{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def score_report_bar():\n",
    "    path=r\"C:\\Users\\Inception\\Desktop\\LLM-based-QA-chatbot-builder\\UI\\score_report\"\n",
    "    import os\n",
    "    import math\n",
    "    dat=[]\n",
    "    for x in os.listdir(path):\n",
    "        wh=[]\n",
    "        flag=0\n",
    "        for x2 in x:\n",
    "            if x2>='a' and x2<='z':\n",
    "                flag=1\n",
    "                wh.append(x2)\n",
    "            elif flag==1:\n",
    "                wh.append(\" \")\n",
    "        wh=''.join(wh)\n",
    "        wh=wh.replace(\"model ans\",\"\")\n",
    "        wh=wh.replace(\"finetuned\",\"\")\n",
    "        wh=wh.replace(\"  \",\" \")\n",
    "        wh=wh.replace(\"xlsx\",\"\")\n",
    "        df_temp=pd.read_excel(os.path.join(path,x))\n",
    "        rating=sum(df_temp[\"rating\"])/len(df_temp)\n",
    "        dat.append({\n",
    "            \"Model Name\":wh,\n",
    "            \"Average Rating\":rating\n",
    "        })\n",
    "    temp=pd.DataFrame(dat)\n",
    "    return temp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import gradio as gr\n",
    "import pandas as pd\n",
    "\n",
    "def bar_plot_fn():\n",
    "    temp=score_report_bar()\n",
    "    return gr.BarPlot(\n",
    "            temp,\n",
    "            x=\"Model Name\",\n",
    "            y=\"Average Rating\",\n",
    "            x_title=\"Model name\",\n",
    "            y_title=\"Average Rating\",\n",
    "            title=\"Simple Bar Plot with made up data\",\n",
    "            tooltip=[\"Model Name\", \"Average Rating\"],\n",
    "            y_lim=[1, 5],\n",
    "            width=200,\n",
    "            height=1000\n",
    "        )\n",
    "with gr.Blocks() as bar_plot:\n",
    "    with gr.Row():\n",
    "        btn=gr.Button(\"test\")\n",
    "    with gr.Row():\n",
    "        plot = gr.BarPlot()\n",
    "    btn.click(bar_plot_fn, None, outputs=plot)\n",
    "\n",
    "bar_plot.launch()\t"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def parse_data(link,num=None):    \n",
    "    from bs4 import BeautifulSoup\n",
    "    import requests\n",
    "    import re\n",
    "    from docx import Document       \n",
    "    from langchain_community.document_loaders import WebBaseLoader\n",
    "    s=set()\n",
    "    import time\n",
    "    start_time = time.time()\n",
    "    duration = 5\n",
    "    def get_links(url):\n",
    "        response = requests.get(url)\n",
    "        data = response.text\n",
    "        soup = BeautifulSoup(data, 'lxml')\n",
    "\n",
    "        links = []\n",
    "        for link in soup.find_all('a'):\n",
    "            link_url = link.get('href')\n",
    "            if link_url is not None and link_url.startswith('http'):\n",
    "                s.add(link_url)\n",
    "                links.append(link_url)\n",
    "        \n",
    "        return links\n",
    "    # def write_to_file(links):\n",
    "    #     with open('data.txt', 'a') as f:\n",
    "    #         f.writelines(links)\n",
    "    def get_all_links(url):\n",
    "            for link in get_links(url):\n",
    "                if (time.time() - start_time) >= duration:\n",
    "                    return\n",
    "                get_all_links(link)\n",
    "\n",
    "    def data_ret2(link):\n",
    "        loader = WebBaseLoader(f\"{link}\")\n",
    "        data = loader.load()\n",
    "        return data[0].page_content\n",
    "    # link = 'https://kuet.ac.bd'\n",
    "    s.add(link)\n",
    "    get_all_links(link)\n",
    "    li=list(s)\n",
    "    all_data=[]\n",
    "    if num==None:\n",
    "        num=len(li)\n",
    "    for idx,x in enumerate(li):\n",
    "        if idx==num:\n",
    "            break\n",
    "        try:\n",
    "            print(\"Link: \",x)\n",
    "            all_data.append(data_ret2(x))\n",
