      <td>http:https://thumbs1.ebaystatic.com/d/l225/m/mcrO4Bk...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Panasonic-ES8103S-Arc3...</td>\n",
       "      <td>$65.88</td>\n",
       "      <td>Panasonic ES8103S Arc3 Electric Shaver Wet/Dry...</td>\n",
       "      <td>shaver panasonic</td>\n",
       "      <td>7</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>84</th>\n",
       "      <td>711158543</td>\n",
       "      <td>4.00</td>\n",
       "      <td>0.000</td>\n",
       "      <td>http:https://thumbs4.ebaystatic.com/d/l225/m/mqWDU2m...</td>\n",
       "      <td>http:https://www.ebay.com/itm/Panasonic-ES3831K-Sing...</td>\n",
       "      <td>$15.76</td>\n",
       "      <td>Panasonic ES3831K Single Blade Wet &amp; Dry Trave...</td>\n",
       "      <td>shaver panasonic</td>\n",
       "      <td>10</td>\n",
       "      <td>eBay</td>\n",
       "      <td>http:https://www.ebay.com/sch/i.html?_from=R40&amp;_trks...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     _unit_id  relevance_score  relevance:variance  \\\n",
       "36  711158495             4.00               0.000   \n",
       "46  711158505             3.67               0.471   \n",
       "59  711158518             3.67               0.471   \n",
       "81  711158540             4.00               0.000   \n",
       "84  711158543             4.00               0.000   \n",
       "\n",
       "                                        product_image  \\\n",
       "36  http:https://thumbs1.ebaystatic.com/d/l225/m/mFuQxKo...   \n",
       "46  http:https://thumbs2.ebaystatic.com/d/l225/m/m5EfWbi...   \n",
       "59  http:https://thumbs2.ebaystatic.com/d/l225/m/mUPdyhb...   \n",
       "81  http:https://thumbs1.ebaystatic.com/d/l225/m/mcrO4Bk...   \n",
       "84  http:https://thumbs4.ebaystatic.com/d/l225/m/mqWDU2m...   \n",
       "\n",
       "                                         product_link product_price  \\\n",
       "36  http:https://www.ebay.com/itm/Lancome-High-Resolutio...       $23.50    \n",
       "46  http:https://www.ebay.com/itm/Jeunesse-Instantly-Age...       $65.99    \n",
       "59  http:https://www.ebay.com/itm/Netgear-N600-300-Mbps-...       $27.00    \n",
       "81  http:https://www.ebay.com/itm/Panasonic-ES8103S-Arc3...       $65.88    \n",
       "84  http:https://www.ebay.com/itm/Panasonic-ES3831K-Sing...       $15.76    \n",
       "\n",
       "                                        product_title             query  rank  \\\n",
       "36  Lancome High Resolution Eye Cream .95 Oz Refil...         eye cream    12   \n",
       "46  Jeunesse Instantly Ageless -- New Box Of 50 Sa...         eye cream    22   \n",
       "59  Netgear N600 300 Mbps 4-Port 10/100 Wireless N...           routers    10   \n",
       "81  Panasonic ES8103S Arc3 Electric Shaver Wet/Dry...  shaver panasonic     7   \n",
       "84  Panasonic ES3831K Single Blade Wet & Dry Trave...  shaver panasonic    10   \n",
       "\n",
       "   source                                                url  \\\n",
       "36   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "46   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "59   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "81   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "84   eBay  http:https://www.ebay.com/sch/i.html?_from=R40&_trks...   \n",
       "\n",
       "   product_description  median_relevance  \n",
       "36                 NaN                 4  \n",
       "46                 NaN                 4  \n",
       "59                 NaN                 4  \n",
       "81                 NaN                 4  \n",
       "84                 NaN                 4  "
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "training_df =  train_df.sample( int(len(train_df) * .9))\n",
    "validation_df = train_df.query(f'_unit_id not in {training_df._unit_id.tolist() }')\n",
    "validation_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "training_file='../data/crowdflower-search-relevance/mytraining.csv'\n",
    "validation_file = '../data/crowdflower-search-relevance/myvalidation.csv'\n",
    "\n",
    "# Do this once\n",
    "# training_df.to_csv( training_file, index=False)\n",
    "# validation_df.to_csv(validation_file, index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "parser = argparse.ArgumentParser()\n",
    "parser.add_argument('-gpu', type = int, default = 0)\n",
