Upload 2 files

insert_into_db_sent_tran_tsdae.ipynb

@@ -0,0 +1,504 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "1aafbf18-de38-4fcf-8245-e2e9a584971f",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "# ! pip install pymilvus==2.3.4\n",
+    "# ! pip install pyarrow==12.0.0\n",
+    "# !pip install -U sentence-transformers"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "f1d8f101-f51b-4a50-b150-86e87c50c453",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "import numpy as np\n",
+    "import tensorflow as tf\n",
+    "from tqdm import tqdm\n",
+    "from dotenv import load_dotenv\n",
+    "import os\n",
+    "import pandas as pd\n",
+    "from pymilvus import connections, utility\n",
+    "from pymilvus import Collection, DataType, FieldSchema, CollectionSchema\n",
+    "import multiprocessing\n",
+    "from sentence_transformers import SentenceTransformer"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "4ad4e3ac-9685-4f12-8043-5fbcc373d3e1",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]"
+      ]
+     },
+     "execution_count": 4,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "tf.config.list_physical_devices('GPU')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "da71d832-b8a7-452b-b736-538a3c069b54",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>index</th>\n",
+       "      <th>category</th>\n",
+       "      <th>short_description</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>0</th>\n",
+       "      <td>0</td>\n",
+       "      <td>SCIENCE</td>\n",
+       "      <td>A closer look at water-splitting's solar fuel ...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1</th>\n",
+       "      <td>1</td>\n",
+       "      <td>SCIENCE</td>\n",
+       "      <td>An irresistible scent makes locusts swarm, stu...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2</th>\n",
+       "      <td>2</td>\n",
+       "      <td>SCIENCE</td>\n",
+       "      <td>Artificial intelligence warning: AI will know ...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>3</th>\n",
+       "      <td>3</td>\n",
+       "      <td>SCIENCE</td>\n",
+       "      <td>Glaciers Could Have Sculpted Mars Valleys: Study</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4</th>\n",
+       "      <td>4</td>\n",
+       "      <td>SCIENCE</td>\n",
+       "      <td>Perseid meteor shower 2020: What time and how ...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>...</th>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "      <td>...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>311171</th>\n",
+       "      <td>311171</td>\n",
+       "      <td>TECH</td>\n",
+       "      <td>RIM CEO Thorsten Heins' 'Significant' Plans Fo...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>311172</th>\n",
+       "      <td>311172</td>\n",
+       "      <td>SPORTS</td>\n",
+       "      <td>Maria Sharapova Stunned By Victoria Azarenka I...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>311173</th>\n",
+       "      <td>311173</td>\n",
+       "      <td>SPORTS</td>\n",
+       "      <td>Giants Over Patriots, Jets Over Colts Among  M...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>311174</th>\n",
+       "      <td>311174</td>\n",
+       "      <td>SPORTS</td>\n",
+       "      <td>Aldon Smith Arrested: 49ers Linebacker Busted ...</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>311175</th>\n",
+       "      <td>311175</td>\n",
+       "      <td>SPORTS</td>\n",
+       "      <td>Dwight Howard Rips Teammates After Magic Loss ...</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "<p>311176 rows × 3 columns</p>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "         index category                                  short_description\n",
+       "0            0  SCIENCE  A closer look at water-splitting's solar fuel ...\n",
+       "1            1  SCIENCE  An irresistible scent makes locusts swarm, stu...\n",
+       "2            2  SCIENCE  Artificial intelligence warning: AI will know ...\n",
+       "3            3  SCIENCE   Glaciers Could Have Sculpted Mars Valleys: Study\n",
+       "4            4  SCIENCE  Perseid meteor shower 2020: What time and how ...\n",
+       "...        ...      ...                                                ...\n",
+       "311171  311171     TECH  RIM CEO Thorsten Heins' 'Significant' Plans Fo...\n",
+       "311172  311172   SPORTS  Maria Sharapova Stunned By Victoria Azarenka I...\n",
