Fine-tuning BERT.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "handmade-folks",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "from transformers import BertForSequenceClassification, BertTokenizer, Trainer, TrainingArguments, \\\n",
    "BertForQuestionAnswering, DistilBertForTokenClassification, AutoTokenizer, AutoModel, AutoModelForTokenClassification, \\\n",
    "pipeline, DataCollatorWithPadding\n",
    "\n",
    "from datasets import Dataset\n",
    "import torch\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import re\n",
    "\n",
    "# for visualizations\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "selective-saturn",
   "metadata": {},
   "source": [
    "# Quick Examples using the pipeline object in transformers\n",
    "\n",
    "We can run some quick examples of fine-tuned models using the nlp module in the transformers package"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "changed-tablet",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No model was supplied, defaulted to distilbert-base-cased-distilled-squad and revision 626af31 (https://huggingface.co/distilbert-base-cased-distilled-squad).\n",
      "Using a pipeline without specifying a model name and revision in production is not recommended.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'transformers.models.distilbert.modeling_distilbert.DistilBertForQuestionAnswering'>\n"
     ]
    }
   ],
   "source": [
    "# list of available pipelines here: https://huggingface.co/transformers/main_classes/pipelines.html\n",
    "nlp = pipeline(\"question-answering\")\n",
    "\n",
    "print(type(nlp.model))  # Using a flavor of BERT called DistilBERT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "anonymous-negotiation",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'score': 0.7949598431587219, 'start': 51, 'end': 54, 'answer': 'dog'}\n",
      "{'score': 0.7224201560020447, 'start': 72, 'end': 75, 'answer': 'cat'}\n",
      "{'score': 0.8167193531990051, 'start': 24, 'end': 33, 'answer': 'Tarantula'}\n"
     ]
    }
   ],
   "source": [
    "context = \"My friend Walker owns a Tarantula. I, Sinan, own a dog. Matt just got a cat!\"\n",
    "\n",
    "print(nlp(question=\"What kind of pet does Sinan have?\", context=context))  # should be dog\n",
    "print(nlp(question=\"What kind of pet does Matt have?\", context=context))  # should be cat\n",
    "print(nlp(question=\"What kind of pet does Walker have?\", context=context))  # should be Tarantula"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "mature-cleveland",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "stupid-determination",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No model was supplied, defaulted to distilbert-base-uncased-finetuned-sst-2-english and revision af0f99b (https://huggingface.co/distilbert-base-uncased-finetuned-sst-2-english).\n",
      "Using a pipeline without specifying a model name and revision in production is not recommended.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'transformers.models.distilbert.modeling_distilbert.DistilBertForSequenceClassification'>\n",
      "[{'label': 'POSITIVE', 'score': 0.9998711347579956}]\n"
     ]
    }
   ],
   "source": [
    "# A sentiment analysis pipeline\n",
    "nlp = pipeline(\"sentiment-analysis\")\n",
    "\n",
    "print(type(nlp.model))  # Using a flavor of BERT called DistilBERT\n",
    "\n",
    "print(nlp(\"My friend told me about this class and I love it so far! She was right.\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "adequate-manual",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "severe-april",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "No model was supplied, defaulted to distilroberta-base and revision ec58a5b (https://huggingface.co/distilroberta-base).\n",
      "Using a pipeline without specifying a model name and revision in production is not recommended.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'transformers.models.roberta.modeling_roberta.RobertaForMaskedLM'>\n",
      "If you don’t *** at the sign, you will get a ticket\n",
      "Token: stop. Score: 42.11%\n",
      "Token: look. Score: 7.53%\n",
      "Token: park. Score: 4.92%\n",
      "Token: arrive. Score: 4.65%\n",
      "Token: sign. Score: 4.27%\n"
     ]
    }
   ],
   "source": [
    "# We can even do a auto-encoder language model task!\n",
    "\n",
    "nlp = pipeline(\"fill-mask\")\n",
    "\n",
    "print(type(nlp.model))  # Using a flavor of BERT called RoBERTa\n",
    "\n",
    "preds = nlp(f\"If you don’t {nlp.tokenizer.mask_token} at the sign, you will get a ticket\")\n",
    "print('If you don’t *** at the sign, you will get a ticket')\n",
    "for p in preds:\n",
    "    print(f\"Token:{p['token_str']}. Score: {100*p['score']:,.2f}%\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "golden-chemistry",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "wicked-punch",
   "metadata": {},
   "source": [
    "# Example 1: Sequence Classification\n",
    "\n",
    "![BERT Sequence Classification](images/bert_seq_class.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "twelve-mining",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(200, 7)\n"
     ]
    },
    {
     "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>index</th>\n",
       "      <th>id</th>\n",
       "      <th>keyword</th>\n",
       "      <th>location</th>\n",
       "      <th>text</th>\n",
       "      <th>target</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7138</td>\n",
       "      <td>10224</td>\n",
       "      <td>volcano</td>\n",
       "      <td>NaN</td>\n",
       "      <td>@MrMikeEaton @Muazimus_Prime hill hill mountai...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2151</td>\n",
       "      <td>3086</td>\n",
       "      <td>deaths</td>\n",
       "      <td>Blackpool</td>\n",
       "      <td>Cancers equate for around 25% of all deaths in...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   index     id  keyword   location  \\\n",
       "0   7138  10224  volcano        NaN   \n",
       "1   2151   3086   deaths  Blackpool   \n",
       "\n",
       "                                                text  target  label  \n",
       "0  @MrMikeEaton @Muazimus_Prime hill hill mountai...       1      1  \n",
       "1  Cancers equate for around 25% of all deaths in...       1      1  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Ingest 100 tweets from the Kaggle disaster tweet comopetition\n",
    "\n",
    "tweets = pd.read_csv('./data/disaster_sample.csv')\n",
    "\n",
    "print(tweets.shape)\n",
    "\n",
    "tweets.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "debaed35",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "3d0a2332",
   "metadata": {},
   "outputs": [],
   "source": [
    "BERT_MODEL = 'bert-large-uncased'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "dadbbb48",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "powerful-machine",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tweet: @MrMikeEaton @Muazimus_Prime hill hill mountain volcano of hell mountain hill hil.. Num tokens: 24\n",
      "Token: 101, subword: [CLS]\n",
      "Token: 1030, subword: @\n",
      "Token: 2720, subword: mr\n",
      "Token: 4328, subword: ##mi\n",
      "Token: 20553, subword: ##kee\n",
      "Token: 22436, subword: ##aton\n",
      "Token: 1030, subword: @\n",
      "Token: 14163, subword: mu\n",
      "Token: 16103, subword: ##azi\n",
      "Token: 7606, subword: ##mus\n",
      "Token: 1035, subword: _\n",
      "Token: 3539, subword: prime\n",
      "Token: 2940, subword: hill\n",
      "Token: 2940, subword: hill\n",
      "Token: 3137, subword: mountain\n",
      "Token: 12779, subword: volcano\n",
      "Token: 1997, subword: of\n",
      "Token: 3109, subword: hell\n",
      "Token: 3137, subword: mountain\n",
      "Token: 2940, subword: hill\n",
      "Token: 7632, subword: hi\n",
      "Token: 2140, subword: ##l\n",
      "Token: 1012, subword: .\n",
      "Token: 102, subword: [SEP]\n"
     ]
    }
   ],
   "source": [
    "# check our tokenizer\n",
    "\n",
    "tokenizer = BertTokenizer.from_pretrained(BERT_MODEL)\n",
    "\n",
    "first_tweet = tweets['text'][0]\n",
    "\n",
    "tokens = tokenizer.encode(first_tweet)\n",
    "\n",
    "print(f'Tweet: {first_tweet}. Num tokens: {len(tokens)}')\n",
    "for t in tokens:\n",
    "    print(f'Token: {t}, subword: {tokenizer.decode([t])}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "demographic-garbage",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "reported-recognition",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    114\n",
       "1     86\n",
       "Name: target, dtype: int64"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tweets['target'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "conventional-asbestos",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    114\n",
       "1     86\n",
       "Name: label, dtype: int64"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# the trainer is expecting a 'label' (see the forward method in the docs)\n",
    "tweets['label'] = tweets['target']\n",
    "\n",
    "tweets['label'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "other-radar",
   "metadata": {},
   "outputs": [],
   "source": [
    "tweet_dataset = Dataset.from_pandas(tweets)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "suburban-arabic",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "datasets.arrow_dataset.Dataset"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(tweet_dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "51279322",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "2a2d8b5f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2b5349c30>"
      ]
     },
     "execution_count": 13,
     "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": [
    "# Histogram of tweet lengths after cleaning. Not very normal\n",
    "\n",
    "tweets['text'].apply(len).hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "409120e7",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "e453be67",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'MEG issues Hazardous Weather Outlook (HWO) url'"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "URL_REGEX = re.compile('http(s)?:\\/\\/t.co\\/\\w+')\n",
    "MENTION_REGEX = re.compile('@\\s?\\w+')\n",
    "\n",
    "def clean_tweet(tweet):\n",
    "    # remove mentions, the pound sign, and replace urls with URL token\n",
