promptllama.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 92,
   "id": "69b6a4eb-dd49-438c-982f-88cc287b5a4f",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import os\n",
    "#import gradio as gr\n",
    "#import fire\n",
    "from enum import Enum\n",
    "from threading import Thread\n",
    "from transformers import AutoModelForCausalLM, AutoTokenizer\n",
    "from auto_gptq import AutoGPTQForCausalLM\n",
    "from llama_cpp import Llama\n",
    "from huggingface_hub import hf_hub_download\n",
    "from transformers import TextIteratorStreamer\n",
    "#from llama_chat_format import format_to_llama_chat_style\n",
    "from ctransformers import AutoModelForCausalLM\n",
    "import json\n",
    "import pandas as pd\n",
    "import os\n",
    "import re\n",
    "from transformers import pipeline, AutoModelForCausalLM, AutoTokenizer\n",
    "import numpy as np\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "id": "ef00ad84-dd71-44bf-808e-47508a1c1c4f",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6b3a7e215db94592a24351777218cc20",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Fetching 1 files:   0%|          | 0/1 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f9b31e794ec24ba68031594eb9a97d2f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Fetching 1 files:   0%|          | 0/1 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "ggml_cuda_set_main_device: using device 0 (NVIDIA A100 80GB PCIe) as main device\n"
     ]
    }
   ],
   "source": [
    "# For demonstraion purpose we use smaller version of LLama\n",
    "# but it can be easily changed with any version depending on the available resources\n",
    "llm = AutoModelForCausalLM.from_pretrained(\"TheBloke/sheep-duck-llama-2-13B-GGUF\", model_file=\"sheep-duck-llama-2-13b.Q4_K_M.gguf\", model_type=\"llama\", gpu_layers=50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "e6f92381-e7b8-44a0-a22c-5c26dd9f24f3",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def fewRel(path):\n",
    "    data_list = []  # Create a list to collect dictionaries for each row\n",
    "    sentences = []\n",
    "    relations = []\n",
    "    entities = []\n",
    "    data = {}\n",
    "    for filename in os.listdir(path):\n",
    "        if not filename.endswith('.json'):\n",
    "            continue\n",
    "        fullname = os.path.join(path, filename)\n",
    "        \n",
    "        with open(fullname) as f:\n",
    "            fewRel_corpus = json.load(f)\n",
    "            \n",
    "            for relation, value in fewRel_corpus.items():\n",
    "                for row in value:\n",
    "                    sentences.append(row['tokens'])\n",
    "                    relations.append(relation)\n",
    "                    entities.append((row['h'][0].strip(),row['t'][0].strip()))\n",
    "        data['sentence'] = sentences\n",
    "        data['relation'] = relations\n",
    "        data['entities'] = entities\n",
    "    \n",
    "    return data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "6cadba59-5c4d-4ae7-912f-ca93f0f31349",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "data = fewRel('data/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "b5f90f89-fa4b-435e-92b6-fe1484a17332",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "df = pd.DataFrame(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "07d45c67-969c-4f2d-a28d-53c3ee0ac66a",
   "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>sentence</th>\n",
       "      <th>relation</th>\n",
       "      <th>entities</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[Merpati, flight, 106, departed, Jakarta, (, C...</td>\n",
       "      <td>P931</td>\n",
       "      <td>(tjq, tanjung pandan)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                            sentence relation  \\\n",
       "0  [Merpati, flight, 106, departed, Jakarta, (, C...     P931   \n",
       "\n",
       "                entities  \n",
       "0  (tjq, tanjung pandan)  "
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "0be6ef4e-2682-4dcd-9ece-cb70415e398f",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def random_rows_per_category(group):\n",