+       "311173  311173   SPORTS  Giants Over Patriots, Jets Over Colts Among  M...\n",
+       "311174  311174   SPORTS  Aldon Smith Arrested: 49ers Linebacker Busted ...\n",
+       "311175  311175   SPORTS  Dwight Howard Rips Teammates After Magic Loss ...\n",
+       "\n",
+       "[311176 rows x 3 columns]"
+      ]
+     },
+     "execution_count": 5,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "data = pd.read_csv('labelled_newscatcher_dataset.csv', sep=\";\", usecols=['title', 'topic'])\n",
+    "json_data=pd.read_json('News_Category_Dataset_v3.json', lines=True)\n",
+    "data.drop_duplicates(subset=['title'], inplace=True)\n",
+    "json_data.drop_duplicates(subset=['headline'], inplace=True)\n",
+    "json_data = json_data[['headline', 'category']].copy()\n",
+    "json_data.rename(columns={'headline': 'title'}, inplace=True)\n",
+    "data.rename(columns={'topic': 'category'}, inplace=True)\n",
+    "data = pd.concat([data, json_data], axis=0)\n",
+    "data.drop_duplicates(subset=['title'], inplace=True)\n",
+    "data.reset_index(drop=True, inplace=True)\n",
+    "data.reset_index(inplace=True)\n",
+    "data.rename(columns={'title': 'short_description'}, inplace=True)\n",
+    "data"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "796f85b1-12dc-42cb-b431-65c88738b607",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "False"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "any(data['short_description'].duplicated())"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "5f46251b-156a-4a72-ab89-6abb6d810006",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "data.to_csv('news_processed.csv', index=False)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "id": "1463ea34-4447-464a-b0a3-da5e09892a09",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "class TextVectorizer:\n",
+    "    '''\n",
+    "    sentence transformers to extract sentence embeddings\n",
+    "    '''\n",
+    "    def vectorize(self, x):\n",
+    "        sent_model = SentenceTransformer('multi-qa-distilbert-cos-v1_finetuned')\n",
+    "        sen_embeddings = sent_model.encode(x)\n",
+    "        return sen_embeddings"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "47a714f3-8948-470b-9caf-93ed2bbf4894",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "vectorizer = TextVectorizer()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "id": "8b1586e5-2923-4632-a3db-fd2364124d6f",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "320"
+      ]
+     },
+     "execution_count": 10,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# getting max length of article descriptions to be used for VARCHAR while defining schema\n",
+    "max_desc_len = max([len(s) for s in data['short_description']])\n",
+    "max_desc_len"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "debe0ef4-b877-495a-872e-47f720b758a9",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "14"
+      ]
+     },
+     "execution_count": 11,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "# getting max length of article categories to be used for VARCHAR while defining schema\n",
+    "max_cat_len = max([len(s) for s in data['category']])\n",
+    "max_cat_len"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "id": "80489f00-e59f-46ab-a933-97145928176c",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "# # Reading milvus URI & API token from secrets.env\n",
+    "load_dotenv('secrets.env')\n",
+    "uri = os.environ.get(\"URI\")\n",
+    "token = os.environ.get(\"TOKEN\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "id": "0bf69f22-e113-43a5-be81-77224cafd856",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Connected to DB\n"
+     ]
+    }
+   ],
+   "source": [
+    "connections.connect(\"default\", uri=uri, token=token)\n",
+    "print(f\"Connected to DB\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "id": "8da06a3b-2005-4c02-a168-dc84bcde7064",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [],
+   "source": [
+    "collection_name = 'news_collection_sent_tran_finetuned'\n",
+    "check_collection = utility.has_collection(collection_name)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "id": "33342612-1380-4d1a-a8e7-931476e07979",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Droped Existing collection\n"
+     ]
+    }
+   ],
+   "source": [
+    "if check_collection:\n",
+    "    drop_result = utility.drop_collection(collection_name)\n",
