Multiclass Classification/Transformer Comparison/Hate & Offensive Speech - hateBERT.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Detecting Hate & Offensive Speech\n",
    "\n",
    "Dataset Source: https://www.kaggle.com/datasets/subhajournal/normal-hate-and-offensive-speeches"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Import Necessary Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "NOTE: Redirects are currently not supported in Windows or MacOs.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Git LFS initialized.\n"
     ]
    }
   ],
   "source": [
    "import os, sys\n",
    "os.environ['TOKENIZERS_PARALLELISM']='false'\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import datasets\n",
    "from datasets import Dataset, DatasetDict\n",
    "\n",
    "import torch\n",
    "\n",
    "import transformers\n",
    "from transformers import AutoModelForSequenceClassification, AutoTokenizer\n",
    "from transformers import TrainingArguments, Trainer, set_seed, DataCollatorWithPadding\n",
    "\n",
    "import evaluate\n",
    "\n",
    "!git lfs install"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Display Library Versions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "        Python: 3.9.12\n",
      "         NumPy: 1.24.2\n",
      "        Pandas: 1.5.0\n",
      "      Datasets: 2.9.0\n",
      "  Transformers: 4.26.1\n",
      "         Torch: 1.12.1\n",
      "      Evaluate: 0.4.0\n"
     ]
    }
   ],
   "source": [
    "print(\"Python:\".rjust(15), sys.version[0:6])\n",
    "print(\"NumPy:\".rjust(15), np.__version__)\n",
    "print(\"Pandas:\".rjust(15), pd.__version__)\n",
    "print(\"Datasets:\".rjust(15), datasets.__version__)\n",
    "print(\"Transformers:\".rjust(15), transformers.__version__)\n",
    "print(\"Torch:\".rjust(15), torch.__version__)\n",
    "print(\"Evaluate:\".rjust(15), evaluate.__version__)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Load & Start Preprocessing Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>text</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Republicans need to go on the offensive! They ...</td>\n",
       "      <td>Offensive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>&amp;gt; Make fun of people who lost their familie...</td>\n",
       "      <td>Offensive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>You're missing my point entirely. I don't give...</td>\n",
       "      <td>Offensive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Wet leg quoting Alex turner’s Brit acceptance ...</td>\n",
       "      <td>Offensive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>@SammiePressdee Sir Stephen Sedley said: \"Free...</td>\n",
       "      <td>Offensive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>218</th>\n",
       "      <td>@KumudShankar @nandtara @MoatasimSolkar @Gauta...</td>\n",
       "      <td>Normal</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>219</th>\n",
       "      <td>There’s no point dropping speech when he is no...</td>\n",
       "      <td>Normal</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>220</th>\n",
       "      <td>harry is one of the most humble musicians i kn...</td>\n",
       "      <td>Normal</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>221</th>\n",
       "      <td>Bro want the biggest most political arc in rec...</td>\n",
       "      <td>Normal</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>222</th>\n",
       "      <td>It's comedic how HxH stans only defenses are t...</td>\n",
       "      <td>Normal</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3061 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  text      label\n",
       "0    Republicans need to go on the offensive! They ...  Offensive\n",
       "1    &gt; Make fun of people who lost their familie...  Offensive\n",
       "2    You're missing my point entirely. I don't give...  Offensive\n",
       "3    Wet leg quoting Alex turner’s Brit acceptance ...  Offensive\n",
       "4    @SammiePressdee Sir Stephen Sedley said: \"Free...  Offensive\n",
       "..                                                 ...        ...\n",
       "218  @KumudShankar @nandtara @MoatasimSolkar @Gauta...     Normal\n",
       "219  There’s no point dropping speech when he is no...     Normal\n",
       "220  harry is one of the most humble musicians i kn...     Normal\n",
       "221  Bro want the biggest most political arc in rec...     Normal\n",