    "parser.add_argument('-freeze_bert', action='store_true', default=False)\n",
    "parser.add_argument('-maxlen', type = int, default= 512)\n",
    "parser.add_argument('-batch_size', type = int, default=8)\n",
    "parser.add_argument('-lr', type = float, default = 2e-5)\n",
    "parser.add_argument('-print_every', type = int, default= 1000)\n",
    "parser.add_argument('-max_eps', type = int, default=4)\n",
    "parser.add_argument('-fp16', type=bool, default=True)\n",
    "args = parser.parse_args([\"-gpu\", \"0\"] )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Creating dataloaders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Creating train and val dataloaders\n",
      "Done in 0.10383796691894531 seconds\n"
     ]
    }
   ],
   "source": [
    "print(\"Creating train and val dataloaders\")\n",
    "st = time.time()\n",
    "train_set = SearchRelevanceDataset(filename = training_file, maxlen = args.maxlen,\n",
    "                                  tokenizer=tokenizer ,\n",
    "                                  data_columns=('query', 'product_title',  'product_description'))\n",
    "val_set = SearchRelevanceDataset(filename = validation_file, maxlen = args.maxlen,\n",
    "                                tokenizer=tokenizer,\n",
    "                                data_columns=('query', 'product_title',  'product_description')) \n",
    "train_loader = DataLoader(train_set, batch_size = args.batch_size, num_workers = 2, shuffle=True) \n",
    "val_loader = DataLoader(val_set, batch_size = args.batch_size, num_workers = 1) \n",
    "print(\"Done in {} seconds\".format(time.time() - st))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "training_stats =[]\n",
    "\n",
    "def flat_accuracy(preds, labels):\n",
    "    \"\"\"Function to calculate the accuracy of our predictions vs labels\"\"\"\n",
    "    pred_flat = np.argmax(preds, axis=1).flatten()\n",
    "    labels_flat = labels.flatten()\n",
    "    return np.sum(pred_flat == labels_flat) / len(labels_flat)\n",
    "\n",
    "def evaluate(model, dataloader, args, ep, avg_train_loss):\n",
    "    model.eval()\n",
    "    total_eval_accuracy = 0\n",
    "    total_eval_loss = 0\n",
    "    nb_eval_steps = 0\n",
    "    t0 = time.time() \n",
    "    with torch.no_grad():\n",
    "        for seq, tok_type_ids, attn_masks, labels in dataloader:\n",
    "            seq, attn_masks = seq.cuda(args.gpu), attn_masks.cuda(args.gpu)            \n",
    "            tok_type_ids, labels = tok_type_ids.cuda(args.gpu), labels.cuda(args.gpu)                        \n",
    "            loss, logits = model(seq, \n",
    "                             token_type_ids=None, \n",
    "                             attention_mask=attn_masks,\n",
    "                                 labels=labels) \n",
    "            # Accumulate the validation loss.\n",
    "            total_eval_loss += loss.item()\n",
    "            # Move logits and labels to CPU\n",
    "            logits = logits.detach().cpu().numpy()\n",
    "            label_ids = labels.to('cpu').numpy()\n",
    "            # Calculate the accuracy for this batch & accumulate it over all batches.\n",
    "            total_eval_accuracy += flat_accuracy(logits, label_ids)       \n",
    "\n",
    "    # Report the final accuracy for this validation run.\n",
    "    avg_val_accuracy = total_eval_accuracy / len(dataloader)\n",
    "    print(\"  Accuracy: {0:.2f}\".format(avg_val_accuracy))\n",
    "    # Calculate the average loss over all of the batches.\n",
    "    avg_val_loss = total_eval_loss / len(dataloader)    \n",
    "    print(\"  Validation Loss: {0:.2f}\".format(avg_val_loss))\n",
    "    \n",
    "    # Record all statistics from this epoch.\n",
    "    training_stats.append(\n",
    "        {\n",
    "            'epoch': ep + 1,\n",
    "            'Training Loss': avg_train_loss,\n",
    "            'Valid. Loss': avg_val_loss,\n",
    "            'Valid. Accur.': avg_val_accuracy,\n",
    "        }\n",
    "    )\n",
    "    return avg_val_accuracy, avg_val_loss\n",
    "\n",
    "\n",
    "def train(model, opti, train_loader, val_loader, args):\n",
    "    model.zero_grad()\n",
    "    best_acc = 0        \n",
    "    for ep in range(args.max_eps):\n",
    "        # Reset the total loss for this epoch.\n",
    "        total_train_loss = 0   \n",
    "        model.train()\n",
