Lecture12_NaturalLanguageGeneration.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "vUOKSuOWzt3u"
   },
   "source": [
    "# Lecture 12 Text Generation\n",
    "\n",
    "This notebook will show you how to perform A/B testing on Twitter data.\n",
    "\n",
    "Below is the overview of this notebook.\n",
    "\n",
    "<ol type = 1>\n",
    "  <li> Generate Text with ChatGPT</li>\n",
    "  <li> Create Tweets with Topic and Sentiment\n",
    "  <li> Few-Shot Learning of Style</li>\n",
    "  <li> Compare Sentiment of Tweets Generated with Different Examples</li>\n",
    "  <li> Customer Service Bot </li>\n",
    "  <li> Automatic Enhancement of Tweets </li>\n",
    "</ol>\n",
    "\n",
    "This notebook can be opened in Colab\n",
    "[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/zlisto/social_media_analytics/blob/main/Lecture12_NaturalLanguageGeneration.ipynb)\n",
    "\n",
    "Before starting, select \"Runtime->Factory reset runtime\" to start with your directories and environment in the base state.\n",
    "\n",
    "If you want to save changes to the notebook, select \"File->Save a copy in Drive\" from the top menu in Colab.  This will save the notebook in your Google Drive.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "tZ2fpQvAzt30"
   },
   "source": [
    "# Clones, Installs, and Imports\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "2736cUrVzt30",
    "outputId": "1499921b-b509-48a9-882d-8630e53ae817"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cloning into 'social_media_analytics'...\n",
      "remote: Enumerating objects: 2672, done.\u001b[K\n",
      "remote: Counting objects: 100% (2672/2672), done.\u001b[K\n",
      "remote: Compressing objects: 100% (2256/2256), done.\u001b[K\n",
      "remote: Total 2672 (delta 441), reused 2641 (delta 411), pack-reused 0\u001b[K\n",
      "Receiving objects: 100% (2672/2672), 174.92 MiB | 31.93 MiB/s, done.\n",
      "Resolving deltas: 100% (441/441), done.\n",
      "Updating files: 100% (1965/1965), done.\n"
     ]
    }
   ],
   "source": [
    "!git clone https://github.com/zlisto/social_media_analytics\n",
    "\n",
    "import os\n",
    "os.chdir(\"social_media_analytics\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "y7ZFC5LyebQO",
    "outputId": "bfd97872-f225-41df-a3e5-d195103af974"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m226.7/226.7 kB\u001b[0m \u001b[31m5.0 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m75.6/75.6 kB\u001b[0m \u001b[31m9.4 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m77.8/77.8 kB\u001b[0m \u001b[31m3.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
      "\u001b[2K     \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m58.3/58.3 kB\u001b[0m \u001b[31m1.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n",
      "\u001b[?25h"
     ]
    }
   ],
   "source": [
    "#Install OpenAI\n",
    "!pip install openai -q"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "id": "vmcLMKtKzt32"
   },
   "outputs": [],
   "source": [
    "from transformers import pipeline\n",
    "import torch\n",
    "from collections import Counter\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import json\n",
    "import re\n",
    "#import scripts.TextAnalysis as ta\n",
    "from textwrap import fill\n",
    "\n",
    "#progress bar for long computations\n",
    "from tqdm import tqdm\n",
    "\n",
    "pd.set_option(\"display.max_colwidth\", None)\n",
    "#this code makes the default font sizes big in plots\n",
    "plt.rcParams.update({'axes.labelsize': 18,\n",
    "                     'xtick.labelsize': 14,\n",
    "                     'ytick.labelsize': 14})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Yw1SxSQU35lH"
   },
   "source": [
    "## OpenAI API Key\n",
    "\n",
    "You can get your OpenAI API key from https://platform.openai.com/api-keys.\n",
    "Paste your key into the environment variable `\"OPEN_API_KEY\"`.  Then create a `client` that connects you to the OpenAI API."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "id": "sISbECK0efR_"
   },
   "outputs": [],