       "222  It's comedic how HxH stans only defenses are t...     Normal\n",
       "\n",
       "[3061 rows x 2 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "parent_dir = \"/Users/briandunn/Desktop/nlp_ds/Hate & Offensive Speech\"\n",
    "\n",
    "unique_classes = [\"Normal\", \"Hate\", \"Offensive\"]\n",
    "\n",
    "data = pd.DataFrame()\n",
    "\n",
    "### loop through all three classes ('Normal', 'Hate', 'Offensive')\n",
    "for class_name in unique_classes:\n",
    "    for x in range(4):\n",
    "        temp_df = pd.DataFrame()\n",
    "        file_name = class_name + \"_Speeches_\" + str(x + 1) + \".csv\"\n",
    "        data_file_name = os.path.join(parent_dir, file_name)\n",
    "        temp_df = pd.read_csv(data_file_name, usecols=[3])\n",
    "        temp_df['label'] = class_name\n",
    "        data = pd.concat([temp_df, data]) \n",
    "\n",
    "data.rename(columns={\"full_text\": \"text\"}, inplace=True)\n",
    "\n",
    "data"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Define Dictionaries to Convert Between String & Integer Class Values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "id2label = {idx:label for idx, label in enumerate(unique_classes)}\n",
    "label2id = {label:idx for idx, label in enumerate(unique_classes)}\n",
    "NUM_OF_LABELS = len(unique_classes)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Visual: Histogram of Text Lengths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1800x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The maximum length of Speeches is 88 words.\n"
     ]
    }
   ],
   "source": [
    "data['text_length'] = data['text'].str.split().str.len()\n",
    "\n",
    "n_bin=60\n",
    "data.hist('text_length', grid=False, figsize=(18,8), range=(0,n_bin), bins=n_bin)\n",
    "plt.title('Histogram: Length of Speeches')\n",
    "plt.xlabel('# of Words in Speeches')\n",
    "plt.ylabel('# of Samples')\n",
    "plt.show()\n",
    "\n",
    "print(f\"The maximum length of Speeches is {max(data['text_length'])} words.\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Visual: Tweet Length by Class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "data.boxplot(\"text_length\", by=\"label\", grid=False, showfliers=False, color=\"Blue\", figsize=(10,4))\n",
    "\n",
    "plt.suptitle('')\n",
    "plt.title('Tweet Length by Type')\n",
    "plt.xlabel('')\n",
    "plt.ylabel('Words per Tweet')\n",
    "plt.show()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Final Data Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 3044 entries, 0 to 222\n",
      "Data columns (total 2 columns):\n",
      " #   Column  Non-Null Count  Dtype \n",
      "---  ------  --------------  ----- \n",
      " 0   text    3044 non-null   object\n",
      " 1   label   3044 non-null   object\n",
      "dtypes: object(2)\n",
      "memory usage: 71.3+ KB\n"
     ]
    }
   ],
   "source": [
    "data = data[data[\"text_length\"] < 57]\n",
    "\n",
    "data = data.drop(columns=[\"text_length\"])\n",
    "\n",
    "data.info()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Number of Samples in Each Class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "class_counts = data.groupby(\"label\").count()\n",
    "\n",
    "class_counts.plot.bar()\n",
    "plt.suptitle('')\n",
    "plt.title('# of Samples of Each Class')\n",
    "plt.xlabel('Class')\n",
    "plt.ylabel('# of Samples')\n",
    "plt.xticks(rotation = 0)\n",
    "plt.legend('', frameon=False)\n",
    "plt.show()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Convert Class Values From String to Integer & Convert DataFrame to Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data.replace(label2id)\n",
    "\n",
    "dataset = Dataset.from_pandas(data)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Split Dataset into Training, Testing & Evaluation Datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Dataset Shape: (2435, 3)\n",
      "Testing Dataset Shape: (304, 3)\n",
      "Evaluation Dataset Shape: (305, 3)\n"
     ]
    }
   ],
   "source": [
    "train_testEval = dataset.train_test_split(train_size=0.80)\n",
    "test_eval = train_testEval['test'].train_test_split(train_size=0.50)\n",
    "\n",
    "ds = DatasetDict({\n",
    "    'train' : train_testEval['train'],\n",
    "    'test' : test_eval['train'],\n",
    "    'eval' : test_eval['test'],\n",
    "})\n",
    "\n",