    "    return group.sample(n=1, random_state=42) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "id": "e9c68c42-90fe-4843-a1ac-5f29f33a781b",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "random_rows = df.groupby('relation', group_keys=False).apply(random_rows_per_category)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "id": "20c9d867-377d-46df-8893-9d8b7dcbd724",
   "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>sentence</th>\n",
       "      <th>relation</th>\n",
       "      <th>entities</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>42858</th>\n",
       "      <td>[The, ICAC, has, jurisdiction, over, state, an...</td>\n",
       "      <td>P1001</td>\n",
       "      <td>(icac, new south wales)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41458</th>\n",
       "      <td>[\", See, also, :, \", Peter, Artedi, (, 1705–17...</td>\n",
       "      <td>P101</td>\n",
       "      <td>(peter artedi, ichthyology)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21158</th>\n",
       "      <td>[Bhupinder, Singh, Hooda, (, born, 15, Septemb...</td>\n",
       "      <td>P102</td>\n",
       "      <td>(bhupinder singh hooda, indian national congress)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40058</th>\n",
       "      <td>[It, belongs, to, the, hobby, group, ,, a, rat...</td>\n",
       "      <td>P105</td>\n",
       "      <td>(hobby, subgenus)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25358</th>\n",
       "      <td>[Jaume, Serra, Serra, (, born, June, 11, ,, 19...</td>\n",
       "      <td>P106</td>\n",
       "      <td>(jaume serra serra, politician)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35858</th>\n",
       "      <td>[Adila, Fachiri, made, a, recording, of, the, ...</td>\n",
       "      <td>P86</td>\n",
       "      <td>(10th violin sonata, beethoven)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>158</th>\n",
       "      <td>[Hill, had, left, his, luggage, at, Lee, Bird,...</td>\n",
       "      <td>P931</td>\n",
       "      <td>(lee bird field, north platte, nebraska)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26758</th>\n",
       "      <td>[He, was, born, in, Christiania, as, a, son, o...</td>\n",
       "      <td>P937</td>\n",
       "      <td>(carl johan michelet, christiania)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33758</th>\n",
       "      <td>[In, Over, the, river, Seeve, flows, into, the...</td>\n",
       "      <td>P974</td>\n",
       "      <td>(elbe, seeve)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17658</th>\n",
       "      <td>[The, Frank, Walsh, -, led, Australian, Labor,...</td>\n",
       "      <td>P991</td>\n",
       "      <td>(1965 election, frank walsh)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>64 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                sentence relation  \\\n",
       "42858  [The, ICAC, has, jurisdiction, over, state, an...    P1001   \n",
       "41458  [\", See, also, :, \", Peter, Artedi, (, 1705–17...     P101   \n",
       "21158  [Bhupinder, Singh, Hooda, (, born, 15, Septemb...     P102   \n",
       "40058  [It, belongs, to, the, hobby, group, ,, a, rat...     P105   \n",
       "25358  [Jaume, Serra, Serra, (, born, June, 11, ,, 19...     P106   \n",
       "...                                                  ...      ...   \n",
       "35858  [Adila, Fachiri, made, a, recording, of, the, ...      P86   \n",
       "158    [Hill, had, left, his, luggage, at, Lee, Bird,...     P931   \n",
       "26758  [He, was, born, in, Christiania, as, a, son, o...     P937   \n",
       "33758  [In, Over, the, river, Seeve, flows, into, the...     P974   \n",
       "17658  [The, Frank, Walsh, -, led, Australian, Labor,...     P991   \n",
       "\n",
       "                                                entities  \n",
       "42858                            (icac, new south wales)  \n",
       "41458                        (peter artedi, ichthyology)  \n",