    "        t0 = time.time()  # Measure how long the training epoch takes.\n",
    "        pbar = tqdm(total=len(train_loader))\n",
    "        for it, (seq, tok_type_ids, attn_masks, labels) in enumerate(train_loader):                        \n",
    "            opti.zero_grad()\n",
    "            seq, attn_masks = seq.cuda(args.gpu), attn_masks.cuda(args.gpu)\n",
    "            tok_type_ids, labels = tok_type_ids.cuda(args.gpu), labels.cuda(args.gpu)            \n",
    "            loss, logits = model(seq, \n",
    "                             token_type_ids=None, # tok_type_ids\n",
    "                             attention_mask=attn_masks, \n",
    "                             labels=labels)                      \n",
    "\n",
    "            # Accumulate the training loss for the epoch over all of the batches \n",
    "            total_train_loss += loss.item()\n",
    "\n",
    "            # Perform a backward pass to calculate the gradients.\n",
    "            if args.fp16:\n",
    "                with amp.scale_loss(loss, opti) as scaled_loss:\n",
    "                    scaled_loss.backward()\n",
    "            else:\n",
    "                loss.backward()    \n",
    "            \n",
    "            # Clip the norm of the gradients to 1.0.; to prevent the \"exploding gradients\" problem.\n",
    "            if args.fp16:\n",
    "                torch.nn.utils.clip_grad_norm_(amp.master_params(opti), 1.0)\n",
    "            else:\n",
    "                torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)\n",
    "                \n",
    "            # Update parameters and take a step using the computed gradient.\n",
    "            opti.step()\n",
    "            # Update the learning rate.\n",
    "            scheduler.step()\n",
    "            model.zero_grad()\n",
    "            pbar.update(1)\n",
    "            if it % args.print_every == 0 and it >0:\n",
    "                # Move logits and labels to CPU\n",
    "                logits = logits.detach().cpu().numpy()\n",
    "                label_ids = labels.to('cpu').numpy()\n",
    "                acc = flat_accuracy(logits, label_ids)                               \n",
    "                print(\"Iteration {} of epoch {} complete. Loss : {} Accuracy : {}\".format(it, ep, loss.item(), acc))\n",
    "        pbar.close()\n",
    "        \n",
    "        # Calculate the average loss over all of the batches.\n",
    "        avg_train_loss = total_train_loss / len(train_loader)            \n",
    "        print(\"Average training loss: {0:.2f}\".format(avg_train_loss))        \n",
    "        try:\n",
    "            val_acc, val_loss = evaluate(model, val_loader, args, ep, avg_train_loss)\n",
    "            print(\"Epoch {} complete! Validation Accuracy : {}, Validation Loss : {}\".format(ep, val_acc, val_loss))\n",
    "            if val_acc > best_acc:\n",
    "                print(\"Best validation accuracy improved from {} to {}, saving model...\".format(best_acc, val_acc))\n",
    "                best_acc = val_acc\n",
    "                torch.save(model.state_dict(), 'bert.cls.search.relevance._{}_freeze_{}.dat'.format(ep, args.freeze_bert))\n",
    "        except Exception as ex:\n",
    "            LOG.exception('failure %s', str(ex))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Number of training epochs. The BERT authors recommend between 2 and 4. \n",
    "epochs = args.max_eps #  4\n",
    "\n",
    "# Total number of training steps is [number of batches] x [number of epochs]. \n",
    "# (Note that this is not the same as the number of training samples).\n",
    "total_steps = len(train_loader) * epochs\n",
    "\n",
    "optimizer = AdamW(model.parameters(), lr = 2e-5, eps = 1e-8)\n",
    "scheduler = get_linear_schedule_with_warmup(optimizer, \n",
    "                                            num_warmup_steps = 0, # Default value in run_glue.py\n",
    "                                            num_training_steps = total_steps)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Do you have Apex? If so, training with mixed precision will speed things up"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Selected optimization level O1:  Insert automatic casts around Pytorch functions and Tensor methods.\n",
      "\n",
      "Defaults for this optimization level are:\n",
      "enabled                : True\n",
      "opt_level              : O1\n",