    "print(\"Training Dataset Shape:\", ds['train'].shape)\n",
    "print(\"Testing Dataset Shape:\", ds['test'].shape)\n",
    "print(\"Evaluation Dataset Shape:\", ds['eval'].shape)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Basic Values/Constants"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "This project will utilize the CPU to train and evaluate.\n",
      "This project will have a maximum length for the input text of 72 tokens.\n",
      "The model checkpoint used for this project is: GroNLP/hateBERT\n"
     ]
    }
   ],
   "source": [
    "set_seed = 42\n",
    "NUM_OF_EPOCHS = 3\n",
    "\n",
    "BATCH_SIZE = 64\n",
    "LEARNING_RATE = 2e-5\n",
    "\n",
    "MAX_LENGTH = int(56 * 1.3)\n",
    "MODEL_CKPT = \"GroNLP/hateBERT\"\n",
    "MODEL_NAME = MODEL_CKPT.split(\"/\")[-1] + \"-Hate_Offensive_or_Normal_Speech\"\n",
    "\n",
    "METRIC_NAME = \"f1\"\n",
    "DEVICE = torch.device(\"cpu\")\n",
    "\n",
    "print(f\"This project will utilize the {str(DEVICE).upper()} to train and evaluate.\")\n",
    "print(f\"This project will have a maximum length for the input text of {MAX_LENGTH} tokens.\")\n",
    "print(f\"The model checkpoint used for this project is: {MODEL_CKPT}\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Tokenize Entire Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "00013e33fc10406f82f6cfd410d5cf79",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6e56177b544d43509d08b2f8a188a45e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c418d437673841009a5bd6cd8d032699",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "afcb534d936a4e728312edaa08101de9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "57bf47925fed4f50bb8d57e925403249",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "852cb3c6094d418088f7f3de02ef41d9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a97da8b4ebfb4a4a904f55c9f1717c5f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "training data shape: (2435, 6)\n",
      "testing data shape: (304, 6)\n",
      "validation data shape: (305, 6)\n",
      "{'text': Value(dtype='string', id=None), 'label': Value(dtype='int64', id=None), '__index_level_0__': Value(dtype='int64', id=None), 'input_ids': Sequence(feature=Value(dtype='int32', id=None), length=-1, id=None), 'token_type_ids': Sequence(feature=Value(dtype='int8', id=None), length=-1, id=None), 'attention_mask': Sequence(feature=Value(dtype='int8', id=None), length=-1, id=None)}\n",
      "Dataset({\n",
      "    features: ['text', 'label', '__index_level_0__', 'input_ids', 'token_type_ids', 'attention_mask'],\n",
      "    num_rows: 2435\n",
      "})\n"
     ]
    }
   ],
   "source": [
    "# Define the Tokenizer\n",
    "tokenizer = AutoTokenizer.from_pretrained(MODEL_CKPT)\n",
    "\n",
    "# Method to Tokenize each Sample in the Dataset\n",
    "def tokenize(batch):\n",
    "    return tokenizer(batch['text'], \n",
    "                     padding=True, \n",
    "                     truncation=True, \n",
    "                     max_length=MAX_LENGTH, \n",
    "                     return_tensors=\"pt\")\n",
    "\n",
    "# Map the 'Tokenize' Method to All Samples in the Entire Dataset\n",
    "encoded_data = ds.map(tokenize, batched=True, batch_size=None)\n",
    "\n",
    "print('training data shape:', encoded_data['train'].shape)\n",
    "print('testing data shape:', encoded_data['test'].shape)\n",
    "print('validation data shape:', encoded_data['eval'].shape)\n",
    "\n",
    "print(encoded_data['train'].features)\n",
    "print(encoded_data['train'])"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Instantiate Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fda906d85fa34acaadc02f6dba42b868",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at GroNLP/hateBERT were not used when initializing BertForSequenceClassification: ['cls.predictions.transform.LayerNorm.weight', 'cls.predictions.bias', 'cls.predictions.transform.dense.weight', 'cls.predictions.transform.LayerNorm.bias', 'cls.predictions.transform.dense.bias', 'cls.predictions.decoder.bias', 'cls.predictions.decoder.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 GroNLP/hateBERT and are newly initialized: ['classifier.bias', 'classifier.weight']\n",
      "You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n"
     ]
    }
   ],
   "source": [
    "model = (\n",
    "    AutoModelForSequenceClassification.from_pretrained(\n",
    "        MODEL_CKPT,\n",
    "        num_labels=NUM_OF_LABELS\n",