       "21158  (bhupinder singh hooda, indian national congress)  \n",
       "40058                                  (hobby, subgenus)  \n",
       "25358                    (jaume serra serra, politician)  \n",
       "...                                                  ...  \n",
       "35858                    (10th violin sonata, beethoven)  \n",
       "158             (lee bird field, north platte, nebraska)  \n",
       "26758                 (carl johan michelet, christiania)  \n",
       "33758                                      (elbe, seeve)  \n",
       "17658                       (1965 election, frank walsh)  \n",
       "\n",
       "[64 rows x 3 columns]"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "random_rows"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "06244330-df7b-4e3c-be5c-e1f0d5b925c1",
   "metadata": {},
   "source": [
    "### Entities remain the same"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "29208aad-da8f-4764-be6c-51adcace314a",
   "metadata": {},
   "outputs": [],
   "source": [
    "example = \"'The FBI has jurisdiction over federal and local government in Texas.' is augumented to 'he CCB has jurisdiction over regional and local government in Brazil.'\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "id": "915c523e-d153-48f4-932f-5e89ede3d464",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "question = 'Please augment 3 syntactically and semantically correct sentences, each distinct from the preceding one,by introducing separate named entities:' + example + ' text ='"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "id": "66f3b305-4c0e-4ee1-86e0-c3fe43fb2ea4",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "all_augumented_data = []\n",
    "for i in range(10):\n",
    "    random_rows = df.groupby('relation', group_keys=False).apply(random_rows_per_category)\n",
    "    for index, row in random_rows.iterrows():\n",
    "        single_aug_data = {}\n",
    "        sent = ' '.join(row['sentence'])\n",
    "        augumented_text = process_data(llm(question + sent))\n",
    "        #print(process_data(augumented_text))\n",
    "        single_aug_data['sentences'] = augumented_text\n",
    "        single_aug_data['relation'] = row['relation']\n",
    "        single_aug_data['entities'] = row['entities']\n",
    "       # output = augumented_text +\", \" + row['relation'] + \", \" +  row['entities'][0] +\", \" +  row['entities'][1]\n",
    "        all_augumented_data.append(single_aug_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "id": "f7483f78-9eb4-47ca-9681-a91a49568f55",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "640"
      ]
     },
     "execution_count": 102,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "id": "e6bcb955-b03a-4b4d-982e-8f3d3cc9e2ae",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def process_data(text):\n",
    "    sentences = re.split(r'\\d+\\.\\s+', text)\n",
    "    sentences = [sentence for sentence in sentences if sentence.strip()]    \n",
    "    return sentences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "id": "4ae85931-7edc-4bcb-b7fe-d6c9fb596584",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "id": "1525a4d4-1cc1-4594-af96-bfa5971f5bb3",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def save(file_path,data):\n",
    "    with open(file_path, \"w\") as json_file:\n",
    "        json.dump(data, json_file)\n",
    "\n",
    "    print(f'Data saved to {file_path}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "id": "4738292e-832b-4e15-8115-e89adeec39b1",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "def nytFb(path):\n",
    "    dist = \"train\"\n",
    "    \n",
    "    temp_sent = \"\"\n",
    "    \n",
    "    temp_rel = \"\"\n",
    "    data = {}\n",
    "    data_list = []  # Create a list to collect dictionaries for each row\n",
    "    sentences = []\n",
    "    relations = []\n",
    "    entities = []\n",
    "   \n",
    "    for filename in os.listdir(path):\n",
    "        if not filename.endswith('.json'): continue\n",
    "        fullname = os.path.join(path, filename)\n",
