PyTorch_NLP_Sentiment_Analysis_sp_rev.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "7248f02a",
   "metadata": {},
   "outputs": [],
   "source": [
    "#basic imports\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from torch.utils.data import DataLoader,Dataset\n",
    "from torch.nn.utils.rnn import pad_sequence\n",
    "from collections import defaultdict\n",
    "import spacy\n",
    "nlp = spacy.load('en_core_web_sm')\n",
    "device = torch.device(\"mps\")\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "fd2c3cdd",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('/Users/aadityajoshi/Downloads/DATASET.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "de4b8105",
   "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>Review</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Great music service, the audio is high quality...</td>\n",
       "      <td>POSITIVE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Please ignore previous negative rating. This a...</td>\n",
       "      <td>POSITIVE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>This pop-up \"Get the best Spotify experience o...</td>\n",
       "      <td>NEGATIVE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Really buggy and terrible to use as of recently</td>\n",
       "      <td>NEGATIVE</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Dear Spotify why do I get songs that I didn't ...</td>\n",
       "      <td>NEGATIVE</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                              Review     label\n",
       "0  Great music service, the audio is high quality...  POSITIVE\n",
       "1  Please ignore previous negative rating. This a...  POSITIVE\n",
       "2  This pop-up \"Get the best Spotify experience o...  NEGATIVE\n",
       "3    Really buggy and terrible to use as of recently  NEGATIVE\n",
       "4  Dear Spotify why do I get songs that I didn't ...  NEGATIVE"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "8e31be64",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 52702 entries, 0 to 52701\n",
      "Data columns (total 2 columns):\n",
      " #   Column  Non-Null Count  Dtype \n",
      "---  ------  --------------  ----- \n",
      " 0   Review  52686 non-null  object\n",
      " 1   label   52702 non-null  object\n",
      "dtypes: object(2)\n",
      "memory usage: 823.6+ KB\n"
     ]
    }
   ],
   "source": [
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "c2e8fc6f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[]\n"
     ]
    }
   ],
   "source": [
    "df.dropna(inplace=True)\n",
    "df.drop_duplicates(inplace=True)\n",
    "blanks = []\n",
    "for i,text,lb in df.itertuples():\n",
    "    if type(text) == str:\n",
    "        if text.isspace() == True:\n",
    "            blanks.append(i)\n",
    "print(blanks)\n",
    "df.drop(blanks,inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "1fd82c90",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['text'] = df['Review']\n",
    "df.drop('Review',axis=1,inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "66d02b87",
   "metadata": {},
   "outputs": [],
   "source": [
    "def clean_lemm(text):\n",
    "    doc = nlp(text.lower())\n",
    "    clean_text = [token.lemma_ for token in doc if not token.is_stop and not token.is_punct]\n",
    "    return ' '.join(clean_text)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "fc380eed",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['text'] = df['text'].apply(clean_lemm)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "e8979e96",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['len'] = df['text'].apply(len)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "a045eeb4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    great music service audio high quality app eas...\n",
       "1    ignore previous negative rating app super grea...\n",
       "2    pop good spotify experience android 12 annoyin...\n",
       "3                          buggy terrible use recently\n",
       "4              dear spotify song playlist shuffle play\n",
       "Name: text, dtype: object"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['text'].head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "a6a89ea9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<seaborn.axisgrid.FacetGrid at 0x105bd5b70>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 500x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.displot(df['len'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "1421f92b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: xlabel='label', ylabel='count'>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 400x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(4,4))\n",
    "sns.countplot(x='label',data=df,palette='viridis')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "43c31555",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['label'] = df['label'].replace({\"POSITIVE\":1,\"NEGATIVE\":0})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "e8c733f2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "324.52900000000227"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['len'].quantile(0.999)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "231697c7",
   "metadata": {},
   "outputs": [],
   "source": [
    "ind_len = list(df[df['len']>=350].index)\n",
    "for i in ind_len:\n",
    "    df['text'].loc[i] = df['text'].loc[i][:350]\n",