+    "    print(\"Droped Existing collection\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "id": "fc8ae048-d586-41e7-9678-75e1752c1693",
+   "metadata": {
+    "tags": []
+   },
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Creating the collection\n",
+      "Schema: {'auto_id': False, 'description': 'collection of news articles', 'fields': [{'name': 'article_id', 'description': 'primary id', 'type': <DataType.INT64: 5>, 'is_primary': True, 'auto_id': False}, {'name': 'article_embed', 'description': '', 'type': <DataType.FLOAT_VECTOR: 101>, 'params': {'dim': 768}}, {'name': 'article_desc', 'description': 'short description of the article', 'type': <DataType.VARCHAR: 21>, 'params': {'max_length': 370}}, {'name': 'article_category', 'description': 'category of the article', 'type': <DataType.VARCHAR: 21>, 'params': {'max_length': 64}}]}\n",
+      "Success!\n"
+     ]
+    }
+   ],
+   "source": [
+    "# Creating collection schema\n",
+    "dim = 768 # embeddings dim\n",
+    "article_id = FieldSchema(name=\"article_id\", dtype=DataType.INT64, is_primary=True, description=\"primary id\") # primary key\n",
+    "article_embed_field = FieldSchema(name=\"article_embed\", dtype=DataType.FLOAT_VECTOR, dim=dim) # description embeddings\n",
+    "article_desc = FieldSchema(name=\"article_desc\", dtype=DataType.VARCHAR, max_length=(max_desc_len + 50), # using max_desc_len to specify VARCHAR len \n",
+    "                           is_primary=False, description=\"short description of the article\") # short description of article\n",
+    "article_cat = FieldSchema(name=\"article_category\", dtype=DataType.VARCHAR, max_length=(max_cat_len + 50), # using max_desc_len to specify VARCHAR len \n",
+    "                           is_primary=False, description=\"category of the article\") # category of article\n",
+    "schema = CollectionSchema(fields=[article_id, article_embed_field, article_desc, article_cat], \n",
+    "                          auto_id=False, description=\"collection of news articles\")\n",
+    "print(f\"Creating the collection\")\n",
+    "collection = Collection(name=collection_name, schema=schema)\n",
+    "print(f\"Schema: {schema}\")\n",
+    "print(\"Success!\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "id": "cca82380-98f6-4c44-aac6-86d4ae3484d0",
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "[0, 1000, 2000, 3000, 4000, 5000, 6000, 7000, 8000, 9000, 10000, 11000, 12000, 13000, 14000, 15000, 16000, 17000, 18000, 19000, 20000, 21000, 22000, 23000, 24000, 25000, 26000, 27000, 28000, 29000, 30000, 31000, 32000, 33000, 34000, 35000, 36000, 37000, 38000, 39000, 40000, 41000, 42000, 43000, 44000, 45000, 46000, 47000, 48000, 49000, 50000, 51000, 52000, 53000, 54000, 55000, 56000, 57000, 58000, 59000, 60000, 61000, 62000, 63000, 64000, 65000, 66000, 67000, 68000, 69000, 70000, 71000, 72000, 73000, 74000, 75000, 76000, 77000, 78000, 79000, 80000, 81000, 82000, 83000, 84000, 85000, 86000, 87000, 88000, 89000, 90000, 91000, 92000, 93000, 94000, 95000, 96000, 97000, 98000, 99000, 100000, 101000, 102000, 103000, 104000, 105000, 106000, 107000, 108000, 109000, 110000, 111000, 112000, 113000, 114000, 115000, 116000, 117000, 118000, 119000, 120000, 121000, 122000, 123000, 124000, 125000, 126000, 127000, 128000, 129000, 130000, 131000, 132000, 133000, 134000, 135000, 136000, 137000, 138000, 139000, 140000, 141000, 142000, 143000, 144000, 145000, 146000, 147000, 148000, 149000, 150000, 151000, 152000, 153000, 154000, 155000, 156000, 157000, 158000, 159000, 160000, 161000, 162000, 163000, 164000, 165000, 166000, 167000, 168000, 169000, 170000, 171000, 172000, 173000, 174000, 175000, 176000, 177000, 178000, 179000, 180000, 181000, 182000, 183000, 184000, 185000, 186000, 187000, 188000, 189000, 190000, 191000, 192000, 193000, 194000, 195000, 196000, 197000, 198000, 199000, 200000, 201000, 202000, 203000, 204000, 205000, 206000, 207000, 208000, 209000, 210000, 211000, 212000, 213000, 214000, 215000, 216000, 217000, 218000, 219000, 220000, 221000, 222000, 223000, 224000, 225000, 226000, 227000, 228000, 229000, 230000, 231000, 232000, 233000, 234000, 235000, 236000, 237000, 238000, 239000, 240000, 241000, 242000, 243000, 244000, 245000, 246000, 247000, 248000, 249000, 250000, 251000, 252000, 253000, 254000, 255000, 256000, 257000, 258000, 259000, 260000, 261000, 262000, 263000, 264000, 265000, 266000, 267000, 268000, 269000, 270000, 271000, 272000, 273000, 274000, 275000, 276000, 277000, 278000, 279000, 280000, 281000, 282000, 283000, 284000, 285000, 286000, 287000, 288000, 289000, 290000, 291000, 292000, 293000, 294000, 295000, 296000, 297000, 298000, 299000, 300000, 301000, 302000, 303000, 304000, 305000, 306000, 307000, 308000, 309000, 310000, 311000, 311176]\n"