    "       \n",
    "        if filename == \"test.json\":\n",
    "            dist = 'test'\n",
    "        elif filename == \"train.json\":\n",
    "            dist = 'train'\n",
    "        else:\n",
    "            dist = 'valid'\n",
    "        data2 = pd.read_json(r''+ fullname, lines=True)\n",
    "        print(data2.head())\n",
    "        return data2\n",
    "       \n",
    "    return data\n",
    "   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "id": "bd5f0bc9-4edd-4e4c-bd02-fb76ffa89c26",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                                            sentText  \\\n",
      "0  Massachusetts ASTON MAGNA Great Barrington ; a...   \n",
      "1  North Carolina EASTERN MUSIC FESTIVAL Greensbo...   \n",
      "2  It will be the final movie credited to Debra H...   \n",
      "3  In a 3-0 victory over the Crew on Saturday in ...   \n",
      "4  The United States ambassador to Mexico , Tony ...   \n",
      "\n",
      "                                           articleId  \\\n",
      "0  /m/vinci8/data1/riedel/projects/relation/kb/ny...   \n",
      "1  /m/vinci8/data1/riedel/projects/relation/kb/ny...   \n",
      "2  /m/vinci8/data1/riedel/projects/relation/kb/ny...   \n",
      "3  /m/vinci8/data1/riedel/projects/relation/kb/ny...   \n",
      "4  /m/vinci8/data1/riedel/projects/relation/kb/ny...   \n",
      "\n",
      "                                    relationMentions  \\\n",
      "0  [{'em1Text': 'Annandale-on-Hudson', 'em2Text':...   \n",
      "1  [{'em1Text': 'North Carolina', 'em2Text': 'Gre...   \n",
      "2  [{'em1Text': 'Debra Hill', 'em2Text': 'Haddonf...   \n",
      "3  [{'em1Text': 'Ohio', 'em2Text': 'Columbus', 'l...   \n",
      "4  [{'em1Text': 'Mexico', 'em2Text': 'Nuevo Lared...   \n",
      "\n",
      "                                      entityMentions  sentId  \n",
      "0  [{'start': 1, 'label': 'ORGANIZATION', 'text':...       1  \n",
      "1  [{'start': 0, 'label': 'LOCATION', 'text': 'No...       2  \n",
      "2  [{'start': 0, 'label': 'LOCATION', 'text': 'De...       1  \n",
      "3  [{'start': 0, 'label': 'LOCATION', 'text': 'Co...       1  \n",
      "4  [{'start': 0, 'label': 'LOCATION', 'text': 'Un...       1  \n"
     ]
    }
   ],
   "source": [
    "data_nyt = nytFb('data/nyt/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "id": "33c8301e-7a48-4d05-bae1-f4ad9a8df2f1",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "nyt_df = pd.DataFrame(columns=['sentences', 'relation', 'entities'])\n",
    "\n",
    "# Loop through the original DataFrame and perform data manipulation\n",
    "for index, row in data_nyt.iterrows():\n",
    "    # Perform your data manipulation here\n",
    "    sentence = row['sentText']\n",
    "    relation = row['relationMentions'][0]['label'].split('/')[-1] \n",
    "    ent1 = row['entityMentions'][0]['text']\n",
    "    ent2 = row['entityMentions'][1]['text']\n",
    "    entities = (ent1, ent2)\n",
    "\n",
    "    # Create a temporary DataFrame for the current row\n",
    "    temp_df = pd.DataFrame({'sentences': [sentence], 'relation': [relation], 'entities': [entities]})\n",
    "\n",
    "    # Concatenate the temporary DataFrame with the new_df\n",
    "    nyt_df = pd.concat([nyt_df, temp_df], ignore_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "id": "26b3b499-f6a6-4ea5-b0bf-4dd2da45a870",
   "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>sentences</th>\n",
       "      <th>relation</th>\n",
       "      <th>entities</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Massachusetts ASTON MAGNA Great Barrington ; a...</td>\n",
       "      <td>contains</td>\n",
       "      <td>(Bard College, Annandale-on-Hudson)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>North Carolina EASTERN MUSIC FESTIVAL Greensbo...</td>\n",
       "      <td>contains</td>\n",
       "      <td>(North Carolina, Greensboro)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>It will be the final movie credited to Debra H...</td>\n",
       "      <td>place_of_birth</td>\n",