      "cast_model_type        : None\n",
      "patch_torch_functions  : True\n",
      "keep_batchnorm_fp32    : None\n",
      "master_weights         : None\n",
      "loss_scale             : dynamic\n",
      "Processing user overrides (additional kwargs that are not None)...\n",
      "After processing overrides, optimization options are:\n",
      "enabled                : True\n",
      "opt_level              : O1\n",
      "cast_model_type        : None\n",
      "patch_torch_functions  : True\n",
      "keep_batchnorm_fp32    : None\n",
      "master_weights         : None\n",
      "loss_scale             : dynamic\n",
      "Warning:  multi_tensor_applier fused unscale kernel is unavailable, possibly because apex was installed without --cuda_ext --cpp_ext. Using Python fallback.  Original ImportError was: ModuleNotFoundError(\"No module named 'amp_C'\")\n"
     ]
    }
   ],
   "source": [
    "# use Apex if available\n",
    "try:\n",
    "    from apex import amp\n",
    "    model, optimizer = amp.initialize(model, optimizer, opt_level=\"O1\") # s[\"fp16_opt_level\"])\n",
    "except ImportError:\n",
    "    raise ImportError(\"Please install apex from https://www.github.com/nvidia/apex to use fp16 training.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\r",
      "  0%|          | 0/2315 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/todd/anaconda3/envs/py37-mlyoucanuse/lib/python3.7/site-packages/torch/optim/lr_scheduler.py:113: UserWarning: Seems like `optimizer.step()` has been overridden after learning rate scheduler initialization. Please, make sure to call `optimizer.step()` before `lr_scheduler.step()`. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate\n",
      "  \"https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate\", UserWarning)\n",
      "\r",
      "  0%|          | 1/2315 [00:00<12:08,  3.18it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient overflow.  Skipping step, loss scaler 0 reducing loss scale to 32768.0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 211/2315 [00:45<06:54,  5.07it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient overflow.  Skipping step, loss scaler 0 reducing loss scale to 16384.0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 43%|████▎     | 1001/2315 [03:36<04:38,  4.71it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 1000 of epoch 0 complete. Loss : 1.4627025127410889 Accuracy : 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▋ | 2001/2315 [07:09<01:07,  4.64it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 2000 of epoch 0 complete. Loss : 0.6471481323242188 Accuracy : 0.875\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2315/2315 [08:15<00:00,  4.67it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average training loss: 0.92\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Accuracy: 0.66\n",
      "  Validation Loss: 0.82\n",
      "Epoch 0 complete! Validation Accuracy : 0.6589147286821705, Validation Loss : 0.823971323957739\n",
      "Best validation accuracy improved from 0 to 0.6589147286821705, saving model...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 41/2315 [00:08<07:21,  5.15it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient overflow.  Skipping step, loss scaler 0 reducing loss scale to 16384.0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 43%|████▎     | 1001/2315 [03:29<04:31,  4.84it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 1000 of epoch 1 complete. Loss : 1.1968224048614502 Accuracy : 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▋ | 2001/2315 [07:01<01:07,  4.62it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 2000 of epoch 1 complete. Loss : 0.811937153339386 Accuracy : 0.625\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 88%|████████▊ | 2046/2315 [07:10<00:52,  5.10it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient overflow.  Skipping step, loss scaler 0 reducing loss scale to 16384.0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2315/2315 [08:06<00:00,  4.76it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average training loss: 0.77\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Accuracy: 0.67\n",
      "  Validation Loss: 0.82\n",
      "Epoch 1 complete! Validation Accuracy : 0.6724806201550387, Validation Loss : 0.8170255799741708\n",