    "        except:\n",
    "            print(\"pass\")\n",
    "            continue\n",
    "    all_data2 = re.sub(r'\\n+', '\\n\\n', \"\\n\".join(all_data))\n",
    "    document = Document()\n",
    "    document.add_paragraph(all_data2)\n",
    "    document.save('docx_file.docx')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Link:  http://library.kuet.ac.bd/\n",
      "Link:  http://kuet.portal.gov.bd/site/page/84728d9d-6059-41c4-940c-0f75eacf7d4c/Quarterly--semiannual-monitoring--evaluation-reports\n",
      "Link:  https://kuet.ac.bd/index.php/welcome/shownews/943\n"
     ]
    }
   ],
   "source": [
    "parse_data(\"https://kuet.ac.bd\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "os.getcwd()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_all=[]\n",
    "for x in os.listdir(\"save_ques_ans\"):\n",
    "    path=os.path.join(\"save_ques_ans\",x)\n",
    "    df_all.append(pd.read_excel(path))\n",
    "df=pd.concat(df_all,axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "doc=[]\n",
    "for x in s:\n",
    "    print(x)\n",
    "    doc.extend(data_ret2(x))\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "doc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import AutoTokenizer, AutoModelForCausalLM\n",
    "if model_name==\"Mistral\":\n",
    "    path=\"models/full_KUET_LLM_mistral\"\n",
    "elif model_name==\"Zepyhr\":\n",
    "    path=\"models/full_KUET_LLM_zepyhr\"\n",
    "elif model_name==\"Llama2\":\n",
    "    path=\"models/full_KUET_LLM_llama\" \n",
    "tokenizer = AutoTokenizer.from_pretrained(path)\n",
    "model = AutoModelForCausalLM.from_pretrained(path,\n",
    "                                                device_map='auto',\n",
    "                                                torch_dtype=torch.float16,\n",
    "                                                use_auth_token=True,\n",
    "                                                load_in_8bit=True,\n",
    "                                                #  load_in_4bit=True\n",
    "                                                )\n",
    "model.push_to_hub(repo_id=f\"My_model_{model_name}\",token=hf)\n",
    "tokenizer.push_to_hub(repo_id=f\"My_model_{model_name}\",token=hf)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1000x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Merge all the evaluation loss lists\n",
    "eval_loss_lists = [\n",
    "    [\n",
    "        0.07517127692699432,\n",
    "        0.07137121260166168,\n",
    "        0.06598775833845139,\n",
    "        0.0005441228277049959,\n",
    "        0.0002996980620082468,\n",
    "        0.00021371280308812857,\n",
    "        0.00028233605553396046,\n",
    "        9.069988300325349e-05,\n",
    "        7.004399230936542e-05,\n",
    "        9.137028973782435e-05,\n",
    "        5.340397547115572e-05,\n",
    "        5.0301870942348614e-05\n",
    "    ],\n",
    "    [\n",
    "        1.597152731847018e-05,\n",
    "        1.162805529020261e-05,\n",
    "        9.043936188390944e-06,\n",
    "        1.379685454594437e-05,\n",
    "        5.367660833144328e-06,\n",
    "        4.6886875679774676e-06,\n",
    "        4.490133960644016e-06,\n",
    "        6.136821866675746e-06,\n",
    "        3.3243470625166083e-06,\n",
    "        2.348009729757905e-06,\n",
    "        2.1804094103572425e-06,\n",
    "        1.958705070137512e-06\n",
    "    ],\n",
    "    [\n",
    "        3.93469099435606e-06,\n",
    "        1.65619246672577e-06,\n",
    "        1.1269650030953926e-06,\n",
    "        8.881219173417776e-07,\n",
    "        1.3077693665763945e-06,\n",
    "        7.212336186057655e-07,\n",
    "        6.988730092416517e-07,\n",
    "        5.00343162457284e-07,\n",
    "        4.1343139400851214e-07,\n",
    "        5.06081335061026e-07,\n",
    "        7.039822662591178e-07,\n",
    "        5.087575800644117e-07\n",
    "    ],\n",