       "      <td>(Debra Hill, Haddonfield)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>In a 3-0 victory over the Crew on Saturday in ...</td>\n",
       "      <td>contains</td>\n",
       "      <td>(Columbus, Ohio)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>The United States ambassador to Mexico , Tony ...</td>\n",
       "      <td>contains</td>\n",
       "      <td>(United States, Mexico)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                           sentences        relation  \\\n",
       "0  Massachusetts ASTON MAGNA Great Barrington ; a...        contains   \n",
       "1  North Carolina EASTERN MUSIC FESTIVAL Greensbo...        contains   \n",
       "2  It will be the final movie credited to Debra H...  place_of_birth   \n",
       "3  In a 3-0 victory over the Crew on Saturday in ...        contains   \n",
       "4  The United States ambassador to Mexico , Tony ...        contains   \n",
       "\n",
       "                              entities  \n",
       "0  (Bard College, Annandale-on-Hudson)  \n",
       "1         (North Carolina, Greensboro)  \n",
       "2            (Debra Hill, Haddonfield)  \n",
       "3                     (Columbus, Ohio)  \n",
       "4              (United States, Mexico)  "
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nyt_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "id": "ea94ccc8-e74f-4102-9101-ca02152840aa",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from tqdm import tqdm\n",
    "def augument(df, num_iterations = 10):\n",
    "# Define the number of iterations\n",
    "    \n",
    "\n",
    "    # Create an empty list to store augmented data\n",
    "    all_augmented_data = []\n",
    "\n",
    "    # Iterate using tqdm for a progress bar\n",
    "    for i in tqdm(range(num_iterations), desc=\"Processing\"):\n",
    "        random_rows = df.groupby('relation', group_keys=False).apply(random_rows_per_category)\n",
    "\n",
    "        for index, row in tqdm(random_rows.iterrows()):\n",
    "            single_aug_data = {}\n",
    "            sent = row['sentences']\n",
    "            augumented_text = process_data(llm(question + sent))\n",
    "\n",
    "            single_aug_data['sentences'] = augumented_text\n",
    "            single_aug_data['relation'] = row['relation']\n",
    "            single_aug_data['entities'] = row['entities']\n",
    "\n",
    "            all_augmented_data.append(single_aug_data)\n",
    "    return all_augmented_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "id": "1f67675b-df2b-4409-bca6-bd666ffecc6e",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "240"
      ]
     },
     "execution_count": 119,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(all_augmented_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "id": "49c76bee-5346-4abb-9969-9c52bc68e9b8",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "relation\n",
      "contains                    30240\n",
      "nationality                  5219\n",
      "place_lived                  5024\n",
      "company                      3971\n",
      "capital                      2430\n",
      "neighborhood_of              2108\n",
      "place_of_birth               1580\n",
      "administrative_divisions     1518\n",
      "country                      1517\n",
      "place_of_death               1023\n",
      "children                      381\n",
      "founders                      377\n",
      "place_founded                 318\n",
      "major_shareholder_of           98\n",
      "teams                          88\n",
      "location                       86\n",
      "major_shareholders             84\n",
      "religion                       53\n",
      "advisors                       34\n",
      "geographic_distribution        26\n",
      "ethnicity                       9\n",
      "people                          9\n",
      "profession                      2\n",
      "industry                        1\n",
      "Name: count, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "category_counts = nyt_df['relation'].value_counts()\n",
    "\n",