    "    tweet = re.sub(URL_REGEX, 'url', tweet)  # replace urls with url. Assumes that the mention of a url is significant\n",
    "    tweet = re.sub(MENTION_REGEX, '', tweet)  # remove mentions entirely\n",
    "    tweet = tweet.replace('#', '')  # remove pound signs\n",
    "    \n",
    "    return tweet.strip()\n",
    "\n",
    "clean_tweet('@prof_oz MEG issues Hazardous #Weather Outlook (HWO) http://t.co/3X6RBQJHn3')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3a7bf0f0",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "01457c7b",
   "metadata": {},
   "outputs": [],
   "source": [
    "tweets['text'] = tweets['text'].apply(clean_tweet)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "6f2fa5f7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x2b4e03e50>"
      ]
     },
     "execution_count": 16,
     "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": [
    "# Histogram of tweet lengths after cleaning. Much more normal :)\n",
    "\n",
    "tweets['text'].apply(len).hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bc97badd",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "ce936014",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Dataset({\n",
       "    features: ['index', 'id', 'keyword', 'location', 'text', 'target', 'label'],\n",
       "    num_rows: 200\n",
       "})"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tweet_dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "da46865f",
   "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>index</th>\n",
       "      <th>id</th>\n",
       "      <th>keyword</th>\n",
       "      <th>location</th>\n",
       "      <th>text</th>\n",
       "      <th>target</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7138</td>\n",
       "      <td>10224</td>\n",
       "      <td>volcano</td>\n",
       "      <td>NaN</td>\n",
       "      <td>hill hill mountain volcano of hell mountain hi...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2151</td>\n",
       "      <td>3086</td>\n",
       "      <td>deaths</td>\n",
       "      <td>Blackpool</td>\n",
       "      <td>Cancers equate for around 25% of all deaths in...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4395</td>\n",
       "      <td>6247</td>\n",
       "      <td>hijacking</td>\n",
       "      <td>World</td>\n",
       "      <td>The Murderous Story Of AmericaÛªs First Hijac...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2508</td>\n",
       "      <td>3602</td>\n",
       "      <td>desolation</td>\n",
       "      <td>Birmingham, UK</td>\n",
       "      <td>The date for the release of EP03 DESOLATION is...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1378</td>\n",
       "      <td>1987</td>\n",
       "      <td>bush%20fires</td>\n",
       "      <td>London/Bristol/Guildford</td>\n",
       "      <td>On holiday to relax sunbathe and drink ... Put...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>195</th>\n",
       "      <td>4522</td>\n",
       "      <td>6425</td>\n",
       "      <td>hurricane</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Stream HYPE HURRICANE</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>196</th>\n",
       "      <td>2049</td>\n",
       "      <td>2941</td>\n",
       "      <td>danger</td>\n",
       "      <td>Instagram: trillrebel_</td>\n",
       "      <td>Guns are for protection.. \\nThat shit really s...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>197</th>\n",
       "      <td>7532</td>\n",
       "      <td>10770</td>\n",
       "      <td>wreckage</td>\n",
       "      <td>iTunes</td>\n",
       "      <td>science Now that a piece of wreckage from flig...</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>198</th>\n",
       "      <td>5345</td>\n",
       "      <td>7630</td>\n",
       "      <td>pandemonium</td>\n",
       "      <td>illinois. united state</td>\n",
       "      <td>Pandemonium In Aba As Woman Delivers Baby With...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>199</th>\n",
       "      <td>1342</td>\n",
       "      <td>1940</td>\n",
       "      <td>burning%20buildings</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Hero's fight wars and save ppl from burning bu...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>200 rows × 7 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     index     id              keyword                  location  \\\n",
       "0     7138  10224              volcano                       NaN   \n",
       "1     2151   3086               deaths                 Blackpool   \n",
       "2     4395   6247            hijacking                     World   \n",
       "3     2508   3602           desolation            Birmingham, UK   \n",
       "4     1378   1987         bush%20fires  London/Bristol/Guildford   \n",
       "..     ...    ...                  ...                       ...   \n",
       "195   4522   6425            hurricane                       NaN   \n",
       "196   2049   2941               danger    Instagram: trillrebel_   \n",
       "197   7532  10770             wreckage                    iTunes   \n",
       "198   5345   7630          pandemonium   illinois. united state    \n",
       "199   1342   1940  burning%20buildings                       NaN   \n",
       "\n",
       "                                                  text  target  label  \n",
       "0    hill hill mountain volcano of hell mountain hi...       1      1  \n",
       "1    Cancers equate for around 25% of all deaths in...       1      1  \n",
       "2    The Murderous Story Of AmericaÛªs First Hijac...       1      1  \n",
       "3    The date for the release of EP03 DESOLATION is...       1      1  \n",
       "4    On holiday to relax sunbathe and drink ... Put...       0      0  \n",
       "..                                                 ...     ...    ...  \n",
       "195                              Stream HYPE HURRICANE       0      0  \n",
       "196  Guns are for protection.. \\nThat shit really s...       0      0  \n",
       "197  science Now that a piece of wreckage from flig...       1      1  \n",
       "198  Pandemonium In Aba As Woman Delivers Baby With...       0      0  \n",
       "199  Hero's fight wars and save ppl from burning bu...       0      0  \n",
       "\n",
       "[200 rows x 7 columns]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tweets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "8d22cf4b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'input_ids': [101, 4931, 2045, 102], 'token_type_ids': [0, 0, 0, 0], 'attention_mask': [1, 1, 1, 1]}"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tokenizer('hey there')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "59c97fb7",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "huggingface/tokenizers: The current process just got forked, after parallelism has already been used. Disabling parallelism to avoid deadlocks...\n",
      "To disable this warning, you can either:\n",
      "\t- Avoid using `tokenizers` before the fork if possible\n",
      "\t- Explicitly set the environment variable TOKENIZERS_PARALLELISM=(true | false)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "15c4d0bf28bc40279264f69f828b9421",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/1 [00:00<?, ?ba/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "{'label': 1,\n",
       " 'input_ids': [101,\n",
       "  1030,\n",
       "  2720,\n",
       "  4328,\n",
       "  20553,\n",
       "  22436,\n",
       "  1030,\n",
       "  14163,\n",
       "  16103,\n",
       "  7606,\n",
       "  1035,\n",
       "  3539,\n",
       "  2940,\n",
       "  2940,\n",
       "  3137,\n",
       "  12779,\n",
       "  1997,\n",
       "  3109,\n",
       "  3137,\n",
       "  2940,\n",
       "  7632,\n",
       "  2140,\n",
       "  1012,\n",
       "  102],\n",
       " 'token_type_ids': [0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0,\n",
       "  0],\n",
       " 'attention_mask': [1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1,\n",
       "  1]}"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# We will not pad our dataset so we can do it dynamically to speed up training\n",
    "def preprocess(data):\n",
    "    return tokenizer(data['text'], truncation=True)\n",
    "\n",
    "tweet_dataset = tweet_dataset.map(\n",
    "    preprocess, \n",
    "    batched=True, \n",
    "    batch_size=len(tweet_dataset),\n",
    "    remove_columns=['index', 'id', 'keyword', 'location', 'text', 'target']\n",
    ")\n",
    "tweet_dataset[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "79fc2acc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Dataset({\n",
       "    features: ['label', 'input_ids', 'token_type_ids', 'attention_mask'],\n",
       "    num_rows: 200\n",
       "})"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tweet_dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "db7a206e",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "informative-nelson",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Dataset has a built in train test split method\n",
    "tweet_dataset = tweet_dataset.train_test_split(test_size=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "48ef4933",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(160, 40)"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(tweet_dataset['train']), len(tweet_dataset['test'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "large-financing",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at bert-large-uncased were not used when initializing BertForSequenceClassification: ['cls.predictions.transform.LayerNorm.bias', 'cls.predictions.decoder.weight', 'cls.predictions.bias', 'cls.predictions.transform.dense.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.dense.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.weight']\n",
      "- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-large-uncased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    }
   ],
   "source": [
    "# using an uncased model means I am assuming that case will not matter when predicting our categories\n",
    "sequence_classification_model = BertForSequenceClassification.from_pretrained(\n",
    "    BERT_MODEL, num_labels=2,\n",
    "    output_attentions=False,\n",
    "    output_hidden_states=False\n",
    ")\n",
    "\n",