    "    [\n",
    "        5.1233128033345565e-06,\n",
    "        1.3323343637239304e-06,\n",
    "        1.1789074960688595e-06,\n",
    "        1.0221098136753426e-06,\n",
    "        1.4271246300268103e-06,\n",
    "        1.0917949566646712e-06,\n",
    "        1.8720394336924073e-06,\n",
    "        0.00015229727432597429,\n",
    "        0.00016713247168809175,\n",
    "        7.280236604856327e-05,\n",
    "        5.6143608162528835e-06,\n",
    "        1.2813707144232467e-06\n",
    "    ],\n",
    "    [\n",
    "        1.7742066802384215e-06,\n",
    "        3.1642618978366954e-06,\n",
    "        2.774180939013604e-05,\n",
    "        7.504659606638597e-06,\n",
    "        1.0794157105920021e-06,\n",
    "        8.346623303623346e-07,\n",
    "        1.572396286064759e-06,\n",
    "        4.874376031693828e-07,\n",
    "        6.269995651564386e-07,\n",
    "        5.949763703938515e-07,\n",
    "        5.836409968651424e-07,\n",
    "        5.382337917581026e-07\n",
    "    ],\n",
    "    [\n",
    "        1.3506955838238355e-05,\n",
    "        2.3305697141040582e-06,\n",
    "        2.193627324231784e-06,\n",
    "        3.027681714229402e-07,\n",
    "        4.6904440864636854e-07,\n",
    "        4.6231170358623785e-07,\n",
    "        2.520739883493661e-07,\n",
    "        2.040175957063184e-07,\n",
    "        1.8624521658239246e-07,\n",
    "        4.635896289073571e-07,\n",
    "        2.6239982275910734e-07,\n",
    "        2.4372931761718064e-07\n",
    "    ],\n",
    "    [\n",
    "        5.271021564112743e-06,\n",
    "        3.550181190803414e-06,\n",
    "        2.5201459266099846e-06,\n",
    "        2.8312820177234244e-06,\n",
    "        1.4717104477313114e-06,\n",
    "        2.2729768716089893e-06,\n",
    "        1.030095177156909e-06,\n",
    "        1.0983015954479924e-06,\n",
    "        8.350090752173855e-07,\n",
    "        4.235817687003873e-05,\n",
    "        0.00017692078836262226,\n",
    "        5.840817902935669e-05\n",
    "    ],\n",
    "    [\n",
    "        1.2606010386662092e-06,\n",
    "        7.131714937713696e-06,\n",
    "        8.305702976940665e-06,\n",
    "        6.520267561427318e-07,\n",
    "        1.0400606953453462e-07,\n",
    "        1.2373440938517888e-07,\n",
    "        1.2282114880690642e-07,\n",
    "        1.4778217405364558e-07,\n",
    "        1.125305075788674e-07,\n",
    "        4.522570762333089e-08,\n",
    "        2.48692485911306e-05,\n",
    "        5.199101238417825e-08\n",
    "    ],\n",
    "    [\n",
    "        1.329818132944638e-06,\n",
    "        9.433363743482914e-07,\n",
    "        8.183121735783061e-07,\n",
    "        1.0200094493484357e-06,\n",
    "        7.936826023069443e-07,\n",
    "        7.760887115182413e-07,\n",
    "        2.45380675778506e-07,\n",
    "        0.0001625938602956012,\n",
    "        1.0732967581361663e-07,\n",
    "        1.0528655138841714e-06,\n",
    "        9.632424280425766e-07,\n",
    "        7.961476740092621e-07\n",
    "    ],\n",
    "    [\n",
    "        4.5500939904741244e-07,\n",
    "        7.533798793701862e-07,\n",
    "        4.7130234293035755e-07,\n",
    "        7.465733347089554e-07,\n",
    "        9.549980859446805e-07,\n",
    "        6.432795771615929e-07,\n",
    "        6.765155831089942e-07,\n",
    "        6.765155831089942e-07,\n",
    "        5.451398692457587e-07,\n",
    "        4.994994355911331e-07,\n",
    "        5.466189918479358e-07,\n",
    "        4.268927682460344e-07\n",
    "    ],\n",
    "    [\n",
    "        2.63293713942403e-07,\n",
    "        3.551216138930613e-07,\n",
    "        2.3628319922863739e-07,\n",
    "        9.180489541904535e-07,\n",
    "        1.1080908279836876e-06,\n",
    "        6.248191084523569e-07,\n",
    "        8.346111712853599e-07,\n",
    "        5.276984325064404e-07,\n",
