training_bert.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "initial_id",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\tqdm\\auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    }
   ],
   "source": [
    "import json\n",
    "import torch\n",
    "import numpy as np\n",
    "from transformers import AdamW\n",
    "from sklearn.metrics import accuracy_score\n",
    "from transformers import BertTokenizer, BertForSequenceClassification\n",
    "from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset\n",
    "from tqdm import tqdm\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import KFold\n",
    "from torch.utils.data import Subset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "336e3f59095cf075",
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "file_path = 'qa_dataset.json'\n",
    "\n",
    "with open(file_path, 'r') as file:\n",
    "    data = json.load(file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "29c40423370cfb0c",
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "30901295",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Original \n",
    "# Prepare data \n",
    "\n",
    "# Tokenizing the data\n",
    "# inputs = tokenizer([x['question'] + \" [SEP] \" + x['answer'] for x in data], padding=True, truncation=True, return_tensors=\"pt\")\n",
    "\n",
    "# Assuming binary classification (change as needed)\n",
    "# labels = torch.tensor([1 if x['answer_length'] > 100 else 0 for x in data])\n",
    "\n",
    "# Create a dataset\n",
    "# dataset = TensorDataset(inputs['input_ids'], inputs['attention_mask'], labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "a37eabf0",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Prepare data\n",
    "\n",
    "# Tokenizing the data\n",
    "max_length = max(len(tokenizer.encode(x['question'] + \" [SEP] \" + x['answer'])) for x in data)\n",
    "tokenized_data = [tokenizer(x['question'] + \" [SEP] \" + x['answer'], \n",
    "                            padding='max_length',  \n",
    "                            max_length=max_length,  \n",
    "                            truncation=True, \n",
    "                            return_tensors=\"pt\") for x in data]\n",
    "\n",
    "# input\n",
    "input_ids = torch.cat([item['input_ids'] for item in tokenized_data], dim=0)\n",
    "attention_masks = torch.cat([item['attention_mask'] for item in tokenized_data], dim=0)\n",
    "labels = torch.tensor([1 if x['answer_length'] > 100 else 0 for x in data])\n",
    "\n",
    "# dataset seprate\n",
    "input_ids_train, input_ids_test, attention_masks_train, attention_masks_test, labels_train, labels_test = train_test_split(\n",
    "    input_ids, attention_masks, labels, train_size=0.7, random_state=42)\n",
    "\n",
    "# create dataset\n",
    "train_dataset = TensorDataset(input_ids_train, attention_masks_train, labels_train)\n",
    "test_dataset = TensorDataset(input_ids_test, attention_masks_test, labels_test)\n",
    "dataset = TensorDataset(input_ids, attention_masks, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "9b2a94ce5138fac5",
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "65fc5a99",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CUDA is available. Using GPU.\n"
     ]
    }
   ],
   "source": [
    "# check GPU is work\n",
    "# Train the model using a suitable optimizer and loss function.\n",
    "if torch.cuda.is_available():\n",
    "    print(\"CUDA is available. Using GPU.\")\n",
    "    device = torch.device(\"cuda\")\n",
    "else:\n",
    "    print(\"CUDA is not available. Using CPU.\")\n",
    "    device = torch.device(\"cpu\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "abc82c94",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1628\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(tensor([[ 101, 2515, 2023,  ...,    0,    0,    0],\n",
       "         [ 101, 2064, 2017,  ...,    0,    0,    0],\n",
       "         [ 101, 2054, 2003,  ...,    0,    0,    0],\n",
       "         ...,\n",
       "         [ 101, 2064, 2009,  ...,    0,    0,    0],\n",
       "         [ 101, 7632, 2515,  ...,    0,    0,    0],\n",
       "         [ 101, 2003, 1996,  ...,    0,    0,    0]]),\n",
       " tensor([[1, 1, 1,  ..., 0, 0, 0],\n",
       "         [1, 1, 1,  ..., 0, 0, 0],\n",
       "         [1, 1, 1,  ..., 0, 0, 0],\n",
       "         ...,\n",
       "         [1, 1, 1,  ..., 0, 0, 0],\n",
       "         [1, 1, 1,  ..., 0, 0, 0],\n",
       "         [1, 1, 1,  ..., 0, 0, 0]]),\n",
       " tensor([1, 0, 1, 0, 1, 0, 0, 0, 0, 0]))"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(len(dataset))\n",
    "dataset[:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a3b226d5e9330d47",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "## Config 1-8"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "5a16023ef51bad72",
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameters: \n",
      "    Batch size: 4\n",
      "    Epochs: 25\n",
      "    Learning rate:1e-05\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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",
      "Epoch 1: 100%|██████████| 285/285 [00:45<00:00,  6.28it/s, loss=0.373]\n",
      "Epoch 2: 100%|██████████| 285/285 [00:44<00:00,  6.38it/s, loss=0.16]  \n",
      "Epoch 3: 100%|██████████| 285/285 [00:44<00:00,  6.39it/s, loss=0.0619]\n",
      "Epoch 4: 100%|██████████| 285/285 [00:45<00:00,  6.28it/s, loss=0.0326] \n",
      "Epoch 5: 100%|██████████| 285/285 [00:45<00:00,  6.32it/s, loss=0.0193] \n",
      "Epoch 6: 100%|██████████| 285/285 [00:44<00:00,  6.33it/s, loss=0.00126] \n",
      "Epoch 7: 100%|██████████| 285/285 [00:44<00:00,  6.34it/s, loss=0.000776]\n",
      "Epoch 8: 100%|██████████| 285/285 [00:44<00:00,  6.38it/s, loss=0.000549]\n",
      "Epoch 9: 100%|██████████| 285/285 [00:44<00:00,  6.38it/s, loss=0.000417]\n",
      "Epoch 10: 100%|██████████| 285/285 [00:44<00:00,  6.40it/s, loss=0.000338]\n",
      "Epoch 11: 100%|██████████| 285/285 [00:44<00:00,  6.39it/s, loss=0.000253]\n",
      "Epoch 12: 100%|██████████| 285/285 [00:45<00:00,  6.26it/s, loss=0.000209]\n",
      "Epoch 13: 100%|██████████| 285/285 [00:44<00:00,  6.41it/s, loss=0.00017] \n",
      "Epoch 14: 100%|██████████| 285/285 [00:45<00:00,  6.25it/s, loss=0.000139]\n",
      "Epoch 15: 100%|██████████| 285/285 [00:45<00:00,  6.33it/s, loss=0.000115]\n",
      "Epoch 16: 100%|██████████| 285/285 [00:44<00:00,  6.38it/s, loss=9.9e-5] \n",
      "Epoch 17: 100%|██████████| 285/285 [00:44<00:00,  6.43it/s, loss=8.34e-5]\n",
      "Epoch 18: 100%|██████████| 285/285 [00:44<00:00,  6.40it/s, loss=0.076] \n",
      "Epoch 19: 100%|██████████| 285/285 [00:44<00:00,  6.42it/s, loss=0.0637] \n",
      "Epoch 20: 100%|██████████| 285/285 [00:44<00:00,  6.39it/s, loss=0.0117]  \n",
      "Epoch 21: 100%|██████████| 285/285 [00:44<00:00,  6.44it/s, loss=0.0163] \n",
      "Epoch 22: 100%|██████████| 285/285 [00:44<00:00,  6.46it/s, loss=0.00106] \n",
      "Epoch 23: 100%|██████████| 285/285 [00:43<00:00,  6.48it/s, loss=0.000257]\n",
      "Epoch 24: 100%|██████████| 285/285 [00:43<00:00,  6.48it/s, loss=0.000176]\n",
      "Epoch 25: 100%|██████████| 285/285 [00:44<00:00,  6.45it/s, loss=0.000137]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss on training: [0.3728603225397436, 0.1595429493424793, 0.061851121319392535, 0.03263494303539853, 0.019264916288911513, 0.0012607566093241697, 0.0007756983959361127, 0.0005494942424998602, 0.0004166449720893676, 0.0003377624296199269, 0.0002534672869124303, 0.00020883632200900792, 0.00016970583842760932, 0.00013902669414069113, 0.0001151190109994148, 9.90145510928542e-05, 8.343303620278133e-05, 0.07604774169001183, 0.06365864756414993, 0.01173559029034151, 0.016309166439092034, 0.0010589074847909311, 0.00025663965168764424, 0.000176008670946191, 0.0001365924863197857]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Evaluating: 100%|██████████| 123/123 [00:05<00:00, 22.60it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test set: 0.9570552147239264\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameters: \n",
      "    Batch size: 4\n",
      "    Epochs: 25\n",
      "    Learning rate:0.0001\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1: 100%|██████████| 285/285 [00:44<00:00,  6.46it/s, loss=0.357]\n",
      "Epoch 2: 100%|██████████| 285/285 [00:43<00:00,  6.48it/s, loss=0.138] \n",
      "Epoch 3: 100%|██████████| 285/285 [00:43<00:00,  6.49it/s, loss=0.0538]\n",
      "Epoch 4: 100%|██████████| 285/285 [00:44<00:00,  6.44it/s, loss=0.0367] \n",
      "Epoch 5: 100%|██████████| 285/285 [00:44<00:00,  6.43it/s, loss=0.0271]\n",
      "Epoch 6: 100%|██████████| 285/285 [00:44<00:00,  6.44it/s, loss=0.00145] \n",
      "Epoch 7: 100%|██████████| 285/285 [00:44<00:00,  6.46it/s, loss=0.000831]\n",
      "Epoch 8: 100%|██████████| 285/285 [00:44<00:00,  6.45it/s, loss=0.000481]\n",
      "Epoch 9: 100%|██████████| 285/285 [00:44<00:00,  6.44it/s, loss=0.000353]\n",
      "Epoch 10: 100%|██████████| 285/285 [00:44<00:00,  6.45it/s, loss=0.000266]\n",
      "Epoch 11: 100%|██████████| 285/285 [00:44<00:00,  6.47it/s, loss=0.000213]\n",
      "Epoch 12: 100%|██████████| 285/285 [00:44<00:00,  6.41it/s, loss=0.000171]\n",
      "Epoch 13: 100%|██████████| 285/285 [00:44<00:00,  6.44it/s, loss=0.000139]\n",
      "Epoch 14: 100%|██████████| 285/285 [00:44<00:00,  6.41it/s, loss=0.000119]\n",
      "Epoch 15: 100%|██████████| 285/285 [00:45<00:00,  6.31it/s, loss=9.74e-5]\n",
      "Epoch 16: 100%|██████████| 285/285 [00:44<00:00,  6.35it/s, loss=8.45e-5]\n",
      "Epoch 17: 100%|██████████| 285/285 [00:43<00:00,  6.56it/s, loss=6.93e-5]\n",
      "Epoch 18: 100%|██████████| 285/285 [00:43<00:00,  6.55it/s, loss=5.81e-5]\n",
      "Epoch 19: 100%|██████████| 285/285 [00:43<00:00,  6.59it/s, loss=4.93e-5]\n",
      "Epoch 20: 100%|██████████| 285/285 [00:42<00:00,  6.63it/s, loss=4.34e-5]\n",
      "Epoch 21: 100%|██████████| 285/285 [00:43<00:00,  6.62it/s, loss=3.77e-5]\n",
      "Epoch 22: 100%|██████████| 285/285 [00:43<00:00,  6.61it/s, loss=3.21e-5]\n",
      "Epoch 23: 100%|██████████| 285/285 [00:43<00:00,  6.59it/s, loss=2.77e-5]\n",
      "Epoch 24: 100%|██████████| 285/285 [00:43<00:00,  6.60it/s, loss=2.45e-5]\n",
      "Epoch 25: 100%|██████████| 285/285 [00:44<00:00,  6.43it/s, loss=2.11e-5]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss on training: [0.35690931618997923, 0.1381112632619446, 0.05376501740364913, 0.036748769423483235, 0.027103278931361018, 0.0014534329182648083, 0.0008310504609988512, 0.0004807028880653282, 0.00035338942140234603, 0.00026572233550005447, 0.00021320766488012546, 0.00017075575168220032, 0.00013928059779573233, 0.00011894981315044947, 9.741357050523603e-05, 8.453160225415782e-05, 6.931088001900271e-05, 5.810359342783494e-05, 4.9327605972250335e-05, 4.341388246980945e-05, 3.76944810436965e-05, 3.2086907639607316e-05, 2.767802957111455e-05, 2.452595636944017e-05, 2.1106108001651846e-05]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Evaluating: 100%|██████████| 123/123 [00:05<00:00, 22.80it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test set: 0.9468302658486708\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameters: \n",