    "sequence_classification_model.config.id2label = {0: 'NOT DISASTER', 1: 'DISASTER'}\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "younger-antique",
   "metadata": {},
   "source": [
    "Let's break down this warning:\n",
    "\n",
    "```\n",
    "Some weights of the model checkpoint at bert-base-uncased were not used when initializing 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",
    "\n",
    "\n",
    "- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
    "- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
    "```\n",
    "\n",
    "This is telling us about the weights that it is not using in this sequence classification version of BERT\n",
    "\n",
    "--\n",
    "\n",
    "```\n",
    "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
    "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n",
    "```\n",
    "\n",
    "The `classifier.weight` and `classifier.bias` are two new matrices we need to learn during fine-tuning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "valid-hundred",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "transformers.models.bert.modeling_bert.BertModel"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(sequence_classification_model.bert)  # normal BERT model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "induced-visit",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "BertForSequenceClassification(\n",
       "  (bert): BertModel(\n",
       "    (embeddings): BertEmbeddings(\n",
       "      (word_embeddings): Embedding(30522, 1024, padding_idx=0)\n",
       "      (position_embeddings): Embedding(512, 1024)\n",
       "      (token_type_embeddings): Embedding(2, 1024)\n",
       "      (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "      (dropout): Dropout(p=0.1, inplace=False)\n",
       "    )\n",
       "    (encoder): BertEncoder(\n",
       "      (layer): ModuleList(\n",
       "        (0): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (1): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (2): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (3): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (4): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (5): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (6): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (7): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (8): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (9): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (10): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (11): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (12): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (13): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (14): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (15): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (16): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (17): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (18): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (19): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (20): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (21): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (22): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "        (23): BertLayer(\n",
       "          (attention): BertAttention(\n",
       "            (self): BertSelfAttention(\n",
       "              (query): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (key): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (value): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "            (output): BertSelfOutput(\n",
       "              (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "              (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "              (dropout): Dropout(p=0.1, inplace=False)\n",
       "            )\n",
       "          )\n",
       "          (intermediate): BertIntermediate(\n",
       "            (dense): Linear(in_features=1024, out_features=4096, bias=True)\n",
       "            (intermediate_act_fn): GELUActivation()\n",
       "          )\n",
       "          (output): BertOutput(\n",
       "            (dense): Linear(in_features=4096, out_features=1024, bias=True)\n",
       "            (LayerNorm): LayerNorm((1024,), eps=1e-12, elementwise_affine=True)\n",
       "            (dropout): Dropout(p=0.1, inplace=False)\n",
       "          )\n",
       "        )\n",
       "      )\n",
       "    )\n",
       "    (pooler): BertPooler(\n",
       "      (dense): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "      (activation): Tanh()\n",
       "    )\n",
       "  )\n",
       "  (dropout): Dropout(p=0.1, inplace=False)\n",
       "  (classifier): Linear(in_features=1024, out_features=2, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sequence_classification_model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "mighty-ordinance",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([2, 1024])\n",
      "torch.Size([2])\n"
     ]
    }
   ],
   "source": [
    "# this classifier is a feed forward layer that is shared across tokens. It squashes the BERT's 768 dimension\n",
    "# into 2 dimensions (disaster or not disaster)\n",
    "print(sequence_classification_model.classifier.weight.shape)\n",
    "\n",
    "print(sequence_classification_model.classifier.bias.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "refined-wrestling",
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = BertTokenizer.from_pretrained(BERT_MODEL)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "plain-championship",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "hill hill mountain volcano of hell mountain hill hil.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "tensor([[[-0.0757, -0.2677, -0.2437,  ..., -0.1609, -0.2434,  0.0849],\n",
       "         [ 0.3218, -0.8701, -1.0655,  ..., -0.5041,  0.7683,  0.6173],\n",
       "         [ 0.1904, -1.1727, -0.6511,  ..., -0.5110,  1.0933,  0.8200],\n",
       "         ...,\n",
       "         [-1.0475,  0.3265, -0.4232,  ...,  0.4013,  0.5853,  0.0258],\n",
       "         [ 0.0767,  0.0371, -0.1873,  ...,  0.4422,  0.4291,  0.4383],\n",
       "         [-0.2622,  0.2595, -0.3092,  ..., -0.1652,  0.3450,  0.2026]]],\n",
       "       grad_fn=<NativeLayerNormBackward0>)"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tweet = tweets['text'][0]\n",
    "print(tweet)\n",
    "\n",
    "token_ids = torch.tensor(tokenizer.encode(tweet)).unsqueeze(0)\n",
    "\n",
    "pre_finetuning_response = sequence_classification_model(\n",
    "    token_ids, output_hidden_states=True\n",
    ")\n",
    "\n",
    "pre_finetuning_response.hidden_states[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "outstanding-denver",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0.8614, 0.1875]], grad_fn=<AddmmBackward0>)"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pre_finetuning_response.logits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bce3aedc",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "c71f6065",
   "metadata": {},
   "outputs": [],
   "source": [
    "# DataCollatorWithPadding creates batch of data. It also dynamically pads text to the \n",
    "#  length of the longest element in the batch, making them all the same length. \n",
    "#  It's possible to pad your text in the tokenizer function with padding=True, dynamic padding is more efficient.\n",
    "data_collator = DataCollatorWithPadding(tokenizer=tokenizer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cb5eec27",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "80e6300e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'input_ids': tensor([[  101,  2158,  2339,  8440,  1005,  1056,  1030,  3698, 12734, 18690,\n",
       "          2100, 10676,  2039,  1029,  2002,  1005,  1055,  2145,  5230,  1012,\n",
       "           102],\n",
       "        [  101, 11561,  1045,  1005,  1049,  6069, 10083,  2017,  2065,  2017,\n",
       "          2292,  2033,  2994,   102,     0,     0,     0,     0,     0,     0,\n",
       "             0]]), 'token_type_ids': tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],\n",
       "        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]), 'attention_mask': tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],\n",
       "        [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0]]), 'labels': tensor([0, 0])}"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch = data_collator(tweet_dataset['train'][:2])\n",
    "\n",
    "batch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "39ba288d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[  101,  2158,  2339,  8440,  1005,  1056,  1030,  3698, 12734, 18690,\n",
       "          2100, 10676,  2039,  1029,  2002,  1005,  1055,  2145,  5230,  1012,\n",
       "           102],\n",
       "        [  101, 11561,  1045,  1005,  1049,  6069, 10083,  2017,  2065,  2017,\n",
       "          2292,  2033,  2994,   102,     0,     0,     0,     0,     0,     0,\n",
       "             0]])"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch['input_ids']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "2aa5dfc7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tokenizer.pad_token_id"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "3cb7fcc1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1],\n",
       "        [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0]])"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch['attention_mask']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "51963510",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],\n",
       "        [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]])"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch['token_type_ids']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "2ebe4684",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([0, 0])"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch['labels']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "e1a20edf",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\"[CLS] babe i'm gonna ruin you if you let me stay [SEP] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD] [PAD]\""