    "    ).to(DEVICE))"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Define Function to Compute Metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_metrics(p):\n",
    "    accuracy_metric = evaluate.load(\"accuracy\")\n",
    "    accuracy = accuracy_metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids)['accuracy']\n",
    "    \n",
    "    ### ------------------- F1 scores -------------------\n",
    "    \n",
    "    f1_score_metric = evaluate.load(\"f1\")\n",
    "    weighted_f1_score = f1_score_metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids, average='weighted')[\"f1\"]\n",
    "    micro_f1_score = f1_score_metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids, average='micro')['f1']\n",
    "    macro_f1_score = f1_score_metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids, average='macro')[\"f1\"]\n",
    "    \n",
    "    ### ------------------- recall -------------------\n",
    "    \n",
    "    recall_metric = evaluate.load(\"recall\")\n",
    "    weighted_recall = recall_metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids, average='weighted')[\"recall\"]\n",
    "    micro_recall = recall_metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids, average='micro')[\"recall\"]\n",
    "    macro_recall = recall_metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids, average='macro')[\"recall\"]\n",
    "    \n",
    "    ### ------------------- precision -------------------\n",
    "    \n",
    "    precision_metric = evaluate.load(\"precision\")\n",
    "    weighted_precision = precision_metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids, average='weighted')[\"precision\"]\n",
    "    micro_precision = precision_metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids, average='micro')[\"precision\"]\n",
    "    macro_precision = precision_metric.compute(predictions=np.argmax(p.predictions, axis=1), references=p.label_ids, average='macro')[\"precision\"]\n",
    "    \n",
    "    return {\"accuracy\" : accuracy, \n",
    "            \"Weighted F1\" : weighted_f1_score,\n",
    "            \"Micro F1\" : micro_f1_score,\n",
    "            \"Macro F1\" : macro_f1_score,\n",
    "            \"Weighted Recall\" : weighted_recall,\n",
    "            \"Micro Recall\" : micro_recall,\n",
    "            \"Macro Recall\" : macro_recall,\n",
    "            \"Weighted Precision\" : weighted_precision,\n",
    "            \"Micro Precision\" : micro_precision,\n",
    "            \"Macro Precision\" : macro_precision\n",
    "            }"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Define TrainingArguments"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "training_args = TrainingArguments(\n",
    "    output_dir=MODEL_NAME,\n",
    "    num_train_epochs=NUM_OF_EPOCHS,\n",
    "    learning_rate=LEARNING_RATE,\n",
    "    per_device_train_batch_size=BATCH_SIZE, \n",
    "    per_device_eval_batch_size=BATCH_SIZE, \n",
    "    weight_decay=0.01,\n",
    "    log_level=\"error\",\n",
    "    logging_strategy='epoch', \n",
    "    evaluation_strategy=\"epoch\",\n",
    "    save_strategy=\"epoch\",\n",
    "    logging_first_step=True,\n",
    "    load_best_model_at_end=True,\n",
    "    disable_tqdm=False,\n",
    "    hub_private_repo=True,\n",
    "    push_to_hub=True\n",
    ")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Define Data Collator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_collator = DataCollatorWithPadding(tokenizer, padding=True)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Subclass Trainer to Handle Class Imbalance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "class CustomTrainer(Trainer):\n",
    "    def compute_loss(self, model, inputs, return_outputs=False):\n",
    "        labels = inputs.get(\"labels\")\n",
    "        outputs = model(**inputs)\n",
    "        logits = outputs.get(\"logits\")\n",
    "        loss_fct = torch.nn.CrossEntropyLoss(weight=torch.tensor([1.0, 2.0, 3.0]))\n",
    "        loss = loss_fct(logits.view(-1, self.model.config.num_labels), labels.view(-1))\n",
    "        return (loss, outputs) if return_outputs else loss"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Instantiate Trainer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Cloning https://huggingface.co/DunnBC22/hateBERT-Hate_Offensive_or_Normal_Speech into local empty directory.\n"
     ]
    }
   ],
   "source": [