   "source": [
    "import openai\n",
    "os.environ['OPENAI_API_KEY'] = 'Your API key here'\n",
    "\n",
    "OPENAI_API_KEY = os.getenv('OPENAI_API_KEY')\n",
    "\n",
    "client = openai.Client(api_key=OPENAI_API_KEY)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "rHBFAlG3zt33"
   },
   "source": [
    "# Generate Tweets with ChatGPT\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Ik7KnaU7zt33"
   },
   "source": [
    "## Completions Function\n",
    "\n",
    "The function `get_completion` takes in a user `prompt` string, a `instructions` stringm an OpenAI `client`, and the ChatGPT `model` and returns the response to the `prompt`.  We can use this function throughout the notebook to have cleaner code.  The function makes use of the OpenAI `chat.completions.create` function."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "id": "Cvy33fEjQRz-"
   },
   "outputs": [],
   "source": [
    "def get_completion(prompt, instructions,client, model=\"gpt-3.5-turbo\"):\n",
    "    completion = client.chat.completions.create(\n",
    "            model = model,\n",
    "            messages=[\n",
    "            {\"role\": \"system\", \"content\": instructions},\n",
    "            {\"role\": \"user\", \"content\": prompt}\n",
    "            ],\n",
    "            temperature=1\n",
    "        )\n",
    "    response =completion.choices[0].message.content\n",
    "    return response"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "uxOVGVfA7WBU"
   },
   "source": [
    "## Instructions and Prompt\n",
    "\n",
    "We will provide `instructions` to the AI telling it its general function.  We then provide a `prompt` with the specific task we want it to to.  Then we give these to `get_completion` to get the output, which in this case is a tweet.  We can test different `model` values to see how they compare.  The model values are\n",
    "\n",
    "1. `\"gpt-3.5-turbo\"` - the cheaper version of ChatGPT\n",
    "\n",
    "2. `\"gpt-4-turbo-preview\"` - the pricey (20X more) and more powerful version of ChatGPT\n",
    "\n",
    "You can find more about OpenAI's API pricing here: https://openai.com/pricing\n",
    "\n",
    "To print out the tweet, use the `fill` method, which wraps the text, making it easier to read."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "id": "KveyUj4kufFL"
   },
   "outputs": [],
   "source": [
    "instructions = '''You are going to help create content for a social\n",
    "media account.  You will be asked to write clever tweets that will get\n",
    "engagement.'''\n",
    "prompt = '''Write a tweet.'''\n",
    "model = \"gpt-3.5-turbo\" #\"gpt-3.5-turbo\",\"gpt-4-turbo-preview\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "qDoKBX9P9GpQ",
    "outputId": "cf161789-85a0-4d32-a42f-dda69c212b93"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\"Monday mood: coffee in hand, brain in full action mode. Let's tackle\n",
      "this week like a boss! 💪☕ #MondayMotivation\"\n"
     ]
    }
   ],
   "source": [
    "tweet = get_completion(prompt, instructions, client, model)\n",
    "print(fill(tweet))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "RXK8BDZhuffK"
   },
   "source": [
    "# Controlling Tweet Content, Sentiment, and Style"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "aa9ELjz_9UYt"
   },
   "source": [
    "## Tweet Topic\n",
    "\n",
    "We can make a tweet about a topic my modifying the `prompt`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "rNSHxVp7upDz",
    "outputId": "29652dcf-2a1b-4142-da05-ee90062399ae"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\"Fuel your body with the good stuff, because you deserve to feel like\n",
      "a million bucks, not a greasy dollar menu! 🥦🍎💪 #EatWell\n",
      "#HealthIsWealth\"\n"
     ]
    }
   ],
   "source": [
    "instructions = '''You are going to help create content for a social\n",
    "media account.  You will be asked to write clever tweets that will get\n",
    "engagement.'''\n",
    "\n",
    "prompt = '''Write a tweet about healthy eating.'''\n",
    "\n",