      "    Batch size: 4\n",
      "    Epochs: 30\n",
      "    Learning rate:1e-05\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1: 100%|██████████| 285/285 [00:44<00:00,  6.41it/s, loss=0.398]\n",
      "Epoch 2: 100%|██████████| 285/285 [00:44<00:00,  6.44it/s, loss=0.238] \n",
      "Epoch 3: 100%|██████████| 285/285 [00:44<00:00,  6.45it/s, loss=0.109] \n",
      "Epoch 4: 100%|██████████| 285/285 [00:44<00:00,  6.38it/s, loss=0.0484] \n",
      "Epoch 5: 100%|██████████| 285/285 [00:44<00:00,  6.37it/s, loss=0.0281] \n",
      "Epoch 6: 100%|██████████| 285/285 [00:44<00:00,  6.36it/s, loss=0.0377] \n",
      "Epoch 7: 100%|██████████| 285/285 [00:44<00:00,  6.37it/s, loss=0.0398] \n",
      "Epoch 8: 100%|██████████| 285/285 [00:44<00:00,  6.45it/s, loss=0.00403]\n",
      "Epoch 9: 100%|██████████| 285/285 [00:43<00:00,  6.53it/s, loss=0.000948]\n",
      "Epoch 10: 100%|██████████| 285/285 [00:44<00:00,  6.42it/s, loss=0.00459] \n",
      "Epoch 11: 100%|██████████| 285/285 [00:44<00:00,  6.44it/s, loss=0.000572]\n",
      "Epoch 12: 100%|██████████| 285/285 [00:43<00:00,  6.49it/s, loss=0.000364]\n",
      "Epoch 13: 100%|██████████| 285/285 [00:43<00:00,  6.55it/s, loss=0.0765]  \n",
      "Epoch 14: 100%|██████████| 285/285 [00:41<00:00,  6.88it/s, loss=0.0201] \n",
      "Epoch 15: 100%|██████████| 285/285 [00:42<00:00,  6.76it/s, loss=0.00141] \n",
      "Epoch 16: 100%|██████████| 285/285 [00:42<00:00,  6.72it/s, loss=0.000378]\n",
      "Epoch 17: 100%|██████████| 285/285 [00:42<00:00,  6.75it/s, loss=0.000277]\n",
      "Epoch 18: 100%|██████████| 285/285 [00:41<00:00,  6.92it/s, loss=0.000208]\n",
      "Epoch 19: 100%|██████████| 285/285 [00:41<00:00,  6.90it/s, loss=0.000155]\n",
      "Epoch 20: 100%|██████████| 285/285 [00:41<00:00,  6.89it/s, loss=0.000126]\n",
      "Epoch 21: 100%|██████████| 285/285 [00:41<00:00,  6.85it/s, loss=0.00011] \n",
      "Epoch 22: 100%|██████████| 285/285 [00:41<00:00,  6.87it/s, loss=8.22e-5]\n",
      "Epoch 23: 100%|██████████| 285/285 [00:41<00:00,  6.90it/s, loss=7.17e-5]\n",
      "Epoch 24: 100%|██████████| 285/285 [00:40<00:00,  6.95it/s, loss=5.88e-5]\n",
      "Epoch 25: 100%|██████████| 285/285 [00:40<00:00,  7.03it/s, loss=5.16e-5]\n",
      "Epoch 26: 100%|██████████| 285/285 [00:40<00:00,  7.02it/s, loss=4.44e-5]\n",
      "Epoch 27: 100%|██████████| 285/285 [00:40<00:00,  7.00it/s, loss=3.86e-5]\n",
      "Epoch 28: 100%|██████████| 285/285 [00:40<00:00,  7.05it/s, loss=3.47e-5]\n",
      "Epoch 29: 100%|██████████| 285/285 [00:40<00:00,  7.01it/s, loss=2.84e-5]\n",
      "Epoch 30: 100%|██████████| 285/285 [00:40<00:00,  7.02it/s, loss=2.49e-5]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss on training: [0.3983791684372383, 0.23780503551426688, 0.10890476805435723, 0.04841021614132874, 0.028136948333121838, 0.037726678576916785, 0.039828131359944674, 0.004031372161504502, 0.0009477131381189745, 0.0045883520946964585, 0.0005715468706942997, 0.0003637235166439623, 0.07650427024391798, 0.020060825686934486, 0.0014054534455920618, 0.00037801638306853805, 0.00027650305549430316, 0.00020797239144724843, 0.00015463403192769506, 0.0001256919729808272, 0.00010997386094746425, 8.215038113504355e-05, 7.174736017253038e-05, 5.881307952222414e-05, 5.1640447494910885e-05, 4.438343501278085e-05, 3.8598988670089873e-05, 3.473970353202037e-05, 2.843091708455815e-05, 2.4920972955761387e-05]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Evaluating: 100%|██████████| 123/123 [00:05<00:00, 24.59it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test set: 0.9366053169734151\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameters: \n",
      "    Batch size: 4\n",
      "    Epochs: 30\n",
      "    Learning rate:0.0001\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1: 100%|██████████| 285/285 [00:40<00:00,  7.02it/s, loss=0.396]\n",
      "Epoch 2: 100%|██████████| 285/285 [00:40<00:00,  7.02it/s, loss=0.164] \n",
      "Epoch 3: 100%|██████████| 285/285 [00:40<00:00,  7.01it/s, loss=0.061]  \n",
      "Epoch 4: 100%|██████████| 285/285 [00:40<00:00,  7.03it/s, loss=0.046] \n",
      "Epoch 5: 100%|██████████| 285/285 [00:40<00:00,  7.05it/s, loss=0.00696]\n",
      "Epoch 6: 100%|██████████| 285/285 [00:40<00:00,  7.05it/s, loss=0.00059] \n",
      "Epoch 7: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=0.000465]\n",
      "Epoch 8: 100%|██████████| 285/285 [00:40<00:00,  7.05it/s, loss=0.000269]\n",
      "Epoch 9: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=0.000203]\n",
      "Epoch 10: 100%|██████████| 285/285 [00:40<00:00,  7.05it/s, loss=0.000198]\n",
      "Epoch 11: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=0.0718]  \n",
      "Epoch 12: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=0.0836] \n",
      "Epoch 13: 100%|██████████| 285/285 [00:40<00:00,  7.02it/s, loss=0.0196] \n",
      "Epoch 14: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=0.0158]\n",
      "Epoch 15: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=0.000621]\n",
      "Epoch 16: 100%|██████████| 285/285 [00:40<00:00,  7.03it/s, loss=0.00024] \n",
      "Epoch 17: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=0.000186]\n",
      "Epoch 18: 100%|██████████| 285/285 [00:40<00:00,  7.03it/s, loss=0.000127]\n",
      "Epoch 19: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=0.000104]\n",
      "Epoch 20: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=8.71e-5]\n",
      "Epoch 21: 100%|██████████| 285/285 [00:40<00:00,  7.03it/s, loss=6.94e-5]\n",
      "Epoch 22: 100%|██████████| 285/285 [00:40<00:00,  7.03it/s, loss=6.05e-5]\n",
      "Epoch 23: 100%|██████████| 285/285 [00:40<00:00,  7.03it/s, loss=5.13e-5]\n",
      "Epoch 24: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=4.2e-5] \n",
      "Epoch 25: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=3.79e-5]\n",
      "Epoch 26: 100%|██████████| 285/285 [00:40<00:00,  7.04it/s, loss=3.16e-5]\n",
      "Epoch 27: 100%|██████████| 285/285 [00:40<00:00,  7.03it/s, loss=2.77e-5]\n",
      "Epoch 28: 100%|██████████| 285/285 [00:40<00:00,  7.02it/s, loss=2.48e-5]\n",
      "Epoch 29: 100%|██████████| 285/285 [00:41<00:00,  6.87it/s, loss=2.12e-5]\n",
      "Epoch 30: 100%|██████████| 285/285 [00:42<00:00,  6.66it/s, loss=1.91e-5]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss on training: [0.39589022244735245, 0.16369920243067962, 0.061010796323483975, 0.046036071573759905, 0.0069605248503329786, 0.000589694246075835, 0.00046498795645515714, 0.00026943534081638385, 0.00020308298326358852, 0.00019750182448963945, 0.07176045661885598, 0.08355385805704843, 0.01957161137946522, 0.015824372113045108, 0.0006205886387441863, 0.00024031586893997564, 0.00018602138006826863, 0.00012690427673780441, 0.00010392707009215659, 8.710881946873933e-05, 6.938176828267037e-05, 6.049591609457581e-05, 5.131427484197877e-05, 4.203188605938442e-05, 3.791109270343696e-05, 3.155920615091944e-05, 2.7707141373212562e-05, 2.4764115254296675e-05, 2.115331638043771e-05, 1.9081308681435124e-05]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Evaluating: 100%|██████████| 123/123 [00:05<00:00, 22.92it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test set: 0.9263803680981595\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameters: \n",
      "    Batch size: 8\n",
      "    Epochs: 25\n",
      "    Learning rate:1e-05\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1: 100%|██████████| 143/143 [00:38<00:00,  3.71it/s, loss=0.465]\n",
      "Epoch 2: 100%|██████████| 143/143 [00:38<00:00,  3.72it/s, loss=0.242]\n",
      "Epoch 3: 100%|██████████| 143/143 [00:38<00:00,  3.75it/s, loss=0.116] \n",
      "Epoch 4: 100%|██████████| 143/143 [00:37<00:00,  3.78it/s, loss=0.0497]\n",
      "Epoch 5: 100%|██████████| 143/143 [00:38<00:00,  3.72it/s, loss=0.054] \n",
      "Epoch 6: 100%|██████████| 143/143 [00:37<00:00,  3.77it/s, loss=0.0257]\n",
      "Epoch 7: 100%|██████████| 143/143 [00:37<00:00,  3.78it/s, loss=0.0142] \n",
      "Epoch 8: 100%|██████████| 143/143 [00:38<00:00,  3.73it/s, loss=0.00734]\n",
      "Epoch 9: 100%|██████████| 143/143 [00:37<00:00,  3.83it/s, loss=0.0309] \n",
      "Epoch 10: 100%|██████████| 143/143 [00:37<00:00,  3.83it/s, loss=0.00383]\n",
      "Epoch 11: 100%|██████████| 143/143 [00:37<00:00,  3.78it/s, loss=0.00314] \n",
      "Epoch 12: 100%|██████████| 143/143 [00:35<00:00,  4.02it/s, loss=0.000851]\n",
      "Epoch 13: 100%|██████████| 143/143 [00:35<00:00,  4.02it/s, loss=0.000748]\n",
      "Epoch 14: 100%|██████████| 143/143 [00:35<00:00,  3.97it/s, loss=0.000526]\n",
      "Epoch 15: 100%|██████████| 143/143 [00:36<00:00,  3.94it/s, loss=0.000441]\n",
      "Epoch 16: 100%|██████████| 143/143 [00:38<00:00,  3.73it/s, loss=0.000361]\n",
      "Epoch 17: 100%|██████████| 143/143 [00:37<00:00,  3.82it/s, loss=0.000306]\n",
      "Epoch 18: 100%|██████████| 143/143 [00:37<00:00,  3.82it/s, loss=0.000263]\n",
      "Epoch 19: 100%|██████████| 143/143 [00:37<00:00,  3.83it/s, loss=0.00023] \n",
      "Epoch 20: 100%|██████████| 143/143 [00:37<00:00,  3.82it/s, loss=0.000214]\n",
      "Epoch 21: 100%|██████████| 143/143 [00:37<00:00,  3.82it/s, loss=0.000431]\n",
      "Epoch 22: 100%|██████████| 143/143 [00:37<00:00,  3.85it/s, loss=0.000229]\n",
      "Epoch 23: 100%|██████████| 143/143 [00:37<00:00,  3.82it/s, loss=0.000154]\n",
      "Epoch 24: 100%|██████████| 143/143 [00:37<00:00,  3.81it/s, loss=0.000134]\n",
      "Epoch 25: 100%|██████████| 143/143 [00:38<00:00,  3.74it/s, loss=0.000118]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss on training: [0.46534037121109195, 0.24213772212396134, 0.11558574031699788, 0.04966446809092914, 0.05395330854698078, 0.02573331059901031, 0.014245532511674857, 0.007338785965295209, 0.030871224329543097, 0.0038344307665849897, 0.0031381822993060563, 0.0008514482606124211, 0.0007482013560121658, 0.0005264691318551378, 0.00044097879150934463, 0.00036123192012700795, 0.000306043433293316, 0.00026259801203590147, 0.00023023421237883452, 0.00021366024960949365, 0.0004312299468539903, 0.0002289887735358817, 0.00015404376044008113, 0.0001335609879286087, 0.00011848464502970903]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Evaluating: 100%|██████████| 62/62 [00:05<00:00, 12.00it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test set: 0.9427402862985685\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameters: \n",
      "    Batch size: 8\n",
      "    Epochs: 25\n",
      "    Learning rate:0.0001\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1: 100%|██████████| 143/143 [00:38<00:00,  3.69it/s, loss=0.372]\n",
      "Epoch 2: 100%|██████████| 143/143 [00:37<00:00,  3.85it/s, loss=0.127] \n",
      "Epoch 3: 100%|██████████| 143/143 [00:37<00:00,  3.79it/s, loss=0.0489]\n",