    "trainer = CustomTrainer(model=model, \n",
    "                  args=training_args, \n",
    "                  compute_metrics=compute_metrics,\n",
    "                  train_dataset=encoded_data['train'], \n",
    "                  eval_dataset=encoded_data['eval'], \n",
    "                  data_collator=data_collator,\n",
    "                  tokenizer=tokenizer)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Train Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/briandunn/Documents/nlpnn/nlp_ml/lib/python3.9/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"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "182da3186db54d0e85cdc22ce96c174a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'loss': 1.1956, 'learning_rate': 1.982905982905983e-05, 'epoch': 0.03}\n",
      "{'loss': 0.8958, 'learning_rate': 1.3333333333333333e-05, 'epoch': 1.0}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6883f97d8d8f4a7dacdcafbcafe1e655",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'eval_loss': 0.6817051768302917, 'eval_accuracy': 0.5508196721311476, 'eval_Weighted F1': 0.4791953207133972, 'eval_Micro F1': 0.5508196721311476, 'eval_Macro F1': 0.43948227999718165, 'eval_Weighted Recall': 0.5508196721311476, 'eval_Micro Recall': 0.5508196721311476, 'eval_Macro Recall': 0.48530194759702955, 'eval_Weighted Precision': 0.7546806994131118, 'eval_Micro Precision': 0.5508196721311476, 'eval_Macro Precision': 0.7906259636139378, 'eval_runtime': 96.6035, 'eval_samples_per_second': 3.157, 'eval_steps_per_second': 0.052, 'epoch': 1.0}\n",
      "{'loss': 0.4625, 'learning_rate': 6.666666666666667e-06, 'epoch': 2.0}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "da3c7ed3fd4e4ef7bd09f726e3aead31",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'eval_loss': 0.24483025074005127, 'eval_accuracy': 0.9245901639344263, 'eval_Weighted F1': 0.9229623065026865, 'eval_Micro F1': 0.9245901639344263, 'eval_Macro F1': 0.9169939643489605, 'eval_Weighted Recall': 0.9245901639344263, 'eval_Micro Recall': 0.9245901639344263, 'eval_Macro Recall': 0.9102864348765989, 'eval_Weighted Precision': 0.9263321742010266, 'eval_Micro Precision': 0.9245901639344263, 'eval_Macro Precision': 0.9296209962876629, 'eval_runtime': 96.1875, 'eval_samples_per_second': 3.171, 'eval_steps_per_second': 0.052, 'epoch': 2.0}\n",
      "{'loss': 0.2071, 'learning_rate': 0.0, 'epoch': 3.0}\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9aa767824c0d4f378eb71dad8d6ec861",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'eval_loss': 0.165512353181839, 'eval_accuracy': 0.940983606557377, 'eval_Weighted F1': 0.939490046121732, 'eval_Micro F1': 0.940983606557377, 'eval_Macro F1': 0.9350655283433574, 'eval_Weighted Recall': 0.940983606557377, 'eval_Micro Recall': 0.940983606557377, 'eval_Macro Recall': 0.9272583518485158, 'eval_Weighted Precision': 0.9447019789154691, 'eval_Micro Precision': 0.940983606557377, 'eval_Macro Precision': 0.9509891570960273, 'eval_runtime': 98.0446, 'eval_samples_per_second': 3.111, 'eval_steps_per_second': 0.051, 'epoch': 3.0}\n",
      "{'train_runtime': 7460.0398, 'train_samples_per_second': 0.979, 'train_steps_per_second': 0.016, 'train_loss': 0.5243594890985733, 'epoch': 3.0}\n"
     ]
    }
   ],
   "source": [
    "train_results = trainer.train()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Save Model & Metrics on Training Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Several commits (2) will be pushed upstream.\n",
      "The progress bars may be unreliable.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "25500098356549ddbf4b2fa12b5842d9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b1965e102d2f47709969bb7169efda57",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "remote: Scanning LFS files of refs/heads/main for validity...        \n",
      "remote: LFS file scan complete.        \n",
      "To https://huggingface.co/DunnBC22/hateBERT-Hate_Offensive_or_Normal_Speech\n",
      "   d346487..a6d3d2b  main -> main\n",
      "\n",
      "To https://huggingface.co/DunnBC22/hateBERT-Hate_Offensive_or_Normal_Speech\n",
      "   a6d3d2b..c71e039  main -> main\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "***** train metrics *****\n",
      "  epoch                    =        3.0\n",
      "  train_loss               =     0.5244\n",
      "  train_runtime            = 2:04:20.03\n",
      "  train_samples_per_second =      0.979\n",