    "print(category_counts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "id": "3eaa7d5c-a680-4432-868d-7caddf288f91",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.barh(category_counts.index, category_counts.values)\n",
    "plt.xscale('log')  # Use a logarithmic scale on the x-axis\n",
    "\n",
    "# Add labels and title\n",
    "plt.ylabel('Relations')\n",
    "plt.xlabel('Logarithmic Scale of Counts')\n",
    "#plt.title('Relation Distribution (Logarithmic Scale)')\n",
    "plt.savefig('nyt_relation_distribution.pdf', format='pdf', dpi=300, bbox_inches='tight')\n",
    "\n",
    "# Show the chart\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "id": "d5836db5-742d-4949-97cc-210f3483f7a8",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "threshold = 100\n",
    "\n",
    "# Filter categories with counts less than the threshold\n",
    "categories_below_threshold = category_counts[category_counts < threshold]\n",
    "\n",
    "# Create two separate DataFrames based on the filter\n",
    "df_below_threshold = nyt_df[nyt_df['relation'].isin(categories_below_threshold.index)]\n",
    "df_above_threshold = nyt_df[~nyt_df['relation'].isin(categories_below_threshold.index)]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "id": "cbc93494-6a51-4cb2-b78e-689eb4c7b053",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Processing:   0%|          | 0/20 [00:00<?, ?it/s]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:34,  8.38s/it]\u001b[A\n",
      "5it [00:42,  8.35s/it]\u001b[A\n",
      "6it [00:49,  7.79s/it]\u001b[A\n",
      "7it [00:56,  7.75s/it]\u001b[A\n",
      "8it [01:02,  7.26s/it]\u001b[A\n",
      "9it [01:10,  7.46s/it]\u001b[A\n",
      "10it [01:19,  7.84s/it]\u001b[A\n",
      "11it [01:27,  7.96s/it]\u001b[A\n",
      "Processing:   5%|▌         | 1/20 [01:27<27:43, 87.56s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:34,  8.38s/it]\u001b[A\n",
      "5it [00:42,  8.35s/it]\u001b[A\n",
      "6it [00:49,  7.89s/it]\u001b[A\n",
      "7it [00:54,  7.10s/it]\u001b[A\n",
      "8it [01:02,  7.40s/it]\u001b[A\n",
      "9it [01:10,  7.56s/it]\u001b[A\n",
      "10it [01:19,  7.91s/it]\u001b[A\n",
      "11it [01:20,  7.30s/it]\u001b[A\n",
      "Processing:  10%|█         | 2/20 [02:47<24:59, 83.30s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.52s/it]\u001b[A\n",
      "4it [00:33,  8.22s/it]\u001b[A\n",
      "5it [00:41,  8.25s/it]\u001b[A\n",
      "6it [00:49,  8.05s/it]\u001b[A\n",
      "7it [00:57,  7.92s/it]\u001b[A\n",
      "8it [01:04,  7.68s/it]\u001b[A\n",
      "9it [01:12,  7.76s/it]\u001b[A\n",
      "10it [01:21,  8.05s/it]\u001b[A\n",
      "11it [01:28,  8.05s/it]\u001b[A\n",
      "Processing:  15%|█▌        | 3/20 [04:16<24:17, 85.72s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.52s/it]\u001b[A\n",
      "4it [00:33,  8.13s/it]\u001b[A\n",
      "5it [00:41,  8.19s/it]\u001b[A\n",
      "6it [00:49,  8.01s/it]\u001b[A\n",
      "7it [00:55,  7.33s/it]\u001b[A\n",
      "8it [01:03,  7.55s/it]\u001b[A\n",
      "9it [01:11,  7.66s/it]\u001b[A\n",
      "10it [01:19,  7.98s/it]\u001b[A\n",
      "11it [01:27,  8.00s/it]\u001b[A\n",
      "Processing:  20%|██        | 4/20 [05:44<23:05, 86.61s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:34,  8.39s/it]\u001b[A\n",
      "5it [00:42,  8.35s/it]\u001b[A\n",
      "6it [00:49,  7.92s/it]\u001b[A\n",
      "7it [00:56,  7.73s/it]\u001b[A\n",
      "8it [01:04,  7.83s/it]\u001b[A\n",
      "9it [01:12,  7.85s/it]\u001b[A\n",
      "10it [01:16,  6.68s/it]\u001b[A\n",
      "11it [01:22,  7.49s/it]\u001b[A\n",
      "Processing:  25%|██▌       | 5/20 [07:06<21:16, 85.11s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:21,  6.46s/it]\u001b[A\n",
      "4it [00:28,  6.73s/it]\u001b[A\n",
      "5it [00:36,  7.30s/it]\u001b[A\n",
      "6it [00:44,  7.42s/it]\u001b[A\n",
      "7it [00:52,  7.50s/it]\u001b[A\n",
      "8it [01:00,  7.67s/it]\u001b[A\n",
      "9it [01:08,  7.74s/it]\u001b[A\n",
      "10it [01:16,  8.04s/it]\u001b[A\n",
      "11it [01:22,  7.52s/it]\u001b[A\n",