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tokenizer.decode(batch['input_ids'][1])  # see the pad tokens. 0 --> [PAD]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "amino-possibility",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/y9/9xqbqkg90tnc0cmm0dxt985m0000gn/T/ipykernel_26907/2408721198.py:22: FutureWarning: load_metric is deprecated and will be removed in the next major version of datasets. Use 'evaluate.load' instead, from the new library 🤗 Evaluate: https://huggingface.co/docs/evaluate\n",
      "  metric = load_metric(\"accuracy\")\n"
     ]
    }
   ],
   "source": [
    "batch_size = 8  # how many examples to train on every batch\n",
    "epochs = 4  # how times to go over the entire dataset\n",
    "\n",
    "\n",
    "training_args = TrainingArguments(\n",
    "    output_dir='./results',\n",
    "    num_train_epochs=epochs,\n",
    "    per_device_train_batch_size=batch_size,\n",
    "    per_device_eval_batch_size=batch_size,\n",
    "    logging_dir='./logs',\n",
    "    logging_strategy='epoch',\n",
    "    evaluation_strategy='epoch',\n",
    "    save_strategy='epoch',\n",
    "    load_best_model_at_end=True,\n",
    "    logging_steps=1\n",
    ")\n",
    "\n",
    "# Define accuracy metric:\n",
    "\n",
    "from datasets import load_metric\n",
    "\n",
    "metric = load_metric(\"accuracy\")\n",
    "\n",
    "def compute_metrics(eval_pred):  # custom method to take in logits and calculate accuracy of the eval set\n",
    "    logits, labels = eval_pred\n",
    "    predictions = np.argmax(logits, axis=-1)\n",
    "    return metric.compute(predictions=predictions, references=labels)\n",
    "\n",
    "# Define the trainer:\n",
    "\n",
    "trainer = Trainer(\n",
    "    model=sequence_classification_model,\n",
    "    args=training_args,\n",
    "    data_collator=data_collator,\n",
    "    train_dataset=tweet_dataset['train'],\n",
    "    eval_dataset=tweet_dataset['test'],\n",
    "    compute_metrics=compute_metrics\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "9ccee263",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='10' max='5' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [5/5 00:44]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "{'eval_loss': 0.6992014646530151,\n",
       " 'eval_accuracy': 0.575,\n",
       " 'eval_runtime': 2.2597,\n",
       " 'eval_samples_per_second': 17.701,\n",
       " 'eval_steps_per_second': 2.213}"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Get initial metrics including accuracy\n",
    "trainer.evaluate()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "considerable-horse",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/homebrew/lib/python3.10/site-packages/transformers/optimization.py:306: FutureWarning: This implementation of AdamW is deprecated and will be removed in a future version. Use the PyTorch implementation torch.optim.AdamW instead, or set `no_deprecation_warning=True` to disable this warning\n",
      "  warnings.warn(\n",
      "***** Running training *****\n",
      "  Num examples = 160\n",
      "  Num Epochs = 4\n",
      "  Instantaneous batch size per device = 8\n",
      "  Total train batch size (w. parallel, distributed & accumulation) = 8\n",
      "  Gradient Accumulation steps = 1\n",
      "  Total optimization steps = 80\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='80' max='80' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [80/80 03:21, Epoch 4/4]\n",
       "    </div>\n",
       "    <table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       " <tr style=\"text-align: left;\">\n",
       "      <th>Epoch</th>\n",
       "      <th>Training Loss</th>\n",
       "      <th>Validation Loss</th>\n",
       "      <th>Accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.643900</td>\n",
       "      <td>0.689123</td>\n",
       "      <td>0.575000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.550000</td>\n",
       "      <td>0.563866</td>\n",
       "      <td>0.750000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.407600</td>\n",
       "      <td>0.688863</td>\n",
       "      <td>0.700000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.225400</td>\n",
       "      <td>0.659879</td>\n",
       "      <td>0.775000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table><p>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n",
      "Saving model checkpoint to ./results/checkpoint-20\n",
      "Configuration saved in ./results/checkpoint-20/config.json\n",
      "Model weights saved in ./results/checkpoint-20/pytorch_model.bin\n",
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n",
      "Saving model checkpoint to ./results/checkpoint-40\n",
      "Configuration saved in ./results/checkpoint-40/config.json\n",
      "Model weights saved in ./results/checkpoint-40/pytorch_model.bin\n",
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n",
      "Saving model checkpoint to ./results/checkpoint-60\n",
      "Configuration saved in ./results/checkpoint-60/config.json\n",
      "Model weights saved in ./results/checkpoint-60/pytorch_model.bin\n",
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n",
      "Saving model checkpoint to ./results/checkpoint-80\n",
      "Configuration saved in ./results/checkpoint-80/config.json\n",
      "Model weights saved in ./results/checkpoint-80/pytorch_model.bin\n",
      "\n",
      "\n",
      "Training completed. Do not forget to share your model on huggingface.co/models =)\n",
      "\n",
      "\n",
      "Loading best model from ./results/checkpoint-40 (score: 0.5638656616210938).\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "TrainOutput(global_step=80, training_loss=0.45673776865005494, metrics={'train_runtime': 205.0346, 'train_samples_per_second': 3.121, 'train_steps_per_second': 0.39, 'total_flos': 57546761685504.0, 'train_loss': 0.45673776865005494, 'epoch': 4.0})"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trainer.train()  # Train our model!!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "alone-florence",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='5' max='5' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [5/5 00:01]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "{'eval_loss': 0.5638656616210938,\n",
       " 'eval_accuracy': 0.75,\n",
       " 'eval_runtime': 2.1564,\n",
       " 'eval_samples_per_second': 18.55,\n",
       " 'eval_steps_per_second': 2.319,\n",
       " 'epoch': 4.0}"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trainer.evaluate()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "circular-toner",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "hill hill mountain volcano of hell mountain hill hil.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "tensor(False)"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(tweet)\n",
    "\n",
    "post_finetuning_response = sequence_classification_model(\n",
    "    token_ids, output_hidden_states=True\n",
    ")\n",
    "\n",
    "# hidden states are different from before fine tuning\n",
    "(post_finetuning_response.hidden_states[0] == pre_finetuning_response.hidden_states[0]).all()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "limited-unknown",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[[-0.0813, -0.2742, -0.2493,  ..., -0.1606, -0.2449,  0.0820],\n",
       "         [ 0.3270, -0.8662, -1.0803,  ..., -0.5040,  0.7609,  0.6166],\n",
       "         [ 0.1879, -1.1690, -0.6640,  ..., -0.5106,  1.0818,  0.8172],\n",
       "         ...,\n",
       "         [-1.0592,  0.3298, -0.4187,  ...,  0.4046,  0.5776,  0.0149],\n",
       "         [ 0.0753,  0.0328, -0.1975,  ...,  0.4356,  0.4278,  0.4396],\n",
       "         [-0.2665,  0.2671, -0.3160,  ..., -0.1716,  0.3324,  0.1828]]],\n",
       "       grad_fn=<NativeLayerNormBackward0>)"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "post_finetuning_response.hidden_states[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "coordinated-anaheim",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[-0.2680,  0.6904]], grad_fn=<AddmmBackward0>)"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "post_finetuning_response.logits"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "4c4e64ed",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Saving model checkpoint to ./results\n",
      "Configuration saved in ./results/config.json\n",
      "Model weights saved in ./results/pytorch_model.bin\n"
     ]
    }
   ],
   "source": [
    "trainer.save_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "4c57bc0d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "loading configuration file ./results/config.json\n",
      "Model config BertConfig {\n",
      "  \"_name_or_path\": \"./results\",\n",
      "  \"architectures\": [\n",
      "    \"BertForSequenceClassification\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"gradient_checkpointing\": false,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 1024,\n",
      "  \"id2label\": {\n",
      "    \"0\": \"NOT DISASTER\",\n",
      "    \"1\": \"DISASTER\"\n",
      "  },\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 4096,\n",
      "  \"label2id\": null,\n",
      "  \"layer_norm_eps\": 1e-12,\n",
      "  \"max_position_embeddings\": 512,\n",
      "  \"model_type\": \"bert\",\n",
      "  \"num_attention_heads\": 16,\n",
      "  \"num_hidden_layers\": 24,\n",
      "  \"pad_token_id\": 0,\n",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"problem_type\": \"single_label_classification\",\n",
      "  \"torch_dtype\": \"float32\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 30522\n",
      "}\n",
      "\n",
      "loading configuration file ./results/config.json\n",
      "Model config BertConfig {\n",
      "  \"_name_or_path\": \"./results\",\n",
      "  \"architectures\": [\n",
      "    \"BertForSequenceClassification\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"gradient_checkpointing\": false,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 1024,\n",
      "  \"id2label\": {\n",
      "    \"0\": \"NOT DISASTER\",\n",
      "    \"1\": \"DISASTER\"\n",
      "  },\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 4096,\n",
      "  \"label2id\": null,\n",
      "  \"layer_norm_eps\": 1e-12,\n",
      "  \"max_position_embeddings\": 512,\n",
      "  \"model_type\": \"bert\",\n",
      "  \"num_attention_heads\": 16,\n",
      "  \"num_hidden_layers\": 24,\n",
      "  \"pad_token_id\": 0,\n",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"problem_type\": \"single_label_classification\",\n",
      "  \"torch_dtype\": \"float32\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 30522\n",