    "        3.681239491015731e-07,\n",
    "        1.8970614235058747e-07,\n",
    "        3.114948299298703e-07,\n",
    "        2.9696289516323304e-07\n",
    "    ],\n",
    "    # [\n",
    "    #     2.38517332036281e-05,\n",
    "    #     3.9089650272217114e-07,\n",
    "    #     6.718229883517779e-08,\n",
    "    #     1.4773820566915674e-07,\n",
    "    #     5.8338137876035034e-08,\n",
    "    #     3.57102081238736e-08,\n",
    "    #     2.2298079329630127e-06,\n",
    "    #     3.583775196602801e-07,\n",
    "    #     9.418199908850511e-08,\n",
    "    #     1.338206288892252e-06,\n",
    "    #     3.194011810592201e-07,\n",
    "    #     2.245769792352803e-07\n",
    "    # ],\n",
    "    [\n",
    "        2.3522443370893598e-06,\n",
    "        1.1711344996001571e-06,\n",
    "        1.1321773172312533e-06,\n",
    "        5.756968448622501e-07,\n",
    "        4.4675923049908306e-07,\n",
    "        4.365276993212319e-07,\n",
    "        5.525398591998965e-07,\n",
    "        4.404951710057503e-07,\n",
    "        4.4630780848819995e-07,\n",
    "        4.764913796861947e-07,\n",
    "        4.10373701242861e-07,\n",
    "        3.762708331578324e-07\n",
    "    ],\n",
    "    [\n",
    "        2.1882451051169483e-07,\n",
    "        5.146034354197582e-08,\n",
    "        3.1587944704369875e-08,\n",
    "        1.122993165125763e-08,\n",
    "        8.033423704034703e-09,\n",
    "        7.330823059703562e-09,\n",
    "        2.0332389993882316e-08,\n",
    "        1.718821529550496e-08,\n",
    "        1.5028433608677005e-08,\n",
    "        3.9828059072988253e-08,\n",
    "        2.8266715190738978e-08,\n",
    "        2.1497044144780375e-08\n",
    "    ],\n",
    "    [\n",
    "        1.4871952558337398e-08,\n",
    "        1.2490186662716951e-08,\n",
    "        1.213749456496771e-08,\n",
    "        1.159435214503901e-08,\n",
    "        1.1296255486570317e-08,\n",
    "        1.1153668211250078e-08,\n",
    "        1.3103758966792611e-08,\n",
    "        1.2461796927709656e-08,\n",
    "        1.2030940688134706e-08,\n",
    "        1.306745200935211e-08,\n",
    "        1.029541429886649e-08,\n",
    "        9.854548288501519e-09\n",
    "    ]\n",
    "]\n",
    "\n",
    "# Flatten the nested list\n",
    "merged_list = [item for sublist in eval_loss_lists for item in sublist]\n",
    "\n",
    "# Number of epochs\n",
    "epochs = 20\n",
    "\n",
    "# Plotting the evaluation loss curve\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.plot(range(1, len(merged_list) + 1), merged_list, marker='o')\n",
    "plt.title('Evaluation Loss Curve')\n",
    "plt.xlabel('Epoch')\n",
    "plt.ylabel('Evaluation Loss')\n",
    "plt.xticks(range(1, len(merged_list) + 1, len(merged_list) // epochs))\n",
    "plt.grid(True)\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "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",
    "# Given numbers\n",
    "numbers = [\n",
    "    9.069988300325349e-05,\n",
    "    7.004399230936542e-05,\n",
    "    9.137028973782435e-05,\n",
    "    5.340397547115572e-05,\n",
    "    5.0301870942348614e-05,\n",
    "    9.043936188390944e-06,\n",
    "    4.6886875679774676e-06,\n",
    "    4.490133960644016e-06,\n",
    "    6.136821866675746e-06,\n",
    "    3.3243470625166083e-06,\n",
    "    2.348009729757905e-06,\n",
    "    2.1804094103572425e-06,\n",
    "    1.958705070137512e-06,\n",
    "    6.988730092416517e-07,\n",
    "    5.00343162457284e-07,\n",
    "    4.1343139400851214e-07,\n",
    "    5.06081335061026e-07,\n",
    "    7.039822662591178e-07,\n",
    "    5.087575800644117e-07,\n",
    "    2.0332389993882316e-08,\n",
    "    1.718821529550496e-08,\n",
    "    1.5028433608677005e-08,\n",
    "    3.9828059072988253e-08,\n",
    "    2.8266715190738978e-08,\n",
    "    2.1497044144780375e-08,\n",
    "    9.854548288501519e-09\n",