      "Epoch 4: 100%|██████████| 143/143 [00:37<00:00,  3.81it/s, loss=0.0407] \n",
      "Epoch 5: 100%|██████████| 143/143 [00:37<00:00,  3.83it/s, loss=0.0161]\n",
      "Epoch 6: 100%|██████████| 143/143 [00:36<00:00,  3.97it/s, loss=0.00471]\n",
      "Epoch 7: 100%|██████████| 143/143 [00:36<00:00,  3.91it/s, loss=0.0214] \n",
      "Epoch 8: 100%|██████████| 143/143 [00:33<00:00,  4.25it/s, loss=0.0251]\n",
      "Epoch 9: 100%|██████████| 143/143 [00:34<00:00,  4.14it/s, loss=0.00869]\n",
      "Epoch 10: 100%|██████████| 143/143 [00:36<00:00,  3.90it/s, loss=0.00448] \n",
      "Epoch 11: 100%|██████████| 143/143 [00:36<00:00,  3.89it/s, loss=0.00109] \n",
      "Epoch 12: 100%|██████████| 143/143 [00:32<00:00,  4.33it/s, loss=0.000475]\n",
      "Epoch 13: 100%|██████████| 143/143 [00:33<00:00,  4.29it/s, loss=0.000328]\n",
      "Epoch 14: 100%|██████████| 143/143 [00:33<00:00,  4.24it/s, loss=0.00031] \n",
      "Epoch 15: 100%|██████████| 143/143 [00:33<00:00,  4.25it/s, loss=0.000276]\n",
      "Epoch 16: 100%|██████████| 143/143 [00:33<00:00,  4.25it/s, loss=0.000191]\n",
      "Epoch 17: 100%|██████████| 143/143 [00:37<00:00,  3.80it/s, loss=0.000174]\n",
      "Epoch 18: 100%|██████████| 143/143 [00:37<00:00,  3.79it/s, loss=0.000167]\n",
      "Epoch 19: 100%|██████████| 143/143 [00:37<00:00,  3.81it/s, loss=0.000139]\n",
      "Epoch 20: 100%|██████████| 143/143 [00:37<00:00,  3.77it/s, loss=0.000157]\n",
      "Epoch 21: 100%|██████████| 143/143 [00:37<00:00,  3.83it/s, loss=0.000116]\n",
      "Epoch 22: 100%|██████████| 143/143 [00:35<00:00,  3.97it/s, loss=0.000103]\n",
      "Epoch 23: 100%|██████████| 143/143 [00:35<00:00,  4.05it/s, loss=8.63e-5]\n",
      "Epoch 24: 100%|██████████| 143/143 [00:36<00:00,  3.97it/s, loss=8.05e-5]\n",
      "Epoch 25: 100%|██████████| 143/143 [00:36<00:00,  3.95it/s, loss=7.09e-5]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss on training: [0.3718735849732286, 0.12737644555281613, 0.048931865974075416, 0.04066898835288895, 0.016083747399254486, 0.004708329549016939, 0.021387514204415328, 0.02514700272683888, 0.008692253232645391, 0.004481279156276618, 0.0010946087296596252, 0.0004752334403696346, 0.0003283684614311474, 0.0003100050039551957, 0.0002763356813095295, 0.00019125158927432698, 0.0001744343053042348, 0.00016725267844528636, 0.00013894247759371292, 0.0001568201109844299, 0.00011622645587732013, 0.00010331661061471706, 8.630061792710412e-05, 8.049877614358446e-05, 7.091635398337149e-05]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Evaluating: 100%|██████████| 62/62 [00:04<00:00, 12.87it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test set: 0.934560327198364\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameters: \n",
      "    Batch size: 8\n",
      "    Epochs: 30\n",
      "    Learning rate:1e-05\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1: 100%|██████████| 143/143 [00:35<00:00,  4.01it/s, loss=0.39] \n",
      "Epoch 2: 100%|██████████| 143/143 [00:35<00:00,  3.99it/s, loss=0.184] \n",
      "Epoch 3: 100%|██████████| 143/143 [00:36<00:00,  3.88it/s, loss=0.0692]\n",
      "Epoch 4: 100%|██████████| 143/143 [00:37<00:00,  3.81it/s, loss=0.0274] \n",
      "Epoch 5: 100%|██████████| 143/143 [00:37<00:00,  3.86it/s, loss=0.0273] \n",
      "Epoch 6: 100%|██████████| 143/143 [00:37<00:00,  3.86it/s, loss=0.019] \n",
      "Epoch 7: 100%|██████████| 143/143 [00:37<00:00,  3.81it/s, loss=0.00491]\n",
      "Epoch 8: 100%|██████████| 143/143 [00:34<00:00,  4.17it/s, loss=0.00876]\n",
      "Epoch 9: 100%|██████████| 143/143 [00:34<00:00,  4.16it/s, loss=0.0284]\n",
      "Epoch 10: 100%|██████████| 143/143 [00:34<00:00,  4.13it/s, loss=0.00584] \n",
      "Epoch 11: 100%|██████████| 143/143 [00:34<00:00,  4.12it/s, loss=0.00213]\n",
      "Epoch 12: 100%|██████████| 143/143 [00:36<00:00,  3.96it/s, loss=0.000589]\n",
      "Epoch 13: 100%|██████████| 143/143 [00:36<00:00,  3.97it/s, loss=0.000254]\n",
      "Epoch 14: 100%|██████████| 143/143 [00:36<00:00,  3.95it/s, loss=0.000187]\n",
      "Epoch 15: 100%|██████████| 143/143 [00:36<00:00,  3.97it/s, loss=0.000175]\n",
      "Epoch 16: 100%|██████████| 143/143 [00:36<00:00,  3.95it/s, loss=0.000141]\n",
      "Epoch 17: 100%|██████████| 143/143 [00:36<00:00,  3.96it/s, loss=0.000152]\n",
      "Epoch 18: 100%|██████████| 143/143 [00:36<00:00,  3.95it/s, loss=0.000101]\n",
      "Epoch 19: 100%|██████████| 143/143 [00:36<00:00,  3.94it/s, loss=7.95e-5]\n",
      "Epoch 20: 100%|██████████| 143/143 [00:36<00:00,  3.92it/s, loss=7.25e-5]\n",
      "Epoch 21: 100%|██████████| 143/143 [00:36<00:00,  3.91it/s, loss=7e-5]   \n",
      "Epoch 22: 100%|██████████| 143/143 [00:37<00:00,  3.79it/s, loss=6.21e-5]\n",
      "Epoch 23: 100%|██████████| 143/143 [00:36<00:00,  3.87it/s, loss=5.45e-5]\n",
      "Epoch 24: 100%|██████████| 143/143 [00:36<00:00,  3.88it/s, loss=5.14e-5]\n",
      "Epoch 25: 100%|██████████| 143/143 [00:37<00:00,  3.79it/s, loss=5.33e-5]\n",
      "Epoch 26: 100%|██████████| 143/143 [00:37<00:00,  3.86it/s, loss=4.57e-5]\n",
      "Epoch 27: 100%|██████████| 143/143 [00:36<00:00,  3.88it/s, loss=3.76e-5]\n",
      "Epoch 28: 100%|██████████| 143/143 [00:36<00:00,  3.92it/s, loss=3.66e-5]\n",
      "Epoch 29: 100%|██████████| 143/143 [00:36<00:00,  3.91it/s, loss=3.25e-5]\n",
      "Epoch 30: 100%|██████████| 143/143 [00:36<00:00,  3.90it/s, loss=3.03e-5]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss on training: [0.3904003064844992, 0.18365405161272397, 0.06921578033898886, 0.027396978305691815, 0.027302698045399143, 0.019034763209121237, 0.004909429539416728, 0.008756036864482909, 0.028448599951493982, 0.0058378839002998706, 0.002129754528217669, 0.0005893640527892467, 0.0002536722139876264, 0.00018672251696755354, 0.0001745361333995426, 0.00014062511972379643, 0.00015174155815511004, 0.0001005324440128235, 7.950291717327297e-05, 7.248241535263377e-05, 7.00487759207714e-05, 6.210519063165997e-05, 5.4500454753251437e-05, 5.135568355106736e-05, 5.3308894427312486e-05, 4.568818770745896e-05, 3.7622912726861626e-05, 3.662106343011789e-05, 3.2500887004812445e-05, 3.0255398685169582e-05]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Evaluating: 100%|██████████| 62/62 [00:05<00:00, 11.92it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test set: 0.9366053169734151\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Parameters: \n",
      "    Batch size: 8\n",
      "    Epochs: 30\n",
      "    Learning rate:0.0001\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1: 100%|██████████| 143/143 [00:38<00:00,  3.75it/s, loss=0.402]\n",
      "Epoch 2: 100%|██████████| 143/143 [00:37<00:00,  3.78it/s, loss=0.181]\n",
      "Epoch 3: 100%|██████████| 143/143 [00:37<00:00,  3.83it/s, loss=0.0827]\n",
      "Epoch 4: 100%|██████████| 143/143 [00:37<00:00,  3.82it/s, loss=0.0523]\n",
      "Epoch 5: 100%|██████████| 143/143 [00:37<00:00,  3.79it/s, loss=0.0173]\n",
      "Epoch 6: 100%|██████████| 143/143 [00:38<00:00,  3.69it/s, loss=0.0368]\n",
      "Epoch 7: 100%|██████████| 143/143 [00:37<00:00,  3.77it/s, loss=0.0143] \n",
      "Epoch 8: 100%|██████████| 143/143 [00:37<00:00,  3.80it/s, loss=0.00771]\n",
      "Epoch 9: 100%|██████████| 143/143 [00:37<00:00,  3.80it/s, loss=0.000812]\n",
      "Epoch 10: 100%|██████████| 143/143 [00:37<00:00,  3.79it/s, loss=0.00128] \n",
      "Epoch 11: 100%|██████████| 143/143 [00:37<00:00,  3.83it/s, loss=0.0069] \n",
      "Epoch 12: 100%|██████████| 143/143 [00:37<00:00,  3.83it/s, loss=0.000412]\n",
      "Epoch 13: 100%|██████████| 143/143 [00:37<00:00,  3.81it/s, loss=0.000301]\n",
      "Epoch 14: 100%|██████████| 143/143 [00:37<00:00,  3.84it/s, loss=0.000254]\n",
      "Epoch 15: 100%|██████████| 143/143 [00:37<00:00,  3.79it/s, loss=0.00021] \n",
      "Epoch 16: 100%|██████████| 143/143 [00:37<00:00,  3.78it/s, loss=0.000183]\n",
      "Epoch 17: 100%|██████████| 143/143 [00:37<00:00,  3.79it/s, loss=0.000168]\n",
      "Epoch 18: 100%|██████████| 143/143 [00:37<00:00,  3.86it/s, loss=0.000135]\n",
      "Epoch 19: 100%|██████████| 143/143 [00:36<00:00,  3.94it/s, loss=0.000116]\n",
      "Epoch 20: 100%|██████████| 143/143 [00:36<00:00,  3.94it/s, loss=0.000108]\n",
      "Epoch 21: 100%|██████████| 143/143 [00:35<00:00,  4.03it/s, loss=0.000101]\n",
      "Epoch 22: 100%|██████████| 143/143 [00:37<00:00,  3.86it/s, loss=8.33e-5]\n",
      "Epoch 23: 100%|██████████| 143/143 [00:36<00:00,  3.97it/s, loss=7.69e-5]\n",
      "Epoch 24: 100%|██████████| 143/143 [00:35<00:00,  4.07it/s, loss=7.04e-5]\n",
      "Epoch 25: 100%|██████████| 143/143 [00:35<00:00,  4.07it/s, loss=6.13e-5]\n",
      "Epoch 26: 100%|██████████| 143/143 [00:33<00:00,  4.25it/s, loss=5.69e-5]\n",
      "Epoch 27: 100%|██████████| 143/143 [00:35<00:00,  4.08it/s, loss=5.4e-5] \n",
      "Epoch 28: 100%|██████████| 143/143 [00:34<00:00,  4.13it/s, loss=4.94e-5]\n",
      "Epoch 29: 100%|██████████| 143/143 [00:34<00:00,  4.11it/s, loss=4.33e-5]\n",
      "Epoch 30: 100%|██████████| 143/143 [00:34<00:00,  4.15it/s, loss=4.07e-5]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss on training: [0.4017190036813279, 0.18092734443505745, 0.08273584879055568, 0.05234834711861933, 0.01734882214601588, 0.03676833902808049, 0.014269355966195047, 0.007714981415107117, 0.000812459602126513, 0.0012799818059584204, 0.0069029644203707705, 0.00041208399656730203, 0.00030127265250434043, 0.0002541157708648214, 0.00020963339320460492, 0.00018280641488418296, 0.00016832554332710074, 0.00013499941661183795, 0.00011643072390982595, 0.0001081050483211126, 0.00010095191055052374, 8.33166490261723e-05, 7.691617994149984e-05, 7.041069708778392e-05, 6.13399147718715e-05, 5.694443179445888e-05, 5.401432500147567e-05, 4.9446489318181155e-05, 4.3349363836028343e-05, 4.066928584990336e-05]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Evaluating: 100%|██████████| 62/62 [00:04<00:00, 13.42it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test set: 0.9304703476482618\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Parameters\n",
    "batch_size_list = [4, 8]\n",
    "epochs_list = [25, 30]\n",
    "learning_rate_list = [1e-5, 1e-4]\n",
    "\n",
    "for batch_size in batch_size_list:\n",
    "    for epochs in epochs_list:\n",
    "        for learning_rate in learning_rate_list:\n",
    "            model = BertForSequenceClassification.from_pretrained('bert-base-uncased').to(device)\n",
    "            \n",
    "            print(f'Parameters: \\n    Batch size: {batch_size}\\n    Epochs: {epochs}\\n    Learning rate:{learning_rate}\\n')\n",
    "            # Prepare for epoch_losses\n",
    "            epoch_losses = []\n",
    "            \n",
    "            # Create dataloader\n",
    "            train_dataloader = DataLoader(train_dataset, sampler=RandomSampler(train_dataset), batch_size=batch_size)\n",
    "            test_dataloader = DataLoader(test_dataset, sampler=SequentialSampler(test_dataset), batch_size=batch_size)\n",
    "            \n",
    "            # Optimizer\n",