    "df['len'] = df['text'].apply(len)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "4bed318b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<seaborn.axisgrid.FacetGrid at 0x3195e9f60>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 500x500 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.displot(df['len'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "93389780",
   "metadata": {},
   "outputs": [],
   "source": [
    "def spacy_tokenizer(text):\n",
    "    clean_text = clean_lemm(text)\n",
    "    return clean_text.split()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "9aafb732",
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_vocab(text_iterator,specials=('<unk>','<pad>'),min_freq=1):\n",
    "    token_count = defaultdict(int)\n",
    "    for text in text_iterator:\n",
    "        for token in spacy_tokenizer(text):\n",
    "            token_count[token] += 1\n",
    "    vocab = {token : idx for idx,(token,count) in enumerate(token_count.items()) if count>=min_freq}\n",
    "    for special in specials:\n",
    "        if special not in token_count:\n",
    "            vocab[special] = len(vocab)\n",
    "    return vocab,len(vocab)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "0828c206",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "22462\n"
     ]
    }
   ],
   "source": [
    "vocab,vocab_size = build_vocab(df['text'])\n",
    "print(vocab_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "89339fb7",
   "metadata": {},
   "outputs": [],
   "source": [
    "text_pipeline = lambda x : [vocab.get(token,vocab['<unk>']) for token in spacy_tokenizer(x)]\n",
    "label_pipeline = lambda x : int(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "0c82b142",
   "metadata": {},
   "outputs": [],
   "source": [
    "class TextDataset(Dataset):\n",
    "    def __init__(self,df,text_pipeline,label_pipeline):\n",
    "        self.texts = df['text'].tolist()\n",
    "        self.labels = df['label'].tolist()\n",
    "        self.text_pipeline = text_pipeline\n",
    "        self.label_pipeline = label_pipeline\n",
    "    def __len__(self):\n",
    "        return len(self.texts)\n",
    "    def __getitem__(self,idx):\n",
    "        text = self.texts[idx]\n",
    "        label = self.labels[idx]\n",
    "        return self.text_pipeline(text),self.label_pipeline(label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "9e313d7c",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "train_df,test_df = train_test_split(df,test_size=0.33,stratify=df['label'],random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "b227e992",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_dataset = TextDataset(train_df,text_pipeline,label_pipeline)\n",
    "test_dataset = TextDataset(test_df,text_pipeline,label_pipeline)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "e3aeecdb",
   "metadata": {},
   "outputs": [],
   "source": [
    "def pad_collate(batch):\n",
    "    text_list = []\n",
    "    label_list = []\n",
    "    for text,label in batch:\n",
    "        text_list.append(torch.LongTensor(text))\n",
    "        label_list.append(label)\n",
    "    text_list = pad_sequence(text_list,batch_first=True)\n",
    "    label_list = torch.LongTensor(label_list)\n",
    "    return text_list.to(device),label_list.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "c5af448a",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_batch_size = 512\n",
    "test_batch_size = 512\n",
    "train_loader = DataLoader(train_dataset,train_batch_size,shuffle=True,collate_fn=pad_collate)\n",
    "test_loader = DataLoader(test_dataset,test_batch_size,shuffle=False,collate_fn=pad_collate)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "39b06270",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LSTM(nn.Module):\n",
    "    def __init__(self,vocab_size,embedding_size,hidden_size,out_size=2,num_layers=1,bidirectional=True,p=0.4):\n",
    "        super(LSTM,self).__init__()\n",
    "        self.hidden_size = hidden_size\n",
    "        self.num_layers = num_layers\n",
    "        self.num_dir = 2 if bidirectional else 1\n",
    "        self.dropout = nn.Dropout(p)\n",
    "        self.emebdding = nn.Embedding(vocab_size,embedding_size)\n",
    "        self.lstm = nn.LSTM(embedding_size,hidden_size,num_layers=num_layers,bidirectional=bidirectional,batch_first=True)\n",
    "        self.fc = nn.Linear(self.num_dir*self.hidden_size,out_size)\n",
    "    def forward(self,x):\n",
    "        hidden = self.init_hidden(x.size(0))\n",
    "        embeddings = self.emebdding(x)\n",
    "        lstm_out,hidden = self.lstm(embeddings,hidden)\n",
    "        lstm_out = self.dropout(lstm_out)\n",
    "        pred = self.fc(lstm_out[:,-1,:])\n",
    "        return pred\n",
    "    def init_hidden(self,batch_size):\n",
    "        hidden = (torch.zeros(self.num_dir*self.num_layers,batch_size,self.hidden_size).to(device),\n",
    "                 torch.zeros(self.num_dir*self.num_layers,batch_size,self.hidden_size).to(device))\n",
    "        return hidden"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "id": "1b7e141e",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = LSTM(vocab_size,128,32,2,3,bidirectional=False).to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "id": "231093cb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LSTM(\n",
       "  (dropout): Dropout(p=0.4, inplace=False)\n",
       "  (emebdding): Embedding(22462, 128)\n",
       "  (lstm): LSTM(128, 32, num_layers=3, batch_first=True)\n",