      "  train_steps_per_second   =      0.016\n"
     ]
    }
   ],
   "source": [
    "trainer.save_model()\n",
    "trainer.log_metrics(\"train\", train_results.metrics)\n",
    "trainer.save_metrics(\"train\", train_results.metrics)\n",
    "trainer.save_state()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Evaluate Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e67733e61db247ffa90350d8494909cd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "***** eval metrics *****\n",
      "  epoch                   =        3.0\n",
      "  eval_Macro F1           =     0.9351\n",
      "  eval_Macro Precision    =      0.951\n",
      "  eval_Macro Recall       =     0.9273\n",
      "  eval_Micro F1           =      0.941\n",
      "  eval_Micro Precision    =      0.941\n",
      "  eval_Micro Recall       =      0.941\n",
      "  eval_Weighted F1        =     0.9395\n",
      "  eval_Weighted Precision =     0.9447\n",
      "  eval_Weighted Recall    =      0.941\n",
      "  eval_accuracy           =      0.941\n",
      "  eval_loss               =     0.1655\n",
      "  eval_runtime            = 0:01:35.92\n",
      "  eval_samples_per_second =       3.18\n",
      "  eval_steps_per_second   =      0.052\n"
     ]
    }
   ],
   "source": [
    "metrics = trainer.evaluate(encoded_data['eval'])\n",
    "trainer.log_metrics(\"eval\", metrics)\n",
    "trainer.save_metrics(\"eval\", metrics)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Push Model to HuggingFace Hub"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9d068dd488214ff4ac48b7623ca375b6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "A Jupyter Widget"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "remote: Scanning LFS files of refs/heads/main for validity...        \n",
      "remote: LFS file scan complete.        \n",
      "To https://huggingface.co/DunnBC22/hateBERT-Hate_Offensive_or_Normal_Speech\n",
      "   c71e039..09d8e0b  main -> main\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'https://huggingface.co/DunnBC22/hateBERT-Hate_Offensive_or_Normal_Speech/commit/09d8e0b8eafbab9f8f5f9fc74a14bb4a407d7d2e'"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trainer.push_to_hub(commit_message=\"All Dunn!!!\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Notes & Other Takeaways From This Project\n",
    "****\n",
    "- This is one of four projects using this same dataset to see which model best predicts if speech is hateful, offensive, or normal.\n",
    "****"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Citations\n",
    "\n",
    "- Model Checkpoint\n",
    "    - @inproceedings{caselli-etal-2021-hatebert,\n",
    "    \\ttitle = \"{H}ate{BERT}: Retraining {BERT} for Abusive Language Detection in {E}nglish\",\n",
    "    \\tauthor = \"Caselli, Tommaso  and\n",
    "      Basile, Valerio  and\n",
    "      Mitrovi{\\'c}, Jelena  and\n",
    "      Granitzer, Michael\",\n",
    "    \\tbooktitle = \"Proceedings of the 5th Workshop on Online Abuse and Harms (WOAH 2021)\",\n",
    "    \\tmonth = aug,\n",
    "    \\tyear = \"2021\",\n",
    "    \\taddress = \"Online\",\n",
    "    \\tpublisher = \"Association for Computational Linguistics\",\n",
    "    \\tturl = \"https://aclanthology.org/2021.woah-1.3\",\n",
    "    \\tdoi = \"10.18653/v1/2021.woah-1.3\",\n",
    "    \\tpages = \"17--25\",\n",
    "    \\tabstract = \"We introduce HateBERT, a re-trained BERT model for abusive language detection in English. The model was trained on RAL-E, a large-scale dataset of Reddit comments in English from communities banned for being offensive, abusive, or hateful that we have curated and made available to the public. We present the results of a detailed comparison between a general pre-trained language model and the retrained version on three English datasets for offensive, abusive language and hate speech detection tasks. In all datasets, HateBERT outperforms the corresponding general BERT model. We also discuss a battery of experiments comparing the portability of the fine-tuned models across the datasets, suggesting that portability is affected by compatibility of the annotated phenomena.\",\n",
    "}"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "nlp_ml",
   "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": "3.9.12"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "0b7a754c5aaaa44a21e2f2148f50b931a5abdffe7ef71537f5954179554f39df"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}