      "Processing:  30%|███       | 6/20 [08:29<19:40, 84.31s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:08,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.52s/it]\u001b[A\n",
      "4it [00:34,  8.38s/it]\u001b[A\n",
      "5it [00:42,  8.35s/it]\u001b[A\n",
      "6it [00:49,  7.95s/it]\u001b[A\n",
      "7it [00:49,  5.52s/it]\u001b[A\n",
      "8it [00:55,  5.67s/it]\u001b[A\n",
      "9it [01:03,  6.37s/it]\u001b[A\n",
      "10it [01:12,  7.09s/it]\u001b[A\n",
      "11it [01:19,  7.22s/it]\u001b[A\n",
      "Processing:  35%|███▌      | 7/20 [09:49<17:55, 82.72s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.83s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:32,  7.72s/it]\u001b[A\n",
      "5it [00:40,  7.93s/it]\u001b[A\n",
      "6it [00:48,  7.84s/it]\u001b[A\n",
      "7it [00:54,  7.44s/it]\u001b[A\n",
      "8it [01:01,  7.02s/it]\u001b[A\n",
      "9it [01:08,  7.30s/it]\u001b[A\n",
      "10it [01:17,  7.73s/it]\u001b[A\n",
      "11it [01:25,  7.79s/it]\u001b[A\n",
      "Processing:  40%|████      | 8/20 [11:14<16:44, 83.68s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.01s/it]\u001b[A\n",
      "2it [00:17,  8.83s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:33,  8.18s/it]\u001b[A\n",
      "5it [00:41,  8.22s/it]\u001b[A\n",
      "6it [00:49,  8.03s/it]\u001b[A\n",
      "7it [00:57,  7.91s/it]\u001b[A\n",
      "8it [01:05,  7.95s/it]\u001b[A\n",
      "9it [01:13,  7.94s/it]\u001b[A\n",
      "10it [01:21,  8.18s/it]\u001b[A\n",
      "11it [01:29,  8.17s/it]\u001b[A\n",
      "Processing:  45%|████▌     | 9/20 [12:44<15:41, 85.61s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.52s/it]\u001b[A\n",
      "4it [00:34,  8.38s/it]\u001b[A\n",
      "5it [00:42,  8.35s/it]\u001b[A\n",
      "6it [00:48,  7.50s/it]\u001b[A\n",
      "7it [00:55,  7.55s/it]\u001b[A\n",
      "8it [01:01,  7.11s/it]\u001b[A\n",
      "9it [01:09,  7.36s/it]\u001b[A\n",
      "10it [01:11,  5.55s/it]\u001b[A\n",
      "11it [01:18,  7.11s/it]\u001b[A\n",
      "Processing:  50%|█████     | 10/20 [14:02<13:53, 83.31s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:34,  8.38s/it]\u001b[A\n",
      "5it [00:42,  8.27s/it]\u001b[A\n",
      "6it [00:49,  8.06s/it]\u001b[A\n",
      "7it [00:54,  7.07s/it]\u001b[A\n",
      "8it [01:02,  7.38s/it]\u001b[A\n",
      "9it [01:10,  7.55s/it]\u001b[A\n",
      "10it [01:15,  6.64s/it]\u001b[A\n",
      "11it [01:21,  7.43s/it]\u001b[A\n",
      "Processing:  55%|█████▌    | 11/20 [15:24<12:25, 82.83s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.62s/it]\u001b[A\n",
      "3it [00:25,  8.42s/it]\u001b[A\n",
      "4it [00:32,  8.00s/it]\u001b[A\n",
      "5it [00:41,  8.11s/it]\u001b[A\n",
      "6it [00:48,  7.95s/it]\u001b[A\n",
      "7it [00:55,  7.66s/it]\u001b[A\n",
      "8it [00:57,  5.80s/it]\u001b[A\n",
      "9it [01:05,  6.46s/it]\u001b[A\n",
      "10it [01:14,  7.15s/it]\u001b[A\n",
      "11it [01:21,  7.44s/it]\u001b[A\n",
      "Processing:  60%|██████    | 12/20 [16:46<11:00, 82.54s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:08,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.52s/it]\u001b[A\n",
      "4it [00:33,  8.34s/it]\u001b[A\n",
      "5it [00:42,  8.33s/it]\u001b[A\n",
      "6it [00:49,  8.10s/it]\u001b[A\n",
      "7it [00:57,  7.95s/it]\u001b[A\n",
      "8it [01:05,  7.98s/it]\u001b[A\n",
      "9it [01:13,  7.96s/it]\u001b[A\n",
      "10it [01:19,  7.25s/it]\u001b[A\n",
      "11it [01:26,  7.83s/it]\u001b[A\n",
      "Processing:  65%|██████▌   | 13/20 [18:12<09:45, 83.63s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:08,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:34,  8.38s/it]\u001b[A\n",
      "5it [00:42,  8.35s/it]\u001b[A\n",
      "6it [00:49,  8.11s/it]\u001b[A\n",
      "7it [00:55,  7.18s/it]\u001b[A\n",
      "8it [01:03,  7.45s/it]\u001b[A\n",
      "9it [01:11,  7.60s/it]\u001b[A\n",
      "10it [01:19,  7.94s/it]\u001b[A\n",
      "11it [01:27,  7.99s/it]\u001b[A\n",
      "Processing:  70%|███████   | 14/20 [19:40<08:29, 84.93s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.81s/it]\u001b[A\n",
      "3it [00:20,  6.19s/it]\u001b[A\n",