      "Best validation accuracy improved from 0.6589147286821705 to 0.6724806201550387, saving model...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 11%|█         | 248/2315 [00:51<06:36,  5.21it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient overflow.  Skipping step, loss scaler 0 reducing loss scale to 8192.0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 43%|████▎     | 1001/2315 [03:35<04:45,  4.60it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 1000 of epoch 2 complete. Loss : 1.0405322313308716 Accuracy : 0.5\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▋ | 2001/2315 [07:04<01:05,  4.78it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 2000 of epoch 2 complete. Loss : 0.8550999164581299 Accuracy : 0.625\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2315/2315 [08:09<00:00,  4.73it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average training loss: 0.68\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Accuracy: 0.68\n",
      "  Validation Loss: 0.81\n",
      "Epoch 2 complete! Validation Accuracy : 0.6841085271317829, Validation Loss : 0.8112265705600266\n",
      "Best validation accuracy improved from 0.6724806201550387 to 0.6841085271317829, saving model...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 23%|██▎       | 542/2315 [01:56<06:26,  4.58it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient overflow.  Skipping step, loss scaler 0 reducing loss scale to 8192.0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 43%|████▎     | 1001/2315 [03:34<04:37,  4.73it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 1000 of epoch 3 complete. Loss : 0.9393256902694702 Accuracy : 0.625\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 61%|██████    | 1408/2315 [04:56<02:54,  5.19it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Gradient overflow.  Skipping step, loss scaler 0 reducing loss scale to 4096.0\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      " 86%|████████▋ | 2001/2315 [07:04<01:07,  4.66it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Iteration 2000 of epoch 3 complete. Loss : 0.32130324840545654 Accuracy : 0.875\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2315/2315 [08:13<00:00,  4.69it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average training loss: 0.59\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  Accuracy: 0.68\n",
      "  Validation Loss: 0.88\n",
      "Epoch 3 complete! Validation Accuracy : 0.6792635658914729, Validation Loss : 0.8823379663995994\n",
      "Done in 2026.2599093914032 seconds\n"
     ]
    }
   ],
   "source": [
    "print(\"Training...\")\n",
    "st = time.time()\n",
    "train(model, optimizer, train_loader, val_loader, args)\n",
    "print(\"Done in {} seconds\".format(time.time() - st))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_train = [ep['Training Loss'] for ep in training_stats]\n",
    "loss_val = [ep['Valid. Loss'] for ep in training_stats]\n",
    "acc_val = [ep['Valid. Accur.'] for ep in training_stats]\n",
    "epochs = list(range(1,5))\n",
    "plt.plot(epochs, loss_train, 'g', label='Training loss')\n",
    "plt.plot(epochs, loss_val, 'b', label='validation loss')\n",
    "plt.plot(epochs, acc_val, 'r', label='validation accuracy')\n",
    "plt.title('Training & Validation loss & accuracy')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss / Acc')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## The graph shows a lot of improvement could be made, but, sadly, we'll see it's good enough to beat the previous SOTA"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Evaluate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.6622934888241011"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "\n",
    "# Put model in evaluation mode\n",
    "model.eval()\n",
    "# Tracking variables \n",
    "predictions, true_labels = [], []\n",
    "# Predict \n",
    "for batch in val_loader:\n",
    "    # Add batch to GPU\n",
    "    batch = tuple(t.to(device) for t in batch)\n",
    "    # Unpack the inputs from our dataloader\n",
    "    b_input_ids, b_tok_types, b_input_mask, b_labels = batch\n",
    "    # Telling the model not to compute or store gradients, saving memory and speeding up prediction\n",
    "    with torch.no_grad():\n",
    "        loss, logits = model(b_input_ids, \n",
    "                             token_type_ids=b_tok_types, \n",
    "                             attention_mask=b_input_mask, \n",