    "            optimizer = AdamW(model.parameters(), lr=2e-5)\n",
    "            \n",
    "            model.train()\n",
    "            \n",
    "            # Training loop\n",
    "            for epoch in range(epochs):\n",
    "                total_loss = 0\n",
    "                progress_bar = tqdm(train_dataloader, desc=f\"Epoch {epoch+1}\")\n",
    "            \n",
    "                for batch in progress_bar:\n",
    "                    # b_input_ids, b_input_mask, b_labels = batch\n",
    "                    b_input_ids, b_input_mask, b_labels = batch[0].to(device), batch[1].to(device), batch[2].to(device)\n",
    "                    model.zero_grad()\n",
    "            \n",
    "                    outputs = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask, labels=b_labels)\n",
    "                    loss = outputs.loss\n",
    "                    total_loss += loss.item()\n",
    "                    loss.backward()\n",
    "                    optimizer.step()\n",
    "            \n",
    "                    # Update the progress bar\n",
    "                    progress_bar.set_postfix({'loss': total_loss/len(train_dataloader)})\n",
    "                    \n",
    "                # Calculate and store the average loss for this epoch\n",
    "                avg_loss = total_loss / len(train_dataloader)\n",
    "                epoch_losses.append(avg_loss)\n",
    "                \n",
    "                # Save the model after each epoch\n",
    "                # model_save_file = os.path.join(model_save_path, f'bert_model_epoch_{epoch+1}.pt')\n",
    "                # torch.save(model.state_dict(), model_save_file)\n",
    "            \n",
    "                # Closing the progress bar and printing the epoch loss\n",
    "                progress_bar.close()\n",
    "                # print(f\"Epoch {epoch+1} finished. Loss: {total_loss/len(train_dataloader)}\")\n",
    "            print(f'Loss on training: {epoch_losses}')\n",
    "\n",
    "            # Switch to evaluation mode\n",
    "            model.eval()\n",
    "            predictions, true_labels = [], []\n",
    "            # Add tqdm progress bar\n",
    "            for batch in tqdm(test_dataloader, desc=\"Evaluating\"):\n",
    "                batch = tuple(t.to(device) for t in batch)\n",
    "                b_input_ids, b_input_mask, b_labels = batch\n",
    "            \n",
    "                with torch.no_grad():\n",
    "                    outputs = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask)\n",
    "                \n",
    "                logits = outputs.logits\n",
    "                logits = logits.detach().cpu().numpy()\n",
    "                label_ids = b_labels.to('cpu').numpy()\n",
    "            \n",
    "                batch_preds = np.argmax(logits, axis=1)\n",
    "                predictions.extend(batch_preds)\n",
    "                true_labels.extend(label_ids)\n",
    "            \n",
    "            # Calculate the accuracy\n",
    "            accuracy = accuracy_score(true_labels, predictions)\n",
    "            print(f\"Accuracy on test set: {accuracy}\")\n",
    "\n",
    "            # Plotting the training loss\n",
    "            plt.plot(epoch_losses, label='Training Loss')\n",
    "            plt.xlabel('Epochs')\n",
    "            plt.ylabel('Loss')\n",
    "            plt.title('Training Loss Over Epochs')\n",
    "            plt.legend()\n",
    "            plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4e32cb9c7a9fea54",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "## Config 9"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "3b1881fd2537daf3",
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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",
      "Epoch 1: 100%|██████████| 570/570 [00:51<00:00, 11.04it/s, loss=0.605]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1 finished. Loss: 0.604719248100331\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 2: 100%|██████████| 570/570 [00:52<00:00, 10.96it/s, loss=0.409] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 2 finished. Loss: 0.4085413656998099\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 3: 100%|██████████| 570/570 [00:52<00:00, 10.95it/s, loss=0.303] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3 finished. Loss: 0.3029960371815322\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 4: 100%|██████████| 570/570 [00:52<00:00, 10.93it/s, loss=0.218] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 4 finished. Loss: 0.21800761774443744\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 5: 100%|██████████| 570/570 [00:51<00:00, 11.05it/s, loss=0.176] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 5 finished. Loss: 0.1761193873981635\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 6: 100%|██████████| 570/570 [00:51<00:00, 11.06it/s, loss=0.143] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 6 finished. Loss: 0.143142576022237\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 7: 100%|██████████| 570/570 [00:52<00:00, 10.88it/s, loss=0.104]  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 7 finished. Loss: 0.1041122071935158\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 8: 100%|██████████| 570/570 [00:53<00:00, 10.69it/s, loss=0.0758]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 8 finished. Loss: 0.07579711706688007\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 9: 100%|██████████| 570/570 [00:52<00:00, 10.76it/s, loss=0.0624] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 9 finished. Loss: 0.062416773550865945\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 10: 100%|██████████| 570/570 [00:52<00:00, 10.79it/s, loss=0.0505]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 10 finished. Loss: 0.05047385521282099\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 11: 100%|██████████| 570/570 [00:52<00:00, 10.89it/s, loss=0.0421] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 11 finished. Loss: 0.04212044457120723\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 12: 100%|██████████| 570/570 [00:52<00:00, 10.88it/s, loss=0.0345] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 12 finished. Loss: 0.03450262523215395\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 13: 100%|██████████| 570/570 [00:51<00:00, 10.98it/s, loss=0.0178] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 13 finished. Loss: 0.01784964993906518\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 14: 100%|██████████| 570/570 [00:52<00:00, 10.96it/s, loss=0.0165] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 14 finished. Loss: 0.01654869586090443\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 15: 100%|██████████| 570/570 [00:51<00:00, 10.99it/s, loss=0.0152] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 15 finished. Loss: 0.015175077006049259\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 16: 100%|██████████| 570/570 [00:51<00:00, 11.02it/s, loss=0.00968]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 16 finished. Loss: 0.009684497025680907\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 17: 100%|██████████| 570/570 [00:51<00:00, 10.98it/s, loss=0.015]  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 17 finished. Loss: 0.015013688844792022\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 18: 100%|██████████| 570/570 [00:51<00:00, 11.02it/s, loss=0.00429] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 18 finished. Loss: 0.004290825295519378\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 19: 100%|██████████| 570/570 [00:51<00:00, 11.02it/s, loss=0.014]   \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 19 finished. Loss: 0.013970253145395729\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 20: 100%|██████████| 570/570 [00:51<00:00, 10.96it/s, loss=0.0162] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 20 finished. Loss: 0.016150779417397357\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 21: 100%|██████████| 570/570 [00:51<00:00, 11.02it/s, loss=0.0187] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 21 finished. Loss: 0.018710570084237446\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 22: 100%|██████████| 570/570 [00:52<00:00, 10.93it/s, loss=0.0078]  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 22 finished. Loss: 0.007799731129310665\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 23: 100%|██████████| 570/570 [00:52<00:00, 10.89it/s, loss=0.00294] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 23 finished. Loss: 0.0029353153321017888\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 24: 100%|██████████| 570/570 [00:53<00:00, 10.64it/s, loss=0.00389] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 24 finished. Loss: 0.00388891459371192\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 25: 100%|██████████| 570/570 [00:53<00:00, 10.57it/s, loss=0.0111]  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 25 finished. Loss: 0.011072590641816262\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Evaluating: 100%|██████████| 245/245 [00:05<00:00, 42.48it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test set: 0.9079754601226994\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "epochs = 25\n",
    "batch_size = 2\n",
    "# Prepare for epoch_losses\n",
    "epoch_losses = []\n",
    "\n",
    "# Create dataloader\n",
    "train_dataloader = DataLoader(train_dataset, sampler=RandomSampler(train_dataset), batch_size=batch_size)\n",
    "test_dataloader = DataLoader(test_dataset, sampler=SequentialSampler(test_dataset), batch_size=batch_size)\n",
    "\n",
    "model = BertForSequenceClassification.from_pretrained('bert-base-uncased').to(device)\n",
    "# Optimizer\n",
    "optimizer = AdamW(model.parameters(), lr=1e-6)\n",
    "\n",
    "model.train()\n",
    "# Training loop\n",
    "for epoch in range(epochs):\n",
    "    total_loss = 0\n",
    "    progress_bar = tqdm(train_dataloader, desc=f\"Epoch {epoch+1}\")\n",
    "\n",
    "    for batch in progress_bar:\n",
    "        # b_input_ids, b_input_mask, b_labels = batch\n",
    "        b_input_ids, b_input_mask, b_labels = batch[0].to(device), batch[1].to(device), batch[2].to(device)\n",
    "        model.zero_grad()\n",
    "\n",
    "        outputs = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask, labels=b_labels)\n",
    "        loss = outputs.loss\n",
    "        total_loss += loss.item()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "        # Update the progress bar\n",
    "        progress_bar.set_postfix({'loss': total_loss/len(train_dataloader)})\n",
    "        \n",
    "    # Calculate and store the average loss for this epoch\n",
    "    avg_loss = total_loss / len(train_dataloader)\n",
    "    epoch_losses.append(avg_loss)\n",
    "    \n",
    "    # Save the model after each epoch\n",
    "    # model_save_file = os.path.join(model_save_path, f'bert_model_epoch_{epoch+1}.pt')\n",