      "4it [00:28,  6.97s/it]\u001b[A\n",
      "5it [00:37,  7.45s/it]\u001b[A\n",
      "6it [00:44,  7.52s/it]\u001b[A\n",
      "7it [00:50,  6.82s/it]\u001b[A\n",
      "8it [00:58,  7.21s/it]\u001b[A\n",
      "9it [01:06,  7.39s/it]\u001b[A\n",
      "10it [01:14,  7.80s/it]\u001b[A\n",
      "11it [01:21,  7.40s/it]\u001b[A\n",
      "Processing:  75%|███████▌  | 15/20 [21:01<06:59, 83.87s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:08,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:33,  8.16s/it]\u001b[A\n",
      "5it [00:41,  8.21s/it]\u001b[A\n",
      "6it [00:49,  8.02s/it]\u001b[A\n",
      "7it [00:57,  7.90s/it]\u001b[A\n",
      "8it [01:05,  7.94s/it]\u001b[A\n",
      "9it [01:12,  7.67s/it]\u001b[A\n",
      "10it [01:20,  7.99s/it]\u001b[A\n",
      "11it [01:28,  8.08s/it]\u001b[A\n",
      "Processing:  80%|████████  | 16/20 [22:30<05:41, 85.39s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:08,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.52s/it]\u001b[A\n",
      "4it [00:26,  5.55s/it]\u001b[A\n",
      "5it [00:33,  6.08s/it]\u001b[A\n",
      "6it [00:41,  6.61s/it]\u001b[A\n",
      "7it [00:46,  6.19s/it]\u001b[A\n",
      "8it [00:54,  6.78s/it]\u001b[A\n",
      "9it [01:02,  7.13s/it]\u001b[A\n",
      "10it [01:11,  7.62s/it]\u001b[A\n",
      "11it [01:18,  7.18s/it]\u001b[A\n",
      "Processing:  85%|████████▌ | 17/20 [23:49<04:10, 83.45s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:33,  7.99s/it]\u001b[A\n",
      "5it [00:41,  8.10s/it]\u001b[A\n",
      "6it [00:46,  7.26s/it]\u001b[A\n",
      "7it [00:54,  7.39s/it]\u001b[A\n",
      "8it [01:00,  7.03s/it]\u001b[A\n",
      "9it [01:08,  7.30s/it]\u001b[A\n",
      "10it [01:17,  7.73s/it]\u001b[A\n",
      "11it [01:25,  7.77s/it]\u001b[A\n",
      "Processing:  90%|█████████ | 18/20 [25:15<02:48, 84.04s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:08,  8.99s/it]\u001b[A\n",
      "2it [00:17,  8.82s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:34,  8.38s/it]\u001b[A\n",
      "5it [00:42,  8.35s/it]\u001b[A\n",
      "6it [00:48,  7.73s/it]\u001b[A\n",
      "7it [00:56,  7.71s/it]\u001b[A\n",
      "8it [01:04,  7.78s/it]\u001b[A\n",
      "9it [01:12,  7.82s/it]\u001b[A\n",
      "10it [01:21,  8.09s/it]\u001b[A\n",
      "11it [01:29,  8.10s/it]\u001b[A\n",
      "Processing:  95%|█████████▌| 19/20 [26:44<01:25, 85.56s/it]\n",
      "0it [00:00, ?it/s]\u001b[A\n",
      "1it [00:09,  9.00s/it]\u001b[A\n",
      "2it [00:17,  8.83s/it]\u001b[A\n",
      "3it [00:25,  8.53s/it]\u001b[A\n",
      "4it [00:34,  8.39s/it]\u001b[A\n",
      "5it [00:42,  8.35s/it]\u001b[A\n",
      "6it [00:50,  8.12s/it]\u001b[A\n",
      "7it [00:55,  7.16s/it]\u001b[A\n",
      "8it [00:58,  6.04s/it]\u001b[A\n",
      "9it [01:06,  6.62s/it]\u001b[A\n",
      "10it [01:15,  7.26s/it]\u001b[A\n",
      "11it [01:23,  7.57s/it]\u001b[A\n",
      "Processing: 100%|██████████| 20/20 [28:07<00:00, 84.38s/it]\n"
     ]
    }
   ],
   "source": [
    "augumented_data = augument(df_below_threshold,20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "id": "c2d45e74-7761-4894-803e-85bd1f85f681",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data saved to augumented_data/nyt_less100_100Plus_llama13_data_updated.json\n"
     ]
    }
   ],
   "source": [
    "file_path = \"augumented_data/nyt_less100_100Plus_llama13_data_updated.json\"\n",
    "save(file_path,augumented_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "id": "42a32e0d-bfd5-4567-8108-6dc0105cdb96",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "480"
      ]
     },
     "execution_count": 138,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(augumented_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "id": "8a6f4f55-1ef9-438a-aec7-bfa9ca55f536",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "random_rows = df_below_threshold.groupby('relation', group_keys=False).apply(random_rows_per_category)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fc0f12f3-0d3e-4984-8244-8607b4dcaeb9",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}