       "  (fc): Linear(in_features=32, out_features=2, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "b5aeb116",
   "metadata": {},
   "outputs": [],
   "source": [
    "num_epochs = 10\n",
    "optimizer = torch.optim.Adam(model.parameters(),lr=0.001)\n",
    "criterion = nn.CrossEntropyLoss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "id": "ae1db153",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1/10: 100%|██████████████████| 68/68 [00:58<00:00,  1.16it/s, loss=0.6889]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10, Average Loss: 0.6844\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 2/10: 100%|██████████████████| 68/68 [00:59<00:00,  1.15it/s, loss=0.5741]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 2/10, Average Loss: 0.6461\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 3/10: 100%|██████████████████| 68/68 [00:59<00:00,  1.13it/s, loss=0.4708]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3/10, Average Loss: 0.4270\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 4/10: 100%|██████████████████| 68/68 [00:59<00:00,  1.14it/s, loss=0.3844]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 4/10, Average Loss: 0.3498\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 5/10: 100%|██████████████████| 68/68 [00:59<00:00,  1.15it/s, loss=0.3864]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 5/10, Average Loss: 0.3200\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 6/10: 100%|██████████████████| 68/68 [00:59<00:00,  1.15it/s, loss=0.3277]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 6/10, Average Loss: 0.2914\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 7/10: 100%|██████████████████| 68/68 [00:59<00:00,  1.15it/s, loss=0.2572]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 7/10, Average Loss: 0.2710\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 8/10: 100%|██████████████████| 68/68 [00:59<00:00,  1.15it/s, loss=0.2785]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 8/10, Average Loss: 0.2493\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 9/10: 100%|██████████████████| 68/68 [00:59<00:00,  1.15it/s, loss=0.2157]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 9/10, Average Loss: 0.2320\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 10/10: 100%|█████████████████| 68/68 [00:58<00:00,  1.16it/s, loss=0.1550]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 10/10, Average Loss: 0.2147\n",
      "Total training time: 9.88 minutes\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "from tqdm import tqdm\n",
    "\n",
    "start = time.time()\n",
    "for epoch in range(num_epochs):\n",
    "    model.train()\n",
    "    # Wrap your train_loader with tqdm\n",
    "    pbar = tqdm(train_loader, desc=f\"Epoch {epoch+1}/{num_epochs}\")\n",
    "    total_loss = 0\n",
    "    for text, label in pbar:\n",
    "        optimizer.zero_grad()\n",
    "        y_pred = model(text)\n",
    "        loss = criterion(y_pred, label)\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        \n",
    "        # Update total loss and progress bar\n",
    "        total_loss += loss.item()\n",
    "        pbar.set_postfix({'loss': f'{loss.item():.4f}'})\n",
    "    \n",
    "    # Calculate and print average loss for the epoch\n",
    "    avg_loss = total_loss / len(train_loader)\n",
    "    print(f\"Epoch {epoch+1}/{num_epochs}, Average Loss: {avg_loss:.4f}\")\n",
    "\n",
    "print(f\"Total training time: {(time.time() - start) / 60:.2f} minutes\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "id": "58dc0558",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.88      0.86      0.87      9684\n",
      "           1       0.83      0.85      0.84      7302\n",
      "\n",
      "    accuracy                           0.86     16986\n",
      "   macro avg       0.86      0.86      0.86     16986\n",
      "weighted avg       0.86      0.86      0.86     16986\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "all_preds = []\n",
    "all_labels = []\n",
    "\n",
    "model.eval()\n",
    "with torch.no_grad():\n",
    "    for text, label in test_loader:\n",
    "        pred = model(text)\n",
    "        pred = torch.argmax(pred, dim=1)\n",
    "        \n",
    "        # Collect predictions and labels\n",
    "        all_preds.extend(pred.cpu().numpy())\n",
    "        all_labels.extend(label.cpu().numpy())\n",
    "\n",
    "# Now, use classification_report with the collected predictions and labels\n",
    "print(classification_report(all_labels, all_preds))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b2385a83",
   "metadata": {},
   "source": [
    "### BiLSTM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "867d69e6",
   "metadata": {},
   "outputs": [],
   "source": [
    "class LSTM(nn.Module):\n",
    "    def __init__(self,vocab_size,embedding_size,hidden_size,out_size=2,num_layers=1,bidirectional=True,p=0.4):\n",
    "        super(LSTM,self).__init__()\n",
    "        self.hidden_size = hidden_size\n",
    "        self.num_layers = num_layers\n",
    "        self.num_dir = 2 if bidirectional else 1\n",
    "        self.dropout = nn.Dropout(p)\n",
    "        self.emebdding = nn.Embedding(vocab_size,embedding_size)\n",
    "        self.lstm = nn.LSTM(embedding_size,hidden_size,num_layers=num_layers,bidirectional=bidirectional,batch_first=True)\n",
    "        self.fc = nn.Linear(self.num_dir*self.hidden_size,out_size)\n",