    "    # torch.save(model.state_dict(), model_save_file)\n",
    "\n",
    "    # Closing the progress bar and printing the epoch loss\n",
    "    progress_bar.close()\n",
    "    print(f\"Epoch {epoch+1} finished. Loss: {total_loss/len(train_dataloader)}\")\n",
    "\n",
    "# Switch to evaluation mode\n",
    "model.eval()\n",
    "predictions, true_labels = [], []\n",
    "# Add tqdm progress bar\n",
    "for batch in tqdm(test_dataloader, desc=\"Evaluating\"):\n",
    "    batch = tuple(t.to(device) for t in batch)\n",
    "    b_input_ids, b_input_mask, b_labels = batch\n",
    "\n",
    "    with torch.no_grad():\n",
    "        outputs = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask)\n",
    "    \n",
    "    logits = outputs.logits\n",
    "    logits = logits.detach().cpu().numpy()\n",
    "    label_ids = b_labels.to('cpu').numpy()\n",
    "\n",
    "    batch_preds = np.argmax(logits, axis=1)\n",
    "    predictions.extend(batch_preds)\n",
    "    true_labels.extend(label_ids)\n",
    "\n",
    "# Calculate the accuracy\n",
    "accuracy = accuracy_score(true_labels, predictions)\n",
    "print(f\"Accuracy on test set: {accuracy}\")\n",
    "\n",
    "# Plotting the training loss\n",
    "plt.plot(epoch_losses, label='Training Loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Training Loss Over Epochs')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a23c15225f1d5a1e",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "## Config 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "55e1130d4b692390",
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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",
      "Epoch 1: 100%|██████████| 143/143 [00:37<00:00,  3.85it/s, loss=0.419]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1 finished. Loss: 0.4189913151199584\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 2: 100%|██████████| 143/143 [00:34<00:00,  4.11it/s, loss=0.217]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 2 finished. Loss: 0.21695001398610486\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 3: 100%|██████████| 143/143 [00:35<00:00,  4.08it/s, loss=0.104] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3 finished. Loss: 0.10395444329775921\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 4: 100%|██████████| 143/143 [00:34<00:00,  4.12it/s, loss=0.0799]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 4 finished. Loss: 0.07987006853979367\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 5: 100%|██████████| 143/143 [00:34<00:00,  4.11it/s, loss=0.029] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 5 finished. Loss: 0.029038808004070934\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 6: 100%|██████████| 143/143 [00:35<00:00,  4.05it/s, loss=0.0313] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 6 finished. Loss: 0.03131093059656686\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 7: 100%|██████████| 143/143 [00:34<00:00,  4.12it/s, loss=0.0492]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 7 finished. Loss: 0.0492200924000553\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 8: 100%|██████████| 143/143 [00:35<00:00,  4.03it/s, loss=0.00214]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 8 finished. Loss: 0.0021441811768963307\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 9: 100%|██████████| 143/143 [00:35<00:00,  4.07it/s, loss=0.00145]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 9 finished. Loss: 0.0014514960336394712\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 10: 100%|██████████| 143/143 [00:35<00:00,  4.06it/s, loss=0.000926]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 10 finished. Loss: 0.0009260845333549448\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 11: 100%|██████████| 143/143 [00:35<00:00,  4.07it/s, loss=0.000731]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 11 finished. Loss: 0.0007314611567425301\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 12: 100%|██████████| 143/143 [00:35<00:00,  4.07it/s, loss=0.000707]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 12 finished. Loss: 0.0007073426624463701\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 13: 100%|██████████| 143/143 [00:34<00:00,  4.14it/s, loss=0.0005]  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 13 finished. Loss: 0.0004997134293751934\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 14: 100%|██████████| 143/143 [00:35<00:00,  4.05it/s, loss=0.000405]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 14 finished. Loss: 0.00040457841420832736\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 15: 100%|██████████| 143/143 [00:34<00:00,  4.09it/s, loss=0.000349]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 15 finished. Loss: 0.00034865639168423944\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 16: 100%|██████████| 143/143 [00:35<00:00,  4.08it/s, loss=0.000303]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 16 finished. Loss: 0.00030294174181502573\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 17: 100%|██████████| 143/143 [00:35<00:00,  4.09it/s, loss=0.00026] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 17 finished. Loss: 0.0002598815123995089\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 18: 100%|██████████| 143/143 [00:35<00:00,  4.08it/s, loss=0.000231]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 18 finished. Loss: 0.00023061735240068496\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 19: 100%|██████████| 143/143 [00:35<00:00,  4.07it/s, loss=0.000207]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 19 finished. Loss: 0.0002069762387716739\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 20: 100%|██████████| 143/143 [00:35<00:00,  4.05it/s, loss=0.000178]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 20 finished. Loss: 0.00017818043933867092\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 21: 100%|██████████| 143/143 [00:35<00:00,  4.07it/s, loss=0.000156]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 21 finished. Loss: 0.00015567666025697157\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 22: 100%|██████████| 143/143 [00:34<00:00,  4.09it/s, loss=0.000142]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 22 finished. Loss: 0.0001417941845696211\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 23: 100%|██████████| 143/143 [00:35<00:00,  4.06it/s, loss=0.000128]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 23 finished. Loss: 0.00012827830897339692\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 24: 100%|██████████| 143/143 [00:35<00:00,  4.08it/s, loss=0.000115]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 24 finished. Loss: 0.0001151122382176076\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 25: 100%|██████████| 143/143 [00:34<00:00,  4.14it/s, loss=0.000103]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 25 finished. Loss: 0.00010258485557502494\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "epochs = 25\n",
    "batch_size = 8\n",
    "# Prepare for epoch_losses\n",
    "epoch_losses = []\n",
    "\n",
    "# Create dataloader\n",
    "train_dataloader = DataLoader(train_dataset, sampler=RandomSampler(train_dataset), batch_size=batch_size)\n",
    "test_dataloader = DataLoader(test_dataset, sampler=SequentialSampler(test_dataset), batch_size=batch_size)\n",
    "\n",
    "model = BertForSequenceClassification.from_pretrained('bert-base-uncased').to(device)\n",
    "# Optimizer\n",
    "optimizer = AdamW(model.parameters(), lr=2e-5)\n",
    "\n",
    "model.train()\n",
    "# Training loop\n",
    "for epoch in range(epochs):\n",
    "    total_loss = 0\n",
    "    progress_bar = tqdm(train_dataloader, desc=f\"Epoch {epoch+1}\")\n",
    "\n",
    "    for batch in progress_bar:\n",
    "        # b_input_ids, b_input_mask, b_labels = batch\n",
    "        b_input_ids, b_input_mask, b_labels = batch[0].to(device), batch[1].to(device), batch[2].to(device)\n",
    "        model.zero_grad()\n",
    "\n",
    "        outputs = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask, labels=b_labels)\n",
    "        loss = outputs.loss\n",
    "        total_loss += loss.item()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "        # Update the progress bar\n",
    "        progress_bar.set_postfix({'loss': total_loss/len(train_dataloader)})\n",
    "        \n",
    "    # Calculate and store the average loss for this epoch\n",
    "    avg_loss = total_loss / len(train_dataloader)\n",
    "    epoch_losses.append(avg_loss)\n",
    "    \n",
    "    # Save the model after each epoch\n",
    "    # model_save_file = os.path.join(model_save_path, f'bert_model_epoch_{epoch+1}.pt')\n",
    "    # torch.save(model.state_dict(), model_save_file)\n",
    "\n",
    "    # Closing the progress bar and printing the epoch loss\n",
    "    progress_bar.close()\n",
    "    print(f\"Epoch {epoch+1} finished. Loss: {total_loss/len(train_dataloader)}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "396c4dac",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Parameters:\n",
      "Batch Size: 8\n",
      "Epochs: 25\n",
      "Learning Rate: 2e-05\n",
      "Beta1: 0.9\n",
      "Beta2: 0.999\n",
      "Training completed. Final model saved to models\\bert_final_model_1.pt\n"
     ]
    }
   ],
   "source": [
    "model_save_path = 'models'  \n",
    "os.makedirs(model_save_path, exist_ok=True)\n",
    "final_model_save_file = os.path.join(model_save_path, 'bert_final_model_1.pt')\n",
    "torch.save(model.state_dict(), final_model_save_file)\n",
    "\n",
    "# print paramter\n",
    "print(\"Training Parameters:\")\n",
    "print(f\"Batch Size: {batch_size}\")\n",
    "print(f\"Epochs: {epochs}\")\n",
    "print(f\"Learning Rate: {optimizer.defaults['lr']}\")\n",
    "print(f\"Beta1: {optimizer.defaults['betas'][0]}\")\n",
    "print(f\"Beta2: {optimizer.defaults['betas'][1]}\")\n",
    "\n",
    "print(f\"Training completed. Final model saved to {final_model_save_file}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "57af6c7d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plotting the training loss\n",
    "plt.plot(epoch_losses, label='Training Loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.title('Training Loss Over Epochs')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f0b6d5d9",
   "metadata": {},
   "source": [
    "### Evulation "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "2598b529f5b9ffaa",
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Evaluating: 100%|██████████| 62/62 [00:04<00:00, 13.04it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on test set: 0.9263803680981595\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
    "\n",
    "# Switch to evaluation mode\n",
    "model.eval()\n",
    "\n",
    "predictions, true_labels = [], []\n",
    "\n",
    "# Add tqdm progress bar\n",
    "for batch in tqdm(test_dataloader, desc=\"Evaluating\"):\n",
    "    batch = tuple(t.to(device) for t in batch)\n",