      "}\n",
      "\n",
      "loading weights file ./results/pytorch_model.bin\n",
      "All model checkpoint weights were used when initializing BertForSequenceClassification.\n",
      "\n",
      "All the weights of BertForSequenceClassification were initialized from the model checkpoint at ./results.\n",
      "If your task is similar to the task the model of the checkpoint was trained on, you can already use BertForSequenceClassification for predictions without further training.\n",
      "loading configuration file config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased/snapshots/80792f8e8216b29f3c846b653a0ff0a37c210431/config.json\n",
      "Model config BertConfig {\n",
      "  \"_name_or_path\": \"bert-large-uncased\",\n",
      "  \"architectures\": [\n",
      "    \"BertForMaskedLM\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"gradient_checkpointing\": false,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 1024,\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 4096,\n",
      "  \"layer_norm_eps\": 1e-12,\n",
      "  \"max_position_embeddings\": 512,\n",
      "  \"model_type\": \"bert\",\n",
      "  \"num_attention_heads\": 16,\n",
      "  \"num_hidden_layers\": 24,\n",
      "  \"pad_token_id\": 0,\n",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 30522\n",
      "}\n",
      "\n",
      "loading file vocab.txt from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased/snapshots/80792f8e8216b29f3c846b653a0ff0a37c210431/vocab.txt\n",
      "loading file tokenizer.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased/snapshots/80792f8e8216b29f3c846b653a0ff0a37c210431/tokenizer.json\n",
      "loading file added_tokens.json from cache at None\n",
      "loading file special_tokens_map.json from cache at None\n",
      "loading file tokenizer_config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased/snapshots/80792f8e8216b29f3c846b653a0ff0a37c210431/tokenizer_config.json\n",
      "loading configuration file config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased/snapshots/80792f8e8216b29f3c846b653a0ff0a37c210431/config.json\n",
      "Model config BertConfig {\n",
      "  \"_name_or_path\": \"bert-large-uncased\",\n",
      "  \"architectures\": [\n",
      "    \"BertForMaskedLM\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"gradient_checkpointing\": false,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 1024,\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 4096,\n",
      "  \"layer_norm_eps\": 1e-12,\n",
      "  \"max_position_embeddings\": 512,\n",
      "  \"model_type\": \"bert\",\n",
      "  \"num_attention_heads\": 16,\n",
      "  \"num_hidden_layers\": 24,\n",
      "  \"pad_token_id\": 0,\n",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 30522\n",
      "}\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# make a classification pipeline\n",
    "pipe = pipeline(\"text-classification\", './results', tokenizer=BERT_MODEL)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "944adf6e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'label': 'NOT DISASTER', 'score': 0.7597532868385315}]"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pipe('Such a terrible earthquake')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "historical-despite",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "persistent-mission",
   "metadata": {},
   "source": [
    "Let's try that sequence again but this time, let's freeze the layers of BERT to try and speed up training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "joined-voluntary",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "loading configuration file config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased/snapshots/80792f8e8216b29f3c846b653a0ff0a37c210431/config.json\n",
      "Model config BertConfig {\n",
      "  \"architectures\": [\n",
      "    \"BertForMaskedLM\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"gradient_checkpointing\": false,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 1024,\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 4096,\n",
      "  \"layer_norm_eps\": 1e-12,\n",
      "  \"max_position_embeddings\": 512,\n",
      "  \"model_type\": \"bert\",\n",
      "  \"num_attention_heads\": 16,\n",
      "  \"num_hidden_layers\": 24,\n",
      "  \"pad_token_id\": 0,\n",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 30522\n",
      "}\n",
      "\n",
      "loading weights file pytorch_model.bin from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased/snapshots/80792f8e8216b29f3c846b653a0ff0a37c210431/pytorch_model.bin\n",
      "Some weights of the model checkpoint at bert-large-uncased were not used when initializing BertForSequenceClassification: ['cls.predictions.transform.LayerNorm.bias', 'cls.predictions.decoder.weight', 'cls.predictions.bias', 'cls.predictions.transform.dense.weight', 'cls.seq_relationship.bias', 'cls.predictions.transform.dense.bias', 'cls.predictions.transform.LayerNorm.weight', 'cls.seq_relationship.weight']\n",
      "- This IS expected if you are initializing BertForSequenceClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing BertForSequenceClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-large-uncased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    }
   ],
   "source": [
    "# re-instantiate BERT\n",
    "new_sequence_classification_model = BertForSequenceClassification.from_pretrained(\n",
    "    BERT_MODEL, num_labels=2,\n",
    "    output_attentions=False,\n",
    "    output_hidden_states=False\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "russian-combine",
   "metadata": {},
   "outputs": [],
   "source": [
    "# freeze all layers in BERT to speed up training\n",
    "for name, param in new_sequence_classification_model.bert.named_parameters():\n",
    "    param.requires_grad = False  # disable training in BERT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0b5b196b",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c5400edc",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "aerial-framework",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='10' max='5' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [5/5 00:13]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "{'eval_loss': 0.7674192190170288,\n",
       " 'eval_accuracy': 0.425,\n",
       " 'eval_runtime': 1.9928,\n",
       " 'eval_samples_per_second': 20.073,\n",
       " 'eval_steps_per_second': 2.509}"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Change the model in the trainer to be our newly instantiated model\n",
    "\n",
    "trainer = Trainer(\n",
    "    model=new_sequence_classification_model,\n",
    "    args=training_args,\n",
    "    data_collator=data_collator,\n",
    "    train_dataset=tweet_dataset['train'],\n",
    "    eval_dataset=tweet_dataset['test'],\n",
    "    compute_metrics=compute_metrics\n",
    ")\n",
    "\n",
    "# Get initial metrics. Similar as before. Check: Why isn't it the same loss?\n",
    "trainer.evaluate()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "opposite-publicity",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/homebrew/lib/python3.10/site-packages/transformers/optimization.py:306: FutureWarning: This implementation of AdamW is deprecated and will be removed in a future version. Use the PyTorch implementation torch.optim.AdamW instead, or set `no_deprecation_warning=True` to disable this warning\n",
      "  warnings.warn(\n",
      "***** Running training *****\n",
      "  Num examples = 160\n",
      "  Num Epochs = 4\n",
      "  Instantaneous batch size per device = 8\n",
      "  Total train batch size (w. parallel, distributed & accumulation) = 8\n",
      "  Gradient Accumulation steps = 1\n",
      "  Total optimization steps = 80\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='80' max='80' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [80/80 00:51, Epoch 4/4]\n",
       "    </div>\n",
       "    <table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       " <tr style=\"text-align: left;\">\n",
       "      <th>Epoch</th>\n",
       "      <th>Training Loss</th>\n",
       "      <th>Validation Loss</th>\n",
       "      <th>Accuracy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.735200</td>\n",
       "      <td>0.707490</td>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.713200</td>\n",
       "      <td>0.707488</td>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.701900</td>\n",
       "      <td>0.708359</td>\n",
       "      <td>0.525000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.713000</td>\n",
       "      <td>0.707446</td>\n",
       "      <td>0.500000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table><p>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n",
      "Saving model checkpoint to ./results/checkpoint-20\n",
      "Configuration saved in ./results/checkpoint-20/config.json\n",
      "Model weights saved in ./results/checkpoint-20/pytorch_model.bin\n",
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n",
      "Saving model checkpoint to ./results/checkpoint-40\n",
      "Configuration saved in ./results/checkpoint-40/config.json\n",
      "Model weights saved in ./results/checkpoint-40/pytorch_model.bin\n",
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n",
      "Saving model checkpoint to ./results/checkpoint-60\n",
      "Configuration saved in ./results/checkpoint-60/config.json\n",
      "Model weights saved in ./results/checkpoint-60/pytorch_model.bin\n",
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n",
      "Saving model checkpoint to ./results/checkpoint-80\n",
      "Configuration saved in ./results/checkpoint-80/config.json\n",
      "Model weights saved in ./results/checkpoint-80/pytorch_model.bin\n",
      "\n",
      "\n",
      "Training completed. Do not forget to share your model on huggingface.co/models =)\n",
      "\n",
      "\n",
      "Loading best model from ./results/checkpoint-80 (score: 0.7074458003044128).\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "TrainOutput(global_step=80, training_loss=0.7158524632453919, metrics={'train_runtime': 51.995, 'train_samples_per_second': 12.309, 'train_steps_per_second': 1.539, 'total_flos': 57546761685504.0, 'train_loss': 0.7158524632453919, 'epoch': 4.0})"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trainer.train()  # MUCH faster training time but training did not go as well!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "induced-addiction",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "***** Running Evaluation *****\n",