    "    def forward(self,x):\n",
    "        hidden = self.init_hidden(x.size(0))\n",
    "        embeddings = self.emebdding(x)\n",
    "        lstm_out,hidden = self.lstm(embeddings,hidden)\n",
    "        lstm_out = self.dropout(lstm_out)\n",
    "        pred = self.fc(lstm_out[:,-1,:])\n",
    "        return pred\n",
    "    def init_hidden(self,batch_size):\n",
    "        hidden = (torch.zeros(self.num_dir*self.num_layers,batch_size,self.hidden_size).to(device),\n",
    "                 torch.zeros(self.num_dir*self.num_layers,batch_size,self.hidden_size).to(device))\n",
    "        return hidden"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "28f3f1fb",
   "metadata": {},
   "outputs": [],
   "source": [
    "Bilstm = LSTM(vocab_size,64,32,out_size=2,num_layers=1,bidirectional=True).to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "f5c68f56",
   "metadata": {},
   "outputs": [],
   "source": [
    "num_epoch = 5\n",
    "opt_2 = torch.optim.Adam(Bilstm.parameters(),lr=0.005)\n",
    "criterion = nn.CrossEntropyLoss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "17fff583",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1/10: 100%|██████████████████| 68/68 [00:58<00:00,  1.16it/s, loss=0.1625]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10, Average Loss: 0.1411\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 2/10: 100%|██████████████████| 68/68 [00:58<00:00,  1.17it/s, loss=0.1074]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 2/10, Average Loss: 0.1226\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 3/10: 100%|██████████████████| 68/68 [00:58<00:00,  1.16it/s, loss=0.0804]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3/10, Average Loss: 0.1103\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 4/10: 100%|██████████████████| 68/68 [00:58<00:00,  1.16it/s, loss=0.0728]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 4/10, Average Loss: 0.1004\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 5/10: 100%|██████████████████| 68/68 [00:57<00:00,  1.18it/s, loss=0.0974]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 5/10, Average Loss: 0.0961\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 6/10: 100%|██████████████████| 68/68 [00:57<00:00,  1.18it/s, loss=0.0670]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 6/10, Average Loss: 0.0897\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 7/10: 100%|██████████████████| 68/68 [00:57<00:00,  1.19it/s, loss=0.0805]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 7/10, Average Loss: 0.0796\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 8/10: 100%|██████████████████| 68/68 [00:57<00:00,  1.19it/s, loss=0.0940]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 8/10, Average Loss: 0.0758\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 9/10: 100%|██████████████████| 68/68 [00:57<00:00,  1.19it/s, loss=0.0146]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 9/10, Average Loss: 0.0650\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 10/10: 100%|█████████████████| 68/68 [00:57<00:00,  1.17it/s, loss=0.0534]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 10/10, Average Loss: 0.0675\n",
      "Total training time: 9.67 minutes\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "from tqdm import tqdm\n",
    "\n",
    "start = time.time()\n",
    "for epoch in range(num_epochs):\n",
    "    Bilstm.train()\n",
    "    # Wrap your train_loader with tqdm\n",
    "    pbar = tqdm(train_loader, desc=f\"Epoch {epoch+1}/{num_epochs}\")\n",
    "    total_loss = 0\n",
    "    for text, label in pbar:\n",
    "        opt_2.zero_grad()\n",
    "        y_pred = Bilstm(text)\n",
    "        loss = criterion(y_pred, label)\n",
    "        loss.backward()\n",
    "        opt_2.step()\n",
    "        \n",
    "        # Update total loss and progress bar\n",
    "        total_loss += loss.item()\n",
    "        pbar.set_postfix({'loss': f'{loss.item():.4f}'})\n",
    "    \n",
    "    # Calculate and print average loss for the epoch\n",
    "    avg_loss = total_loss / len(train_loader)\n",
    "    print(f\"Epoch {epoch+1}/{num_epochs}, Average Loss: {avg_loss:.4f}\")\n",
    "\n",
    "print(f\"Total training time: {(time.time() - start) / 60:.2f} minutes\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "8aec7faa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.87      0.85      0.86      9684\n",
      "           1       0.81      0.83      0.82      7302\n",
      "\n",
      "    accuracy                           0.84     16986\n",
      "   macro avg       0.84      0.84      0.84     16986\n",
      "weighted avg       0.84      0.84      0.84     16986\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "all_preds = []\n",
    "all_labels = []\n",
    "\n",
    "Bilstm.eval()\n",
    "with torch.no_grad():\n",
    "    for text, label in test_loader:\n",
    "        pred = Bilstm(text)\n",
    "        pred = torch.argmax(pred, dim=1)\n",
    "        \n",
    "        # Collect predictions and labels\n",
    "        all_preds.extend(pred.cpu().numpy())\n",
    "        all_labels.extend(label.cpu().numpy())\n",
    "\n",
    "# Now, use classification_report with the collected predictions and labels\n",
    "print(classification_report(all_labels, all_preds))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c2ab6b3b",
   "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": "3.10.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}