    "                             labels=b_labels)   \n",
    "        # Move logits and labels to CPU\n",
    "        logits = logits.detach().cpu().numpy()\n",
    "        label_ids = b_labels.cpu().numpy()  \n",
    "        # Store predictions and true labels\n",
    "        predictions.append(logits)\n",
    "        true_labels.append(label_ids)\n",
    "\n",
    "# Flatten the predictions and true values for evaluation on the whole dataset\n",
    "flat_predictions = [item for sublist in predictions for item in sublist]\n",
    "flat_predictions = np.argmax(flat_predictions, axis=1).flatten()\n",
    "flat_true_labels = [item for sublist in true_labels for item in sublist]\n",
    "\n",
    "accuracy_score(flat_true_labels, flat_predictions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "labels [1 2 3 4]\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           1       0.64      0.50      0.56       111\n",
      "           2       0.49      0.47      0.48       260\n",
      "           3       0.49      0.45      0.47       568\n",
      "           4       0.78      0.83      0.80      1119\n",
      "\n",
      "    accuracy                           0.66      2058\n",
      "   macro avg       0.60      0.56      0.58      2058\n",
      "weighted avg       0.65      0.66      0.66      2058\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print('labels', val_set.label_encoder.classes_)\n",
    "\n",
    "target_names = [ str(tmp) for tmp in val_set.label_encoder.classes_.tolist()]\n",
    "print(classification_report(flat_true_labels, flat_predictions, target_names=target_names))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "I0820 01:11:56.681124 139904012736320 configuration_utils.py:144] Configuration saved in mymodel/config.json\n",
      "I0820 01:11:57.829241 139904012736320 modeling_utils.py:483] Model weights saved in mymodel/pytorch_model.bin\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "('mymodel/vocab.txt',\n",
       " 'mymodel/special_tokens_map.json',\n",
       " 'mymodel/added_tokens.json')"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.save_pretrained('mymodel')\n",
    "tokenizer.save_pretrained('mymodel')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Let's load the test data and prepare it to make a submission"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "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>id</th>\n",
       "      <th>query</th>\n",
       "      <th>product_title</th>\n",
       "      <th>product_description</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>electric griddle</td>\n",
       "      <td>Star-Max 48 in Electric Griddle</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6</td>\n",
       "      <td>phillips coffee maker</td>\n",
       "      <td>Philips SENSEO HD7810 WHITE Single Serve Pod C...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>9</td>\n",
       "      <td>san francisco 49ers</td>\n",
       "      <td>2013 San Francisco 49ers Clock</td>\n",
       "      <td>A 2013 San Francisco 49ers clock is the ultima...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>11</td>\n",
       "      <td>aveeno shampoo</td>\n",
       "      <td>AVEENO       10.5FLOZ NRSH SHINE SH</td>\n",
       "      <td>Water, Ammonium Lauryl Sulfate, Dimethicone, S...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>12</td>\n",
       "      <td>flea and tick control for dogs</td>\n",
       "      <td>Merial Frontline Plus Flea and Tick Control fo...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   id                           query  \\\n",
       "0   3                electric griddle   \n",
       "1   6           phillips coffee maker   \n",
       "2   9             san francisco 49ers   \n",
       "3  11                  aveeno shampoo   \n",
       "4  12  flea and tick control for dogs   \n",
       "\n",
       "                                       product_title  \\\n",
       "0                    Star-Max 48 in Electric Griddle   \n",
       "1  Philips SENSEO HD7810 WHITE Single Serve Pod C...   \n",
       "2                     2013 San Francisco 49ers Clock   \n",
       "3                AVEENO       10.5FLOZ NRSH SHINE SH   \n",
       "4  Merial Frontline Plus Flea and Tick Control fo...   \n",
       "\n",
       "                                 product_description  \n",
       "0                                                NaN  \n",