      "  Num examples = 40\n",
      "  Batch size = 8\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "      \n",
       "      <progress value='5' max='5' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      [5/5 00:01]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "{'eval_loss': 0.7074458003044128,\n",
       " 'eval_accuracy': 0.5,\n",
       " 'eval_runtime': 1.9442,\n",
       " 'eval_samples_per_second': 20.574,\n",
       " 'eval_steps_per_second': 2.572,\n",
       " 'epoch': 4.0}"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trainer.evaluate()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "liberal-champion",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(True)"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "post_finetuning_response = new_sequence_classification_model(\n",
    "    token_ids, output_hidden_states=True\n",
    ")\n",
    "\n",
    "# hidden states are same as before fine tuning\n",
    "(post_finetuning_response.hidden_states[0] == pre_finetuning_response.hidden_states[0]).all()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "million-pathology",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(tensor([[0.8614, 0.1875]], grad_fn=<AddmmBackward0>),\n",
       " tensor([[ 0.0131, -0.1344]], grad_fn=<AddmmBackward0>))"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# the logits are different as the pre-trained model because the classifier weights are initialized randomly every time\n",
    "#  this is also why the initial logits weren't the same\n",
    "\n",
    "pre_finetuning_response.logits, post_finetuning_response.logits  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "varying-actor",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "sustainable-field",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "dimensional-drain",
   "metadata": {},
   "source": [
    "# Example 2. Question/Answering with a model already fine-tuned using SQuAD\n",
    "\n",
    "\n",
    "![BERT QA](images/bert_qa.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "norwegian-fortune",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "loading file vocab.txt from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased/snapshots/80792f8e8216b29f3c846b653a0ff0a37c210431/vocab.txt\n",
      "loading file added_tokens.json from cache at None\n",
      "loading file special_tokens_map.json from cache at None\n",
      "loading file tokenizer_config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased/snapshots/80792f8e8216b29f3c846b653a0ff0a37c210431/tokenizer_config.json\n",
      "loading configuration file config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased/snapshots/80792f8e8216b29f3c846b653a0ff0a37c210431/config.json\n",
      "Model config BertConfig {\n",
      "  \"_name_or_path\": \"bert-large-uncased\",\n",
      "  \"architectures\": [\n",
      "    \"BertForMaskedLM\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"gradient_checkpointing\": false,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 1024,\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 4096,\n",
      "  \"layer_norm_eps\": 1e-12,\n",
      "  \"max_position_embeddings\": 512,\n",
      "  \"model_type\": \"bert\",\n",
      "  \"num_attention_heads\": 16,\n",
      "  \"num_hidden_layers\": 24,\n",
      "  \"pad_token_id\": 0,\n",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 30522\n",
      "}\n",
      "\n",
      "loading configuration file config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased-whole-word-masking-finetuned-squad/snapshots/242d9dbb66bb5033025196d5678907307f8fb098/config.json\n",
      "Model config BertConfig {\n",
      "  \"architectures\": [\n",
      "    \"BertForQuestionAnswering\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 1024,\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 4096,\n",
      "  \"layer_norm_eps\": 1e-12,\n",
      "  \"max_position_embeddings\": 512,\n",
      "  \"model_type\": \"bert\",\n",
      "  \"num_attention_heads\": 16,\n",
      "  \"num_hidden_layers\": 24,\n",
      "  \"pad_token_id\": 0,\n",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 30522\n",
      "}\n",
      "\n",
      "loading weights file pytorch_model.bin from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--bert-large-uncased-whole-word-masking-finetuned-squad/snapshots/242d9dbb66bb5033025196d5678907307f8fb098/pytorch_model.bin\n",
      "All model checkpoint weights were used when initializing BertForQuestionAnswering.\n",
      "\n",
      "All the weights of BertForQuestionAnswering were initialized from the model checkpoint at bert-large-uncased-whole-word-masking-finetuned-squad.\n",
      "If your task is similar to the task the model of the checkpoint was trained on, you can already use BertForQuestionAnswering for predictions without further training.\n"
     ]
    }
   ],
   "source": [
    "# there are pre fine-tuned models that learned from the SQuAD dataset\n",
    "\n",
    "large_bert_tokenizer = BertTokenizer.from_pretrained('bert-large-uncased', return_token_type_ids=True)\n",
    "qa_bert = BertForQuestionAnswering.from_pretrained('bert-large-uncased-whole-word-masking-finetuned-squad')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "distinguished-element",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "transformers.models.bert.modeling_bert.BertModel"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(qa_bert.bert)  # BERT model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "optimum-arena",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of layers in the encoder stack: 24\n"
     ]
    }
   ],
   "source": [
    "print(f'number of layers in the encoder stack: {len(qa_bert.bert.encoder.layer)}')  #  LARGE BERT"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "active-involvement",
   "metadata": {},
   "source": [
    "### Count the Parameters in Large BERT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "comparative-franklin",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Parameters: 333,770,752\n"
     ]
    }
   ],
   "source": [
    "total_params = 0\n",
    "for p in qa_bert.bert.parameters():\n",
    "    if len(p.shape) == 2:\n",
    "        total_params += p.shape[0] * p.shape[1]\n",
    "        \n",
    "print(f'Total Parameters: {total_params:,}')  # This is where the 110M parameter comes from"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "occupied-guard",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The BERT model has 391 different named parameters.\n",
      "\n",
      "==== Embedding Layer ====\n",
      "\n",
      "bert.embeddings.word_embeddings.weight                  (30522, 1024)\n",
      "bert.embeddings.position_embeddings.weight               (512, 1024)\n",
      "bert.embeddings.token_type_embeddings.weight               (2, 1024)\n",
      "bert.embeddings.LayerNorm.weight                             (1024,)\n",
      "bert.embeddings.LayerNorm.bias                               (1024,)\n",
      "\n",
      "==== First Transformer ====\n",
      "\n",
      "bert.encoder.layer.0.attention.self.query.weight        (1024, 1024)\n",
      "bert.encoder.layer.0.attention.self.query.bias               (1024,)\n",
      "bert.encoder.layer.0.attention.self.key.weight          (1024, 1024)\n",
      "bert.encoder.layer.0.attention.self.key.bias                 (1024,)\n",
      "bert.encoder.layer.0.attention.self.value.weight        (1024, 1024)\n",
      "bert.encoder.layer.0.attention.self.value.bias               (1024,)\n",
      "bert.encoder.layer.0.attention.output.dense.weight      (1024, 1024)\n",
      "bert.encoder.layer.0.attention.output.dense.bias             (1024,)\n",
      "bert.encoder.layer.0.attention.output.LayerNorm.weight       (1024,)\n",
      "bert.encoder.layer.0.attention.output.LayerNorm.bias         (1024,)\n",
      "bert.encoder.layer.0.intermediate.dense.weight          (4096, 1024)\n",
      "bert.encoder.layer.0.intermediate.dense.bias                 (4096,)\n",
      "bert.encoder.layer.0.output.dense.weight                (1024, 4096)\n",
      "bert.encoder.layer.0.output.dense.bias                       (1024,)\n",
      "bert.encoder.layer.0.output.LayerNorm.weight                 (1024,)\n",
      "bert.encoder.layer.0.output.LayerNorm.bias                   (1024,)\n",
      "\n",
      "==== Output Layer ====\n",
      "\n",
      "qa_outputs.weight                                          (2, 1024)\n",
      "qa_outputs.bias                                                 (2,)\n"
     ]
    }
   ],
   "source": [
    "# Get all of the model's parameters as a list of tuples.\n",
    "params = list(qa_bert.named_parameters())\n",
    "\n",
    "print('The BERT model has {:} different named parameters.\\n'.format(len(params)))\n",
    "\n",
    "print('==== Embedding Layer ====\\n')\n",
    "\n",
    "for p in params[0:5]:\n",
    "    print(\"{:<55} {:>12}\".format(p[0], str(tuple(p[1].size()))))\n",
    "\n",
    "print('\\n==== First Transformer ====\\n')\n",
    "\n",
    "for p in params[5:21]:\n",
    "    print(\"{:<55} {:>12}\".format(p[0], str(tuple(p[1].size()))))\n",
    "\n",
    "print('\\n==== Output Layer ====\\n')\n",
    "\n",
    "for p in params[-2:]:\n",
    "    print(\"{:<55} {:>12}\".format(p[0], str(tuple(p[1].size()))))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "handy-grammar",
   "metadata": {},
   "outputs": [],
   "source": [
    "# this qa_outputs is a feed forward layer that is shared across tokens. It squashes the BERT-large's 1024 dimension\n",
    "# into 2 dimensions (start position and end position)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "copyrighted-allergy",
   "metadata": {},
   "outputs": [],
   "source": [
    "sns.set(rc={\"figure.figsize\":(20, 5)}) \n",
    "\n",
    "def q_a(question, context):\n",
    "    encodings = large_bert_tokenizer.encode_plus(question, context)\n",
    "\n",
    "    inputIds, attentionMask = encodings[\"input_ids\"], encodings[\"attention_mask\"]\n",
    "\n",
    "    response = qa_bert(torch.tensor([inputIds]), attention_mask=torch.tensor([attentionMask]))\n",
    "    scoresStart, scoresEnd = response['start_logits'], response['end_logits']\n",