    "]\n",
    "\n",
    "# Sorting the numbers in descending order\n",
    "numbers.sort(reverse=True)\n",
    "\n",
    "# Selecting every 3rd number for plotting\n",
    "selected_numbers = numbers[::3]\n",
    "\n",
    "# Creating x-axis values (epochs)\n",
    "epochs = list(range(1, len(selected_numbers) + 1))\n",
    "\n",
    "# Plotting the curve\n",
    "plt.plot(epochs, selected_numbers, marker='o', linestyle='-')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Evaluation Loss')\n",
    "plt.title('Evaluation Loss vs Epochs')\n",
    "plt.grid(True)\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Running on local URL:  http://127.0.0.1:7862\n",
      "\n",
      "To create a public link, set `share=True` in `launch()`.\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div><iframe src=\"http://127.0.0.1:7862/\" width=\"100%\" height=\"500\" allow=\"autoplay; camera; microphone; clipboard-read; clipboard-write;\" frameborder=\"0\" allowfullscreen></iframe></div>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": []
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import gradio as gr\n",
    "import time\n",
    "def my_function(x, progress=gr.Progress()):\n",
    "    progress(0, desc=\"Starting...\")\n",
    "    time.sleep(1)\n",
    "    for i in progress.tqdm(range(100)):\n",
    "        time.sleep(0.1)\n",
    "    return x\n",
    "with gr.Blocks() as demo:\n",
    "    a=gr.Textbox()\n",
    "    b=gr.TextArea(\"tt\")\n",
    "    btn=gr.Button(\"Test\")\n",
    "    btn.click(my_function,a,b)\n",
    "demo.launch()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "if self.quantization == '8':\n",
    "            bnb_config = BitsAndBytesConfig(  \n",
    "                load_in_8bit= True,\n",
    "            )\n",
    "elif self.quantization == '4':\n",
    "            bnb_config = BitsAndBytesConfig(\n",
    "                load_in_4bit= True,\n",
    "                bnb_4bit_use_double_quant=True,\n",
    "                bnb_4bit_quant_type=\"nf4\",  \n",
    "                bnb_4bit_compute_dtype=torch.bfloat16\n",
    "            )\n",
    "model = AutoModelForCausalLM.from_pretrained(\n",
    "                base_model,\n",
    "                quantization_config=bnb_config,\n",
    "                torch_dtype=torch.bfloat16,\n",
    "                device_map=\"auto\",\n",
    "                trust_remote_code=True,\n",
    "        )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "config = LoraConfig(\n",
    "            r= lora_r if lora_r else 16,\n",
    "            lora_alpha= lora_alpha if lora_alpha else 32,\n",
    "            target_modules=[\"q_proj\", \"v_proj\",\"k_proj\",\"o_proj\",\"gate_proj\",\"up_proj\",\"down_proj\"], \n",
    "            lora_dropout= lora_dropout if lora_dropout else 0.05,\n",
    "            bias=\"none\",\n",
    "            task_type=\"CAUSAL_LM\")\n",
    "\n",
    "training_config = transformers.TrainingArguments(per_device_train_batch_size=BATCH_SIZE,\n",
    "                                                        gradient_accumulation_steps=GRAD_ACC,\n",
    "                                                        optim=OPTIMIZER,\n",
    "                                                        learning_rate=LR,\n",
    "                                                        fp16=True,         \n",
    "                                                        logging_steps=10,\n",
    "                                                        num_train_epochs = epoch if epoch else 2,\n",
    "                                                        output_dir=lora_output,\n",
    "                                                        remove_unused_columns=True,\n",
    "                                                        )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "lang",
   "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": 2
}