    "    b_input_ids, b_input_mask, b_labels = batch\n",
    "\n",
    "    with torch.no_grad():\n",
    "        outputs = model(b_input_ids, token_type_ids=None, attention_mask=b_input_mask)\n",
    "    \n",
    "    logits = outputs.logits\n",
    "    logits = logits.detach().cpu().numpy()\n",
    "    label_ids = b_labels.to('cpu').numpy()\n",
    "\n",
    "    batch_preds = np.argmax(logits, axis=1)\n",
    "    predictions.extend(batch_preds)\n",
    "    true_labels.extend(label_ids)\n",
    "\n",
    "# Calculate the accuracy\n",
    "accuracy = accuracy_score(true_labels, predictions)\n",
    "print(f\"Accuracy on test set: {accuracy}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e4de4e7b",
   "metadata": {},
   "source": [
    "### K-Fold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "c1e96fd8",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FOLD 0\n",
      "-------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "c:\\Users\\Andy Cui\\.conda\\envs\\nlp\\lib\\site-packages\\transformers\\optimization.py:411: 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",
      "Training Epoch 1/25, Fold 1/10: 100%|██████████| 184/184 [00:45<00:00,  4.01it/s, loss=0.306]\n",
      "Training Epoch 2/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.14] \n",
      "Training Epoch 3/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.0561]\n",
      "Training Epoch 4/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.17it/s, loss=0.0293]\n",
      "Training Epoch 5/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.17it/s, loss=0.0245]\n",
      "Training Epoch 6/25, Fold 1/10: 100%|██████████| 184/184 [00:43<00:00,  4.18it/s, loss=0.0301] \n",
      "Training Epoch 7/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.15it/s, loss=0.0765]\n",
      "Training Epoch 8/25, Fold 1/10: 100%|██████████| 184/184 [00:43<00:00,  4.19it/s, loss=0.00652]\n",
      "Training Epoch 9/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.17it/s, loss=0.00269]\n",
      "Training Epoch 10/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.13it/s, loss=0.000858]\n",
      "Training Epoch 11/25, Fold 1/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000503]\n",
      "Training Epoch 12/25, Fold 1/10: 100%|██████████| 184/184 [00:43<00:00,  4.18it/s, loss=0.00675] \n",
      "Training Epoch 13/25, Fold 1/10: 100%|██████████| 184/184 [00:43<00:00,  4.19it/s, loss=0.0198] \n",
      "Training Epoch 14/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.17it/s, loss=0.0346]\n",
      "Training Epoch 15/25, Fold 1/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.00553]\n",
      "Training Epoch 16/25, Fold 1/10: 100%|██████████| 184/184 [00:43<00:00,  4.24it/s, loss=0.00051] \n",
      "Training Epoch 17/25, Fold 1/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.00041] \n",
      "Training Epoch 18/25, Fold 1/10: 100%|██████████| 184/184 [00:46<00:00,  3.97it/s, loss=0.000215]\n",
      "Training Epoch 19/25, Fold 1/10: 100%|██████████| 184/184 [00:45<00:00,  4.06it/s, loss=0.0003]  \n",
      "Training Epoch 20/25, Fold 1/10: 100%|██████████| 184/184 [00:45<00:00,  4.07it/s, loss=0.000145]\n",
      "Training Epoch 21/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.000111]\n",
      "Training Epoch 22/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.15it/s, loss=9.31e-5]\n",
      "Training Epoch 23/25, Fold 1/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=9.44e-5]\n",
      "Training Epoch 24/25, Fold 1/10: 100%|██████████| 184/184 [00:45<00:00,  4.06it/s, loss=7.21e-5]\n",
      "Training Epoch 25/25, Fold 1/10: 100%|██████████| 184/184 [00:47<00:00,  3.90it/s, loss=6.29e-5]\n",
      "Validation, Fold 1/10: 100%|██████████| 21/21 [00:01<00:00, 12.44it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation accuracy: 0.9464285714285714\n",
      "FOLD 1\n",
      "-------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "Training Epoch 1/25, Fold 2/10: 100%|██████████| 184/184 [00:46<00:00,  3.93it/s, loss=0.379]\n",
      "Training Epoch 2/25, Fold 2/10: 100%|██████████| 184/184 [00:45<00:00,  4.06it/s, loss=0.163]\n",
      "Training Epoch 3/25, Fold 2/10: 100%|██████████| 184/184 [00:45<00:00,  4.07it/s, loss=0.0703]\n",
      "Training Epoch 4/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.0383]\n",
      "Training Epoch 5/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.16it/s, loss=0.0272]\n",
      "Training Epoch 6/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.0127] \n",
      "Training Epoch 7/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.0363]\n",
      "Training Epoch 8/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.00631]\n",
      "Training Epoch 9/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.15it/s, loss=0.0152]\n",
      "Training Epoch 10/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.16it/s, loss=0.00369]\n",
      "Training Epoch 11/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.15it/s, loss=0.01]  \n",
      "Training Epoch 12/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.15it/s, loss=0.00791]\n",
      "Training Epoch 13/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.15it/s, loss=0.00767]\n",
      "Training Epoch 14/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.13it/s, loss=0.00459]\n",
      "Training Epoch 15/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.15it/s, loss=0.0015] \n",
      "Training Epoch 16/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.16it/s, loss=0.000303]\n",
      "Training Epoch 17/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.000255]\n",
      "Training Epoch 18/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.0002]  \n",
      "Training Epoch 19/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.000175]\n",
      "Training Epoch 20/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.000151]\n",
      "Training Epoch 21/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.000138]\n",
      "Training Epoch 22/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.000119]\n",
      "Training Epoch 23/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.000108]\n",
      "Training Epoch 24/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.0001]  \n",
      "Training Epoch 25/25, Fold 2/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=8.78e-5]\n",
      "Validation, Fold 2/10: 100%|██████████| 21/21 [00:01<00:00, 13.34it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation accuracy: 0.9246031746031746\n",
      "FOLD 2\n",
      "-------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "Training Epoch 1/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.361]\n",
      "Training Epoch 2/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.13it/s, loss=0.17] \n",
      "Training Epoch 3/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.0709]\n",
      "Training Epoch 4/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.13it/s, loss=0.0294]\n",
      "Training Epoch 5/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.0305]\n",
      "Training Epoch 6/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.00319]\n",
      "Training Epoch 7/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.000628]\n",
      "Training Epoch 8/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.09it/s, loss=0.00044] \n",
      "Training Epoch 9/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.0039]  \n",
      "Training Epoch 10/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.15it/s, loss=0.0329]\n",
      "Training Epoch 11/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.0155]\n",
      "Training Epoch 12/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.0288] \n",
      "Training Epoch 13/25, Fold 3/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=0.019]  \n",
      "Training Epoch 14/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.00394]\n",
      "Training Epoch 15/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.000607]\n",
      "Training Epoch 16/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.000251]\n",
      "Training Epoch 17/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.000155]\n",
      "Training Epoch 18/25, Fold 3/10: 100%|██████████| 184/184 [00:45<00:00,  4.06it/s, loss=0.000112]\n",
      "Training Epoch 19/25, Fold 3/10: 100%|██████████| 184/184 [00:47<00:00,  3.88it/s, loss=8.64e-5]\n",
      "Training Epoch 20/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=7.51e-5]\n",
      "Training Epoch 21/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=6.63e-5]\n",
      "Training Epoch 22/25, Fold 3/10: 100%|██████████| 184/184 [00:45<00:00,  4.04it/s, loss=6.01e-5]\n",
      "Training Epoch 23/25, Fold 3/10: 100%|██████████| 184/184 [00:45<00:00,  4.07it/s, loss=5.3e-5] \n",
      "Training Epoch 24/25, Fold 3/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=4.55e-5]\n",
      "Training Epoch 25/25, Fold 3/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=4.08e-5]\n",
      "Validation, Fold 3/10: 100%|██████████| 21/21 [00:01<00:00, 13.27it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation accuracy: 0.9642857142857143\n",
      "FOLD 3\n",
      "-------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "Training Epoch 1/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.373]\n",
      "Training Epoch 2/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.16] \n",
      "Training Epoch 3/25, Fold 4/10: 100%|██████████| 184/184 [00:45<00:00,  4.09it/s, loss=0.0588]\n",
      "Training Epoch 4/25, Fold 4/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=0.0376]\n",
      "Training Epoch 5/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.0133]\n",
      "Training Epoch 6/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.13it/s, loss=0.0195]\n",
      "Training Epoch 7/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.0227]\n",
      "Training Epoch 8/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.13it/s, loss=0.0197]\n",
      "Training Epoch 9/25, Fold 4/10: 100%|██████████| 184/184 [00:45<00:00,  4.04it/s, loss=0.00857]\n",
      "Training Epoch 10/25, Fold 4/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=0.0217]  \n",
      "Training Epoch 11/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.0071] \n",
      "Training Epoch 12/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.000989]\n",
      "Training Epoch 13/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.000447]\n",
      "Training Epoch 14/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.000331]\n",
      "Training Epoch 15/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.000232]\n",
      "Training Epoch 16/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.000213]\n",
      "Training Epoch 17/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.000168]\n",
      "Training Epoch 18/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.0394]  \n",
      "Training Epoch 19/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.0354]\n",