       "1                                                NaN  \n",
       "2  A 2013 San Francisco 49ers clock is the ultima...  \n",
       "3  Water, Ammonium Lauryl Sulfate, Dimethicone, S...  \n",
       "4                                                NaN  "
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_file ='../data/crowdflower-search-relevance/test.csv'\n",
    "test_df = pd.read_csv(test_file)\n",
    "test_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_df['product_title'].fillna('', inplace=True)\n",
    "test_df['product_description'].fillna('', inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_set = SearchRelevanceDataset(filename = test_file, maxlen = args.maxlen,\n",
    "                                tokenizer=tokenizer,\n",
    "                                data_columns=('query', 'product_title',  'product_description')\n",
    "                                 , label_name=None)  \n",
    "test_loader = DataLoader(test_set, batch_size = args.batch_size, num_workers = 1, shuffle=False)  "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Let's prove we can reconstitute our model and make predictions for a submission"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 2815/2815 [02:37<00:00, 17.82it/s]\n"
     ]
    }
   ],
   "source": [
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "\n",
    "# Put model in evaluation mode\n",
    "model.eval()\n",
    "\n",
    "# Tracking variables \n",
    "predictions, true_labels = [], []\n",
    "\n",
    "# Predict \n",
    "for batch in tqdm(test_loader, total=len(test_loader)):\n",
    "    # Add batch to GPU\n",
    "    batch = tuple(t.to(device) for t in batch)\n",
    "    # Unpack the inputs from our dataloader\n",
    "    b_input_ids, b_tok_types, b_input_mask, b_labels = batch\n",
    "    # Telling the model not to compute or store gradients, saving memory and speeding up prediction\n",
    "    with torch.no_grad():\n",
    "        loss, logits = model(b_input_ids, \n",
    "                             token_type_ids=b_tok_types, \n",
    "                             attention_mask=b_input_mask, \n",
    "                             labels=b_labels)   \n",
    "        # Move logits and labels to CPU\n",
    "        logits = logits.detach().cpu().numpy()\n",
    "        label_ids = b_labels.cpu().numpy()  \n",
    "        # Store predictions and true labels\n",
    "        predictions.append(logits)\n",
    "        true_labels.append(label_ids)\n",
    "\n",
    "flat_predictions = [item for sublist in predictions for item in sublist]\n",
    "flat_predictions = np.argmax(flat_predictions, axis=1).flatten()\n",
    "\n",
    "submission_file = 'my.kaggle.searchRelevance.submission.csv'\n",
    "ids = test_df['id'].tolist()\n",
    "corrected_preds =[tmp +1 for tmp in flat_predictions]\n",
    "my_submission = pd.DataFrame({\"id\": ids, \"prediction\": corrected_preds})\n",
    "my_submission.to_csv(submission_file, index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "my.kaggle.searchRelevance.submission.csv\r\n"
     ]
    }
   ],
   "source": [
    "! ls my.kaggle.searchRelevance.submission.csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "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>id</th>\n",
       "      <th>prediction</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>9</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>11</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>12</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   id  prediction\n",
       "0   3           4\n",
       "1   6           4\n",
       "2   9           3\n",
       "3  11           4\n",
       "4  12           4"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "my_submission.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22513\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "4    0.570959\n",
       "3    0.260916\n",
       "2    0.131035\n",
       "1    0.037090\n",
       "Name: prediction, dtype: float64"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(len(my_submission))\n",
    "my_submission['prediction'].value_counts() / len(my_submission)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "source": [
    "## Now, in this next cell, we will invoke a time machine to take us back to when the contest was still running...\n",
    "Let me know if you beat this baseline."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}