+     ]
+    }
+   ],
+   "source": [
+    "cuts = [*range(0, len(data), 1000)]\n",
+    "cuts.append(len(data))\n",
+    "print(cuts)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "066c67ac-01a6-4151-8e85-5869ddce1c0a",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "article_id = []\n",
+    "article_desc = []\n",
+    "article_embed = []\n",
+    "article_cat = []\n",
+    "try:\n",
+    "    for i in tqdm(range(len(cuts)-1)):\n",
+    "        df = data.iloc[cuts[i]: cuts[i+1]].copy()\n",
+    "        article_id = [*df['index']]\n",
+    "        article_desc = [*df['short_description']]\n",
+    "        article_cat = [*df['category']]\n",
+    "        results = []\n",
+    "        article_embed = vectorizer.vectorize(article_desc)\n",
+    "        docs = [article_id, article_embed, article_desc, article_cat]\n",
+    "        ins_resp = collection.insert(docs)\n",
+    "        print(ins_resp)\n",
+    "        article_id = []\n",
+    "        article_desc = []\n",
+    "        article_embed = []\n",
+    "        article_cat = []\n",
+    "        if i == 0:\n",
+    "            index_params = {\"index_type\": \"AUTOINDEX\", \"metric_type\": \"L2\", \"params\": {}} \n",
+    "            collection.create_index(field_name='article_embed', index_params=index_params)\n",
+    "            collection = Collection(name=collection_name)\n",
+    "            collection.load()\n",
+    "except:\n",
+    "    raise"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "d50177fa-fd0c-48ad-bc9b-a7bdc826a628",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python (tf_gpu)",
+   "language": "python",
+   "name": "tf_gpu"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.9.18"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

tsdae_finetune_sent_transformer.ipynb

@@ -0,0 +1,245 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "f0dc396f",
+   "metadata": {},
+   "source": [
+    "### TSDAE: Fine-tune sentence transformers using unsupervised learning with Pytorch\n",
+    "https://www.sbert.net/examples/unsupervised_learning/TSDAE/README.html"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "id": "34329058",
+   "metadata": {
+    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
+    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5",
+    "execution": {
+     "iopub.execute_input": "2024-01-13T16:51:37.887254Z",
+     "iopub.status.busy": "2024-01-13T16:51:37.886390Z",
+     "iopub.status.idle": "2024-01-13T16:51:37.891827Z",
+     "shell.execute_reply": "2024-01-13T16:51:37.890706Z",
+     "shell.execute_reply.started": "2024-01-13T16:51:37.887212Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "# !pip install sentence_transformers==2.2.2"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "id": "ebd138d8",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2024-01-13T16:51:50.310221Z",
+     "iopub.status.busy": "2024-01-13T16:51:50.309586Z",
+     "iopub.status.idle": "2024-01-13T16:51:50.315850Z",
+     "shell.execute_reply": "2024-01-13T16:51:50.314927Z",
+     "shell.execute_reply.started": "2024-01-13T16:51:50.310185Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "import numpy as np\n",
+    "import string\n",
+    "from tqdm import tqdm\n",
+    "from numpy.linalg import norm\n",
+    "from sentence_transformers import SentenceTransformer, LoggingHandler\n",
+    "from sentence_transformers import models, util, datasets, evaluation, losses\n",
+    "from torch.utils.data import DataLoader"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "7ef2d063",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2024-01-13T13:24:44.770211Z",
+     "iopub.status.busy": "2024-01-13T13:24:44.769806Z",
+     "iopub.status.idle": "2024-01-13T13:24:44.775042Z",
+     "shell.execute_reply": "2024-01-13T13:24:44.773860Z",
+     "shell.execute_reply.started": "2024-01-13T13:24:44.770177Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "# import nltk\n",
+    "# nltk.download('punkt')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "id": "453c1add",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2024-01-13T16:51:54.070689Z",