      "Training Epoch 20/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.00523]\n",
      "Training Epoch 21/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.000778]\n",
      "Training Epoch 22/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.13it/s, loss=0.000187]\n",
      "Training Epoch 23/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.000154]\n",
      "Training Epoch 24/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.15it/s, loss=0.000131]\n",
      "Training Epoch 25/25, Fold 4/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.000117]\n",
      "Validation, Fold 4/10: 100%|██████████| 21/21 [00:01<00:00, 13.27it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation accuracy: 0.8769841269841271\n",
      "FOLD 4\n",
      "-------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "Training Epoch 1/25, Fold 5/10: 100%|██████████| 184/184 [00:44<00:00,  4.16it/s, loss=0.336]\n",
      "Training Epoch 2/25, Fold 5/10: 100%|██████████| 184/184 [00:44<00:00,  4.18it/s, loss=0.141]\n",
      "Training Epoch 3/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.065] \n",
      "Training Epoch 4/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.0314]\n",
      "Training Epoch 5/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.00404]\n",
      "Training Epoch 6/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0299] \n",
      "Training Epoch 7/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.21it/s, loss=0.0718]\n",
      "Training Epoch 8/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.21it/s, loss=0.0152]\n",
      "Training Epoch 9/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.00134]\n",
      "Training Epoch 10/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000748]\n",
      "Training Epoch 11/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.20it/s, loss=0.000537]\n",
      "Training Epoch 12/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000339]\n",
      "Training Epoch 13/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000426]\n",
      "Training Epoch 14/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.00032] \n",
      "Training Epoch 15/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0002]  \n",
      "Training Epoch 16/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.00017] \n",
      "Training Epoch 17/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000152]\n",
      "Training Epoch 18/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000132]\n",
      "Training Epoch 19/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.000117]\n",
      "Training Epoch 20/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000102]\n",
      "Training Epoch 21/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0177]  \n",
      "Training Epoch 22/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0426]\n",
      "Training Epoch 23/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.00932]\n",
      "Training Epoch 24/25, Fold 5/10: 100%|██████████| 184/184 [00:43<00:00,  4.20it/s, loss=0.00429]\n",
      "Training Epoch 25/25, Fold 5/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.000425]\n",
      "Validation, Fold 5/10: 100%|██████████| 21/21 [00:01<00:00, 13.40it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation accuracy: 0.9107142857142857\n",
      "FOLD 5\n",
      "-------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "Training Epoch 1/25, Fold 6/10: 100%|██████████| 184/184 [00:44<00:00,  4.17it/s, loss=0.326]\n",
      "Training Epoch 2/25, Fold 6/10: 100%|██████████| 184/184 [00:44<00:00,  4.16it/s, loss=0.12] \n",
      "Training Epoch 3/25, Fold 6/10: 100%|██████████| 184/184 [00:44<00:00,  4.16it/s, loss=0.0637]\n",
      "Training Epoch 4/25, Fold 6/10: 100%|██████████| 184/184 [00:44<00:00,  4.17it/s, loss=0.0557]\n",
      "Training Epoch 5/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.00867]\n",
      "Training Epoch 6/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.0202]\n",
      "Training Epoch 7/25, Fold 6/10: 100%|██████████| 184/184 [00:44<00:00,  4.16it/s, loss=0.0225]\n",
      "Training Epoch 8/25, Fold 6/10: 100%|██████████| 184/184 [00:46<00:00,  3.98it/s, loss=0.00272]\n",
      "Training Epoch 9/25, Fold 6/10: 100%|██████████| 184/184 [00:44<00:00,  4.17it/s, loss=0.0256] \n",
      "Training Epoch 10/25, Fold 6/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.00808]\n",
      "Training Epoch 11/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.006]   \n",
      "Training Epoch 12/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.00177]\n",
      "Training Epoch 13/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.000298]\n",
      "Training Epoch 14/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000276]\n",
      "Training Epoch 15/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.19it/s, loss=0.000175]\n",
      "Training Epoch 16/25, Fold 6/10: 100%|██████████| 184/184 [00:44<00:00,  4.13it/s, loss=0.000145]\n",
      "Training Epoch 17/25, Fold 6/10: 100%|██████████| 184/184 [00:44<00:00,  4.14it/s, loss=0.000133]\n",
      "Training Epoch 18/25, Fold 6/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.00012] \n",
      "Training Epoch 19/25, Fold 6/10: 100%|██████████| 184/184 [00:45<00:00,  4.06it/s, loss=0.000103]\n",
      "Training Epoch 20/25, Fold 6/10: 100%|██████████| 184/184 [00:45<00:00,  4.01it/s, loss=0.0649]\n",
      "Training Epoch 21/25, Fold 6/10: 100%|██████████| 184/184 [00:45<00:00,  4.09it/s, loss=0.00694] \n",
      "Training Epoch 22/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.0371] \n",
      "Training Epoch 23/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.00249]\n",
      "Training Epoch 24/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.00027] \n",
      "Training Epoch 25/25, Fold 6/10: 100%|██████████| 184/184 [00:43<00:00,  4.21it/s, loss=0.000185]\n",
      "Validation, Fold 6/10: 100%|██████████| 21/21 [00:01<00:00, 13.72it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation accuracy: 0.9166666666666666\n",
      "FOLD 6\n",
      "-------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "Training Epoch 1/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.357]\n",
      "Training Epoch 2/25, Fold 7/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.155]\n",
      "Training Epoch 3/25, Fold 7/10: 100%|██████████| 184/184 [00:45<00:00,  4.07it/s, loss=0.0704]\n",
      "Training Epoch 4/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.22it/s, loss=0.0297]\n",
      "Training Epoch 5/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0193]\n",
      "Training Epoch 6/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.24it/s, loss=0.0279]\n",
      "Training Epoch 7/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0259] \n",
      "Training Epoch 8/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0171]\n",
      "Training Epoch 9/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.00202]\n",
      "Training Epoch 10/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0011] \n",
      "Training Epoch 11/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000488]\n",
      "Training Epoch 12/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000372]\n",
      "Training Epoch 13/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0003]  \n",
      "Training Epoch 14/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000256]\n",
      "Training Epoch 15/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000227]\n",
      "Training Epoch 16/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.24it/s, loss=0.000343]\n",
      "Training Epoch 17/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.24it/s, loss=0.0333]  \n",
      "Training Epoch 18/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.24it/s, loss=0.0264]\n",
      "Training Epoch 19/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0157] \n",
      "Training Epoch 20/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000648]\n",
      "Training Epoch 21/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.24it/s, loss=0.000351]\n",
      "Training Epoch 22/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.00023] \n",
      "Training Epoch 23/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000262]\n",
      "Training Epoch 24/25, Fold 7/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.0636] \n",
      "Training Epoch 25/25, Fold 7/10: 100%|██████████| 184/184 [00:41<00:00,  4.42it/s, loss=0.0141] \n",
      "Validation, Fold 7/10: 100%|██████████| 21/21 [00:01<00:00, 14.68it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation accuracy: 0.9404761904761905\n",
      "FOLD 7\n",
      "-------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "Training Epoch 1/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.334]\n",
      "Training Epoch 2/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.145]\n",
      "Training Epoch 3/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0606]\n",
      "Training Epoch 4/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0364]\n",
      "Training Epoch 5/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.00626]\n",
      "Training Epoch 6/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0289] \n",
      "Training Epoch 7/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0281]\n",
      "Training Epoch 8/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0118] \n",
      "Training Epoch 9/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.00134]\n",
      "Training Epoch 10/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.000731]\n",
      "Training Epoch 11/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.00806]\n",
      "Training Epoch 12/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.000643]\n",
      "Training Epoch 13/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0278] \n",
      "Training Epoch 14/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.024] \n",
      "Training Epoch 15/25, Fold 8/10: 100%|██████████| 184/184 [00:41<00:00,  4.48it/s, loss=0.015] \n",
      "Training Epoch 16/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.00652]\n",
      "Training Epoch 17/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.00169]\n",
      "Training Epoch 18/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.000253]\n",
      "Training Epoch 19/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.000171]\n",
      "Training Epoch 20/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.000428]\n",
      "Training Epoch 21/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0354]\n",
      "Training Epoch 22/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.00511]\n",
      "Training Epoch 23/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.00607]\n",
      "Training Epoch 24/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.00811]\n",
      "Training Epoch 25/25, Fold 8/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0046] \n",