    "    tokens = inputIds[torch.argmax(scoresStart): torch.argmax(scoresEnd) + 1]\n",
    "    answerTokens = large_bert_tokenizer.convert_ids_to_tokens(tokens, skip_special_tokens=True)\n",
    "    \n",
    "    token_labels = large_bert_tokenizer.convert_ids_to_tokens(inputIds)\n",
    "    \n",
    "    \n",
    "    # bar plot code courtesy of Chris McCormick\n",
    "    # https://mccormickml.com/2020/03/10/question-answering-with-a-fine-tuned-BERT\n",
    "    \n",
    "    # Create a barplot showing the start word score for all of the tokens.\n",
    "    ax = sns.barplot(x=[f'{i}_{t}' for i, t in enumerate(token_labels)], y=scoresStart.squeeze().tolist(), ci=None)\n",
    "    # Turn the xlabels vertical.\n",
    "    ax.set_xticklabels(ax.get_xticklabels(), rotation=90, ha=\"center\")\n",
    "    # Turn on the vertical grid to help align words to scores.\n",
    "    plt.title('Start Word Scores')\n",
    "    \n",
    "    plt.show()\n",
    "    \n",
    "    # Create a barplot showing the start word score for all of the tokens.\n",
    "    ax = sns.barplot(x=[f'{i}_{t}' for i, t in enumerate(token_labels)], y=scoresEnd.squeeze().tolist(), ci=None)\n",
    "    # Turn the xlabels vertical.\n",
    "    ax.set_xticklabels(ax.get_xticklabels(), rotation=90, ha=\"center\")\n",
    "    # Turn on thaae vertical grid to help align words to scores.\n",
    "    plt.title('End Word Scores')\n",
    "\n",
    "    plt.show()\n",
    "    \n",
    "    return large_bert_tokenizer.convert_tokens_to_string(answerTokens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "hydraulic-detective",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/y9/9xqbqkg90tnc0cmm0dxt985m0000gn/T/ipykernel_26907/2807390656.py:20: FutureWarning: \n",
      "\n",
      "The `ci` parameter is deprecated. Use `errorbar=None` for the same effect.\n",
      "\n",
      "  ax = sns.barplot(x=[f'{i}_{t}' for i, t in enumerate(token_labels)], y=scoresStart.squeeze().tolist(), ci=None)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2000x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/y9/9xqbqkg90tnc0cmm0dxt985m0000gn/T/ipykernel_26907/2807390656.py:29: FutureWarning: \n",
      "\n",
      "The `ci` parameter is deprecated. Use `errorbar=None` for the same effect.\n",
      "\n",
      "  ax = sns.barplot(x=[f'{i}_{t}' for i, t in enumerate(token_labels)], y=scoresEnd.squeeze().tolist(), ci=None)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2000x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'canada'"
      ]
     },
     "execution_count": 63,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "q_a(\"Where does Matt live?\", \"Matt lives in Canada and Sinan lives in California\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "commercial-basement",
   "metadata": {},
   "outputs": [],
   "source": [
    "# From https://docs.microsoft.com/en-us/azure/databricks/kb/jobs/active-vs-dead-jobs\n",
    "\n",
    "longer_context = \"\"\"\n",
    "Whenever there are too many concurrent jobs running on a cluster, there is a chance that the Spark internal \n",
    "eventListenerBus drops events. These events are used to track job progress in the Spark UI. Whenever the event \n",
    "listener drops events you start seeing dead jobs/stages in Spark UI, which never finish. \n",
    "The jobs are actually finished but not shown as completed in the Spark UI.\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "fitting-tribute",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "78\n",
      "Token: 101, subword: [CLS]\n",
      "Token: 7188, subword: whenever\n",
      "Token: 2045, subword: there\n",
      "Token: 2024, subword: are\n",
      "Token: 2205, subword: too\n",
      "Token: 2116, subword: many\n",
      "Token: 16483, subword: concurrent\n",
      "Token: 5841, subword: jobs\n",
      "Token: 2770, subword: running\n",
      "Token: 2006, subword: on\n",
      "Token: 1037, subword: a\n",
      "Token: 9324, subword: cluster\n",
      "Token: 1010, subword: ,\n",
      "Token: 2045, subword: there\n",
      "Token: 2003, subword: is\n",
      "Token: 1037, subword: a\n",
      "Token: 3382, subword: chance\n",
      "Token: 2008, subword: that\n",
      "Token: 1996, subword: the\n",
      "Token: 12125, subword: spark\n",
      "Token: 4722, subword: internal\n",
      "Token: 2724, subword: event\n",
      "Token: 9863, subword: ##list\n",
      "Token: 24454, subword: ##ener\n",
      "Token: 8286, subword: ##bus\n",
      "Token: 9010, subword: drops\n",
      "Token: 2824, subword: events\n",
      "Token: 1012, subword: .\n",
      "Token: 2122, subword: these\n",
      "Token: 2824, subword: events\n"
     ]
    }
   ],
   "source": [
    "print(len(tokenizer(longer_context)['input_ids']))  # 78 tokens\n",
    "      \n",
    "for t in tokenizer(longer_context)['input_ids'][:30]:\n",
    "    print(f'Token: {t}, subword: {tokenizer.decode([t])}')\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "enhanced-renewal",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/y9/9xqbqkg90tnc0cmm0dxt985m0000gn/T/ipykernel_26907/2807390656.py:20: FutureWarning: \n",
      "\n",
      "The `ci` parameter is deprecated. Use `errorbar=None` for the same effect.\n",
      "\n",
      "  ax = sns.barplot(x=[f'{i}_{t}' for i, t in enumerate(token_labels)], y=scoresStart.squeeze().tolist(), ci=None)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2000x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/y9/9xqbqkg90tnc0cmm0dxt985m0000gn/T/ipykernel_26907/2807390656.py:29: FutureWarning: \n",
      "\n",
      "The `ci` parameter is deprecated. Use `errorbar=None` for the same effect.\n",
      "\n",
      "  ax = sns.barplot(x=[f'{i}_{t}' for i, t in enumerate(token_labels)], y=scoresEnd.squeeze().tolist(), ci=None)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 2000x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "'there is a chance that the spark internal eventlistenerbus drops events'"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "q_a(\"what if I run a bunch of jobs at the same time?\", longer_context)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bigger-wheel",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "respected-bunny",
   "metadata": {},
   "source": [
    "# Working with flavors of BERT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "maritime-visitor",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "loading configuration file config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--distilbert-base-uncased/snapshots/1c4513b2eedbda136f57676a34eea67aba266e5c/config.json\n",
      "Model config DistilBertConfig {\n",
      "  \"activation\": \"gelu\",\n",
      "  \"architectures\": [\n",
      "    \"DistilBertForMaskedLM\"\n",
      "  ],\n",
      "  \"attention_dropout\": 0.1,\n",
      "  \"dim\": 768,\n",
      "  \"dropout\": 0.1,\n",
      "  \"hidden_dim\": 3072,\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"max_position_embeddings\": 512,\n",
      "  \"model_type\": \"distilbert\",\n",
      "  \"n_heads\": 12,\n",
      "  \"n_layers\": 6,\n",
      "  \"pad_token_id\": 0,\n",
      "  \"qa_dropout\": 0.1,\n",
      "  \"seq_classif_dropout\": 0.2,\n",
      "  \"sinusoidal_pos_embds\": false,\n",
      "  \"tie_weights_\": true,\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"vocab_size\": 30522\n",
      "}\n",
      "\n",
      "loading weights file pytorch_model.bin from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--distilbert-base-uncased/snapshots/1c4513b2eedbda136f57676a34eea67aba266e5c/pytorch_model.bin\n",
      "Some weights of the model checkpoint at distilbert-base-uncased were not used when initializing DistilBertForTokenClassification: ['vocab_transform.bias', 'vocab_layer_norm.bias', 'vocab_layer_norm.weight', 'vocab_projector.weight', 'vocab_transform.weight', 'vocab_projector.bias']\n",
      "- This IS expected if you are initializing DistilBertForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing DistilBertForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "Some weights of DistilBertForTokenClassification were not initialized from the model checkpoint at distilbert-base-uncased and are newly initialized: ['classifier.weight', 'classifier.bias']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    }
   ],
   "source": [
    "distil_model = DistilBertForTokenClassification.from_pretrained('distilbert-base-uncased')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "fitted-incident",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "transformers.models.distilbert.modeling_distilbert.DistilBertForTokenClassification"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(distil_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "stainless-belize",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The DistilBERT model has 102 different named parameters.\n",
      "\n",
      "==== Embedding Layer ====\n",
      "\n",
      "distilbert.embeddings.word_embeddings.weight            (30522, 768)\n",
      "distilbert.embeddings.position_embeddings.weight          (512, 768)\n",
      "distilbert.embeddings.LayerNorm.weight                        (768,)\n",
      "distilbert.embeddings.LayerNorm.bias                          (768,)\n",
      "\n",
      "==== First Transformer ====\n",
      "\n",
      "distilbert.transformer.layer.0.attention.q_lin.bias           (768,)\n",
      "distilbert.transformer.layer.0.attention.k_lin.weight     (768, 768)\n",
      "distilbert.transformer.layer.0.attention.k_lin.bias           (768,)\n",
      "distilbert.transformer.layer.0.attention.v_lin.weight     (768, 768)\n",
      "distilbert.transformer.layer.0.attention.v_lin.bias           (768,)\n",
      "distilbert.transformer.layer.0.attention.out_lin.weight   (768, 768)\n",
      "distilbert.transformer.layer.0.attention.out_lin.bias         (768,)\n",
      "distilbert.transformer.layer.0.sa_layer_norm.weight           (768,)\n",
      "distilbert.transformer.layer.0.sa_layer_norm.bias             (768,)\n",
      "distilbert.transformer.layer.0.ffn.lin1.weight           (3072, 768)\n",
      "distilbert.transformer.layer.0.ffn.lin1.bias                 (3072,)\n",
      "distilbert.transformer.layer.0.ffn.lin2.weight           (768, 3072)\n",
      "distilbert.transformer.layer.0.ffn.lin2.bias                  (768,)\n",
      "distilbert.transformer.layer.0.output_layer_norm.weight       (768,)\n",