+     "iopub.status.busy": "2024-01-13T16:51:54.069945Z",
+     "iopub.status.idle": "2024-01-13T16:51:54.809726Z",
+     "shell.execute_reply": "2024-01-13T16:51:54.808920Z",
+     "shell.execute_reply.started": "2024-01-13T16:51:54.070657Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "data = pd.read_csv('news_processed.csv', usecols=['short_description'])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "id": "61629b79",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2024-01-13T16:51:55.125758Z",
+     "iopub.status.busy": "2024-01-13T16:51:55.124990Z",
+     "iopub.status.idle": "2024-01-13T16:51:55.180470Z",
+     "shell.execute_reply": "2024-01-13T16:51:55.179559Z",
+     "shell.execute_reply.started": "2024-01-13T16:51:55.125716Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "data.dropna(inplace=True)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "id": "14162e2c",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2024-01-13T16:51:55.817535Z",
+     "iopub.status.busy": "2024-01-13T16:51:55.816764Z",
+     "iopub.status.idle": "2024-01-13T16:51:55.836578Z",
+     "shell.execute_reply": "2024-01-13T16:51:55.835697Z",
+     "shell.execute_reply.started": "2024-01-13T16:51:55.817499Z"
+    }
+   },
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "'Experimental coronavirus vaccine prevents severe disease in mice'"
+      ]
+     },
+     "execution_count": 7,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "data['short_description'][1000]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "id": "8fd1a801",
+   "metadata": {
+    "execution": {
+     "iopub.execute_input": "2024-01-13T17:08:09.906826Z",
+     "iopub.status.busy": "2024-01-13T17:08:09.906182Z",
+     "iopub.status.idle": "2024-01-13T17:08:09.914834Z",
+     "shell.execute_reply": "2024-01-13T17:08:09.913737Z",
+     "shell.execute_reply.started": "2024-01-13T17:08:09.906795Z"
+    }
+   },
+   "outputs": [],
+   "source": [
+    "def finetune_model(data: pd.DataFrame, col_to_use: str='short_description', \n",
+    "                   model_id: str=\"sentence-transformers/multi-qa-distilbert-cos-v1\", \n",
+    "                   batch_size: int=8, epochs: int=2):\n",
+    "    \n",
+    "#     https://www.sbert.net/examples/unsupervised_learning/TSDAE/README.html\n",
+    "    \n",
+    "    word_embedding_model = models.Transformer(model_id)\n",
+    "    pooling_model = models.Pooling(word_embedding_model.get_word_embedding_dimension(), 'cls')\n",
+    "    model = SentenceTransformer(modules=[word_embedding_model, pooling_model])\n",
+    "    \n",
+    "    train_examples = data[col_to_use].tolist()\n",
+    "    train_dataset = datasets.DenoisingAutoEncoderDataset(train_examples)\n",
+    "    train_dataloader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True)\n",
+    "    train_loss = losses.DenoisingAutoEncoderLoss(model, decoder_name_or_path=\"sentence-transformers/paraphrase-distilroberta-base-v2\", tie_encoder_decoder=False)\n",
+    "#     train_loss = losses.DenoisingAutoEncoderLoss(model, decoder_name_or_path=model_id, tie_encoder_decoder=True)\n",
+    "\n",
+    "    \n",
+    "    model.fit(\n",
+    "    train_objectives=[(train_dataloader, train_loss)],\n",
+    "    epochs=epochs,\n",
+    "    weight_decay=0,\n",
+    "    scheduler='constantlr',\n",
+    "    optimizer_params={'lr': 3e-5},\n",
+    "    show_progress_bar=True\n",
+    "    )\n",
+    "    model_save_path = model_id + '_finetuned'\n",
+    "    model.save(model_save_path)\n",
+    "    return model_save_path"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "b17a9779",
+   "metadata": {
+    "scrolled": true
+   },
+   "outputs": [],
+   "source": [
+    "# fine-tune sentence transformer\n",
+    "finetuned_model_id = finetune_model(data=data)\n",
+    "finetuned_model = SentenceTransformer(finetuned_model_id)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "4d87c128",
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kaggle": {
+   "accelerator": "gpu",
+   "dataSources": [
+    {
+     "datasetId": 4298708,
+     "sourceId": 7394110,
+     "sourceType": "datasetVersion"
+    }
+   ],
+   "dockerImageVersionId": 30636,
+   "isGpuEnabled": true,
+   "isInternetEnabled": true,
+   "language": "python",
+   "sourceType": "notebook"
+  },
+  "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": "3.10.12"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}