      "Validation, Fold 8/10: 100%|██████████| 21/21 [00:01<00:00, 14.67it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation accuracy: 0.9285714285714286\n",
      "FOLD 8\n",
      "-------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "Training Epoch 1/25, Fold 9/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.358]\n",
      "Training Epoch 2/25, Fold 9/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.178]\n",
      "Training Epoch 3/25, Fold 9/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0708]\n",
      "Training Epoch 4/25, Fold 9/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0701]\n",
      "Training Epoch 5/25, Fold 9/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0349]\n",
      "Training Epoch 6/25, Fold 9/10: 100%|██████████| 184/184 [00:40<00:00,  4.50it/s, loss=0.0336]\n",
      "Training Epoch 7/25, Fold 9/10: 100%|██████████| 184/184 [00:44<00:00,  4.17it/s, loss=0.00749]\n",
      "Training Epoch 8/25, Fold 9/10: 100%|██████████| 184/184 [00:44<00:00,  4.18it/s, loss=0.00161]\n",
      "Training Epoch 9/25, Fold 9/10: 100%|██████████| 184/184 [00:44<00:00,  4.16it/s, loss=0.000849]\n",
      "Training Epoch 10/25, Fold 9/10: 100%|██████████| 184/184 [00:44<00:00,  4.15it/s, loss=0.000693]\n",
      "Training Epoch 11/25, Fold 9/10: 100%|██████████| 184/184 [00:43<00:00,  4.24it/s, loss=0.000506]\n",
      "Training Epoch 12/25, Fold 9/10: 100%|██████████| 184/184 [00:43<00:00,  4.23it/s, loss=0.000391]\n",
      "Training Epoch 13/25, Fold 9/10: 100%|██████████| 184/184 [00:43<00:00,  4.18it/s, loss=0.000306]\n",
      "Training Epoch 14/25, Fold 9/10: 100%|██████████| 184/184 [00:45<00:00,  4.06it/s, loss=0.000257]\n",
      "Training Epoch 15/25, Fold 9/10: 100%|██████████| 184/184 [00:46<00:00,  3.93it/s, loss=0.000215]\n",
      "Training Epoch 16/25, Fold 9/10: 100%|██████████| 184/184 [00:46<00:00,  3.92it/s, loss=0.000182]\n",
      "Training Epoch 17/25, Fold 9/10: 100%|██████████| 184/184 [00:47<00:00,  3.91it/s, loss=0.000187]\n",
      "Training Epoch 18/25, Fold 9/10: 100%|██████████| 184/184 [00:47<00:00,  3.89it/s, loss=0.000143]\n",
      "Training Epoch 19/25, Fold 9/10: 100%|██████████| 184/184 [00:46<00:00,  3.93it/s, loss=0.000123]\n",
      "Training Epoch 20/25, Fold 9/10: 100%|██████████| 184/184 [00:45<00:00,  4.07it/s, loss=0.000108]\n",
      "Training Epoch 21/25, Fold 9/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=9.5e-5] \n",
      "Training Epoch 22/25, Fold 9/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=8.4e-5] \n",
      "Training Epoch 23/25, Fold 9/10: 100%|██████████| 184/184 [00:45<00:00,  4.07it/s, loss=7.25e-5]\n",
      "Training Epoch 24/25, Fold 9/10: 100%|██████████| 184/184 [00:45<00:00,  4.06it/s, loss=6.55e-5]\n",
      "Training Epoch 25/25, Fold 9/10: 100%|██████████| 184/184 [00:44<00:00,  4.09it/s, loss=5.98e-5]\n",
      "Validation, Fold 9/10: 100%|██████████| 21/21 [00:01<00:00, 12.68it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation accuracy: 0.9404761904761905\n",
      "FOLD 9\n",
      "-------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of BertForSequenceClassification were not initialized from the model checkpoint at bert-base-uncased 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",
      "Training Epoch 1/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.09it/s, loss=0.336]\n",
      "Training Epoch 2/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.138]\n",
      "Training Epoch 3/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=0.0379]\n",
      "Training Epoch 4/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.0484]\n",
      "Training Epoch 5/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=0.0143]\n",
      "Training Epoch 6/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=0.0142]\n",
      "Training Epoch 7/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.06it/s, loss=0.00176]\n",
      "Training Epoch 8/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.09it/s, loss=0.000949]\n",
      "Training Epoch 9/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.04it/s, loss=0.000477]\n",
      "Training Epoch 10/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=0.0198]  \n",
      "Training Epoch 11/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=0.0313]\n",
      "Training Epoch 12/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.07it/s, loss=0.0115] \n",
      "Training Epoch 13/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.07it/s, loss=0.00843]\n",
      "Training Epoch 14/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.10it/s, loss=0.00388] \n",
      "Training Epoch 15/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=0.000444]\n",
      "Training Epoch 16/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.11it/s, loss=0.000253]\n",
      "Training Epoch 17/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.03it/s, loss=0.000286]\n",
      "Training Epoch 18/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.18it/s, loss=0.000144]\n",
      "Training Epoch 19/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.13it/s, loss=0.000111]\n",
      "Training Epoch 20/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=9.17e-5]\n",
      "Training Epoch 21/25, Fold 10/10: 100%|██████████| 184/184 [00:45<00:00,  4.08it/s, loss=7.83e-5]\n",
      "Training Epoch 22/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.12it/s, loss=7.06e-5]\n",
      "Training Epoch 23/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.18it/s, loss=6.17e-5]\n",
      "Training Epoch 24/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.17it/s, loss=5.56e-5]\n",
      "Training Epoch 25/25, Fold 10/10: 100%|██████████| 184/184 [00:44<00:00,  4.17it/s, loss=4.57e-5]\n",
      "Validation, Fold 10/10: 100%|██████████| 21/21 [00:01<00:00, 13.68it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation accuracy: 0.9107142857142857\n",
      "Average performance across all folds: {'avg_train_loss': 0.001972767296634313, 'avg_val_loss': 0.445684893727481, 'avg_val_accuracy': 0.9259920634920634}\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "# Define the KFold cross-validator\n",
    "kf = KFold(n_splits=10, shuffle=True, random_state=42)\n",
    "\n",
    "# Initialize performance metric list\n",
    "fold_performance = []\n",
    "\n",
    "# Parameters\n",
    "batch_size = 8\n",
    "epochs = 25\n",
    "learning_rate = 2e-5\n",
    "\n",
    "# Define the device\n",
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "\n",
    "# Start the KFold cross-validation\n",
    "for fold, (train_ids, test_ids) in enumerate(kf.split(dataset)):\n",
    "    print(f\"FOLD {fold}\")\n",
    "    print(\"-------------------------------\")\n",
    "\n",
    "    # Split the data into training and validation sets for the current fold\n",
    "    train_subset = Subset(dataset, train_ids)\n",
    "    test_subset = Subset(dataset, test_ids)\n",
    "    \n",
    "    train_dataloader = DataLoader(train_subset, sampler=RandomSampler(train_subset), batch_size=batch_size)\n",
    "    test_dataloader = DataLoader(test_subset, sampler=SequentialSampler(test_subset), batch_size=batch_size)\n",
    "    \n",
    "    # Initialize the BERT model for sequence classification\n",
    "    model = BertForSequenceClassification.from_pretrained('bert-base-uncased')\n",
    "    model.to(device)\n",
    "    \n",
    "    # Define the optimizer\n",
    "    optimizer = AdamW(model.parameters(), lr=learning_rate)\n",
    "    model.train()\n",
    "    \n",
    "    # Training loop for the current fold\n",
    "    for epoch in range(epochs):\n",
    "        total_loss = 0\n",
    "        \n",
    "        train_progress_bar = tqdm(train_dataloader, desc=f\"Training Epoch {epoch+1}/{epochs}, Fold {fold+1}/10\")\n",
    "        \n",
    "        for step, batch in enumerate(train_progress_bar):\n",
    "            batch = tuple(t.to(device) for t in batch)\n",
    "            b_input_ids, b_input_mask, b_labels = batch\n",
    "            \n",
    "            model.zero_grad()\n",
    "            \n",
    "            outputs = model(b_input_ids, attention_mask=b_input_mask, labels=b_labels)\n",
    "            loss = outputs.loss\n",
    "            total_loss += loss.item()\n",
    "            \n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            \n",
    "            train_progress_bar.set_postfix(loss=total_loss/(step+1))\n",
    "        \n",
    "        train_progress_bar.close()\n",
    "        \n",
    "    # Calculate the average loss over the training data.\n",
    "    avg_train_loss = total_loss / len(train_dataloader)\n",
    "        \n",
    "    # Validation step\n",
    "    model.eval()\n",
    "    total_eval_accuracy = 0\n",
    "    total_eval_loss = 0\n",
    "\n",
    "    eval_progress_bar = tqdm(test_dataloader, desc=f\"Validation, Fold {fold+1}/10\")\n",
    "\n",
    "    for batch in eval_progress_bar:\n",
    "        batch = tuple(t.to(device) for t in batch)\n",
    "        b_input_ids, b_input_mask, b_labels = batch\n",
    "    \n",
    "        with torch.no_grad():\n",
    "            outputs = model(b_input_ids, attention_mask=b_input_mask, labels=b_labels)\n",
    "    \n",
    "        loss = outputs.loss\n",
    "        total_eval_loss += loss.item()\n",
    "    \n",
    "        # Move logits and labels to CPU\n",
    "        logits = outputs.logits.detach().cpu().numpy()\n",
    "        label_ids = b_labels.to('cpu').numpy()\n",
    "\n",
    "        # Calculate the accuracy for this batch of test sentences\n",
    "        preds = np.argmax(logits, axis=1)\n",
    "        batch_accuracy = np.mean(preds == label_ids)\n",
    "        total_eval_accuracy += batch_accuracy\n",
    "        \n",
    "    # Report the final accuracy for this validation run.\n",
    "    avg_val_accuracy = total_eval_accuracy / len(test_dataloader)\n",
    "    print(f\"Validation accuracy: {avg_val_accuracy}\")\n",
    "\n",
    "\n",
    "    # Calculate the average loss over all of the batches.\n",
    "    avg_val_loss = total_eval_loss / len(test_dataloader)\n",
    "    \n",
    "    # Record all statistics from this epoch.\n",
    "    fold_performance.append({\n",
    "        'fold': fold,\n",
    "        'train_loss': avg_train_loss,\n",
    "        'val_loss': avg_val_loss,\n",
    "        'val_accuracy': avg_val_accuracy\n",
    "    })\n",
    "\n",
    "    eval_progress_bar.close()\n",
    "\n",
    "# Calculate and print the average performance across all folds\n",
    "average_performance = {\n",
    "    'avg_train_loss': np.mean([x['train_loss'] for x in fold_performance]),\n",
    "    'avg_val_loss': np.mean([x['val_loss'] for x in fold_performance]),\n",
    "    'avg_val_accuracy': np.mean([x['val_accuracy'] for x in fold_performance])\n",
    "}\n",
    "print(f\"Average performance across all folds: {average_performance}\")"
   ]
  }
 ],
 "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": "3.10.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}