      "distilbert.transformer.layer.0.output_layer_norm.bias         (768,)\n",
      "\n",
      "==== Output Layer ====\n",
      "\n",
      "classifier.weight                                           (2, 768)\n",
      "classifier.bias                                                 (2,)\n"
     ]
    }
   ],
   "source": [
    "# Get all of the model's parameters as a list of tuples.\n",
    "named_params = list(distil_model.named_parameters())\n",
    "\n",
    "print('The DistilBERT model has {:} different named parameters.\\n'.format(len(named_params)))\n",
    "\n",
    "print('==== Embedding Layer ====\\n')\n",
    "for p in named_params[0:4]:\n",
    "    print(\"{:<55} {:>12}\".format(p[0], str(tuple(p[1].size()))))\n",
    "\n",
    "print('\\n==== First Transformer ====\\n')\n",
    "for p in named_params[5:20]:\n",
    "    print(\"{:<55} {:>12}\".format(p[0], str(tuple(p[1].size()))))\n",
    "\n",
    "print('\\n==== Output Layer ====\\n')\n",
    "for p in named_params[-2:]:\n",
    "    print(\"{:<55} {:>12}\".format(p[0], str(tuple(p[1].size()))))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "smart-layer",
   "metadata": {},
   "source": [
    "# Count the Parameters in DistilBERT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "preceding-internship",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Parameters: 66,302,976\n"
     ]
    }
   ],
   "source": [
    "total_params = 0\n",
    "for p in distil_model.parameters():\n",
    "    if len(p.shape) == 2:\n",
    "        total_params += p.shape[0] * p.shape[1]\n",
    "        \n",
    "print(f'Total Parameters: {total_params:,}')  # This is where the 110M parameter comes from"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "personal-thesaurus",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "advanced-thinking",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "suspected-brave",
   "metadata": {},
   "source": [
    "![Transformers](images/hf_models.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "97b29596",
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import AutoModelForSequenceClassification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "complimentary-demographic",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "loading configuration file config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--savasy--bert-base-turkish-ner-cased/snapshots/d2853558b8a3b19639dce6da2d8a5b6d8f0102a0/config.json\n",
      "Model config BertConfig {\n",
      "  \"_name_or_path\": \"savasy/bert-base-turkish-ner-cased\",\n",
      "  \"_num_labels\": 7,\n",
      "  \"architectures\": [\n",
      "    \"BertForTokenClassification\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 768,\n",
      "  \"id2label\": {\n",
      "    \"0\": \"B-LOC\",\n",
      "    \"1\": \"B-ORG\",\n",
      "    \"2\": \"B-PER\",\n",
      "    \"3\": \"I-LOC\",\n",
      "    \"4\": \"I-ORG\",\n",
      "    \"5\": \"I-PER\",\n",
      "    \"6\": \"O\"\n",
      "  },\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 3072,\n",
      "  \"label2id\": {\n",
      "    \"B-LOC\": 0,\n",
      "    \"B-ORG\": 1,\n",
      "    \"B-PER\": 2,\n",
      "    \"I-LOC\": 3,\n",
      "    \"I-ORG\": 4,\n",
      "    \"I-PER\": 5,\n",
      "    \"O\": 6\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",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 32000\n",
      "}\n",
      "\n",
      "loading file vocab.txt from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--savasy--bert-base-turkish-ner-cased/snapshots/d2853558b8a3b19639dce6da2d8a5b6d8f0102a0/vocab.txt\n",
      "loading file tokenizer.json from cache at None\n",
      "loading file added_tokens.json from cache at None\n",
      "loading file special_tokens_map.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--savasy--bert-base-turkish-ner-cased/snapshots/d2853558b8a3b19639dce6da2d8a5b6d8f0102a0/special_tokens_map.json\n",
      "loading file tokenizer_config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--savasy--bert-base-turkish-ner-cased/snapshots/d2853558b8a3b19639dce6da2d8a5b6d8f0102a0/tokenizer_config.json\n",
      "loading configuration file config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--savasy--bert-base-turkish-ner-cased/snapshots/d2853558b8a3b19639dce6da2d8a5b6d8f0102a0/config.json\n",
      "Model config BertConfig {\n",
      "  \"_name_or_path\": \"savasy/bert-base-turkish-ner-cased\",\n",
      "  \"_num_labels\": 7,\n",
      "  \"architectures\": [\n",
      "    \"BertForTokenClassification\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 768,\n",
      "  \"id2label\": {\n",
      "    \"0\": \"B-LOC\",\n",
      "    \"1\": \"B-ORG\",\n",
      "    \"2\": \"B-PER\",\n",
      "    \"3\": \"I-LOC\",\n",
      "    \"4\": \"I-ORG\",\n",
      "    \"5\": \"I-PER\",\n",
      "    \"6\": \"O\"\n",
      "  },\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 3072,\n",
      "  \"label2id\": {\n",
      "    \"B-LOC\": 0,\n",
      "    \"B-ORG\": 1,\n",
      "    \"B-PER\": 2,\n",
      "    \"I-LOC\": 3,\n",
      "    \"I-ORG\": 4,\n",
      "    \"I-PER\": 5,\n",
      "    \"O\": 6\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",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 32000\n",
      "}\n",
      "\n",
      "loading configuration file config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--savasy--bert-base-turkish-ner-cased/snapshots/d2853558b8a3b19639dce6da2d8a5b6d8f0102a0/config.json\n",
      "Model config BertConfig {\n",
      "  \"_name_or_path\": \"savasy/bert-base-turkish-ner-cased\",\n",
      "  \"_num_labels\": 7,\n",
      "  \"architectures\": [\n",
      "    \"BertForTokenClassification\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 768,\n",
      "  \"id2label\": {\n",
      "    \"0\": \"B-LOC\",\n",
      "    \"1\": \"B-ORG\",\n",
      "    \"2\": \"B-PER\",\n",
      "    \"3\": \"I-LOC\",\n",
      "    \"4\": \"I-ORG\",\n",
      "    \"5\": \"I-PER\",\n",
      "    \"6\": \"O\"\n",
      "  },\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 3072,\n",
      "  \"label2id\": {\n",
      "    \"B-LOC\": 0,\n",
      "    \"B-ORG\": 1,\n",
      "    \"B-PER\": 2,\n",
      "    \"I-LOC\": 3,\n",
      "    \"I-ORG\": 4,\n",
      "    \"I-PER\": 5,\n",
      "    \"O\": 6\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",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 32000\n",
      "}\n",
      "\n",
      "loading configuration file config.json from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--savasy--bert-base-turkish-ner-cased/snapshots/d2853558b8a3b19639dce6da2d8a5b6d8f0102a0/config.json\n",
      "Model config BertConfig {\n",
      "  \"_name_or_path\": \"savasy/bert-base-turkish-ner-cased\",\n",
      "  \"_num_labels\": 7,\n",
      "  \"architectures\": [\n",
      "    \"BertForTokenClassification\"\n",
      "  ],\n",
      "  \"attention_probs_dropout_prob\": 0.1,\n",
      "  \"classifier_dropout\": null,\n",
      "  \"hidden_act\": \"gelu\",\n",
      "  \"hidden_dropout_prob\": 0.1,\n",
      "  \"hidden_size\": 768,\n",
      "  \"id2label\": {\n",
      "    \"0\": \"B-LOC\",\n",
      "    \"1\": \"B-ORG\",\n",
      "    \"2\": \"B-PER\",\n",
      "    \"3\": \"I-LOC\",\n",
      "    \"4\": \"I-ORG\",\n",
      "    \"5\": \"I-PER\",\n",
      "    \"6\": \"O\"\n",
      "  },\n",
      "  \"initializer_range\": 0.02,\n",
      "  \"intermediate_size\": 3072,\n",
      "  \"label2id\": {\n",
      "    \"B-LOC\": 0,\n",
      "    \"B-ORG\": 1,\n",
      "    \"B-PER\": 2,\n",
      "    \"I-LOC\": 3,\n",
      "    \"I-ORG\": 4,\n",
      "    \"I-PER\": 5,\n",
      "    \"O\": 6\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",
      "  \"position_embedding_type\": \"absolute\",\n",
      "  \"transformers_version\": \"4.23.1\",\n",
      "  \"type_vocab_size\": 2,\n",
      "  \"use_cache\": true,\n",
      "  \"vocab_size\": 32000\n",
      "}\n",
      "\n",
      "loading weights file pytorch_model.bin from cache at /Users/sinanozdemir/.cache/huggingface/hub/models--savasy--bert-base-turkish-ner-cased/snapshots/d2853558b8a3b19639dce6da2d8a5b6d8f0102a0/pytorch_model.bin\n",
      "All model checkpoint weights were used when initializing BertForTokenClassification.\n",
      "\n",
      "All the weights of BertForTokenClassification were initialized from the model checkpoint at savasy/bert-base-turkish-ner-cased.\n",
      "If your task is similar to the task the model of the checkpoint was trained on, you can already use BertForTokenClassification for predictions without further training.\n"
     ]
    }
   ],
   "source": [
    "# https://huggingface.co/savasy/bert-base-turkish-ner-cased\n",
    "custom_module = 'savasy/bert-base-turkish-ner-cased'\n",
    "\n",
    "turkish_ner_tokenizer = AutoTokenizer.from_pretrained(custom_module)\n",
    "turkish_ner_model = AutoModelForTokenClassification.from_pretrained(custom_module)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "everyday-madison",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "transformers.models.bert.modeling_bert.BertModel"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(turkish_ner_model.bert)  # standard BERT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "hispanic-gross",
   "metadata": {},
   "outputs": [],
   "source": [
    "sequence = \"Merhaba! Benim adım Sinan. San Francisco'dan geliyorum\" # Hi! I'm Sinan. I come from San Francisco"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "sixth-construction",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'entity': 'B-PER',\n",
       "  'score': 0.724247,\n",
       "  'index': 5,\n",
       "  'word': 'Sinan',\n",
       "  'start': 20,\n",
       "  'end': 25},\n",
       " {'entity': 'B-LOC',\n",
       "  'score': 0.99879956,\n",
       "  'index': 7,\n",
       "  'word': 'San',\n",
       "  'start': 27,\n",
       "  'end': 30},\n",
       " {'entity': 'I-LOC',\n",
       "  'score': 0.99770975,\n",
       "  'index': 8,\n",
       "  'word': 'Francisco',\n",
       "  'start': 31,\n",
       "  'end': 40}]"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ner=pipeline('ner', model=turkish_ner_model, tokenizer=turkish_ner_tokenizer)\n",
    "ner(sequence)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "hollow-savannah",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "70db93ea",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}