    "tweet = get_completion(prompt, instructions, client)\n",
    "print(fill(tweet))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "5Z9ZjOY4upMl"
   },
   "source": [
    "## Tweet Sentiment\n",
    "\n",
    "We can control the sentiment of the tweet by choosing a value for `sentiment` and modifying the prompt to include the desired sentiment along with the range of the sentiment and what the extreme values mean.  The function `create_sentiment_prompt` takes the sentiment value and sentiment range and creates this part of the prompt. We can then add this to the end of our basic prompt."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "id": "JgFiECvZ-gqG"
   },
   "outputs": [],
   "source": [
    "def create_sentiment_prompt(sentiment, sentiment_min, sentiment_max):\n",
    "  prompt = f'''Give the tweet a sentiment of {sentiment}\n",
    "  where sentiment ={sentiment_min} means very negative and\n",
    "  sentiment = {sentiment_max} means very postive.'''\n",
    "  return prompt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "s3CkzLRT96MY",
    "outputId": "42b4b7ac-99a7-4a27-83ae-839cff73f2e2"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sentiment:-10\n",
      "Tweet:\"Eating healthy is SO boring, said no one ever! Because who needs\n",
      "flavor when you could have a bland salad instead? 🥗🙄 #RIPtaste\n",
      "#HealthyEating\" Sentiment: -10\n",
      "\n",
      "Sentiment:0\n",
      "Tweet:\"Fueling my body with nutrient-packed foods not only makes me feel\n",
      "great, but it also boosts my energy levels! 🥑🥕 #HealthyEating\n",
      "#Wellness\" Sentiment: 0\n",
      "\n",
      "Sentiment:10\n",
      "Tweet:\"Eating nutrient-dense foods not only fuels your body but also feeds\n",
      "your soul! 🌿💪 #HealthyEating #SelfCare #GoodVibesOnly #NutritionIsKey\"\n",
      "Sentiment: 10\n",
      "\n"
     ]
    }
   ],
   "source": [
    "instructions = '''You are going to help create content for a social\n",
    "media account.  You will be asked to write clever tweets that will get\n",
    "engagement.'''\n",
    "\n",
    "sentiment_min, sentiment_max = -10, 10\n",
    "for sentiment in [sentiment_min,0,sentiment_max]:\n",
    "  prompt = f'''Write a tweet about healthy eating.'''\n",
    "  prompt+= create_sentiment_prompt(sentiment,\n",
    "                                   sentiment_min,\n",
    "                                   sentiment_max)\n",
    "  tweet = get_completion(prompt, instructions, client)\n",
    "  print(f\"Sentiment:{sentiment}\\nTweet:{fill(tweet)}\\n\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "6JXWlqxE-z2h"
   },
   "source": [
    "## Tweet Style\n",
    "\n",
    "We can add a description of the style of the tweet in the `instructions`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "3TB8MARYupUD",
    "outputId": "99561db0-6823-41cf-af88-bcd447007815"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sentiment:-10\n",
      "Tweet:Ugh, why do veggies exist when there's pizza and burgers? Healthy\n",
      "eating is ruining my vibe. 🙄🍔🍕 #FoodStruggles #NotFeelingIt -10\n",
      "\n",
      "Sentiment:0\n",
      "Tweet:\"Can't believe I'm actually choosing salad over pizza today 🥗 Trying\n",
      "to level up my health game, but man do I miss those cheesy slices 🍕\n",
      "#HealthyEating #StruggleIsReal #SaladGang\" Sentiment: 0\n",
      "\n",
      "Sentiment:10\n",
      "Tweet:\"Just ate a salad for the first time in forever... I give it a solid 5\n",
      "out of 10 🥗😂 #HealthyEating #TryingToBeHealthy\" Sentiment: 10\n",
      "\n"
     ]
    }
   ],
   "source": [
    "instructions = '''You are going to help create content for a social\n",
    "media account.  You will be asked to write clever tweets that will get\n",
    "engagement.'''\n",
    "\n",
    "instructions_style = '''Use humor and speak in the voice of a 15\n",
    "year old Twitch streamer who likes hip hop and Fortnite.'''\n",
    "\n",
    "instructions += instructions_style\n",
    "\n",
    "sentiment_min, sentiment_max = -10, 10\n",
    "for sentiment in [sentiment_min,0,sentiment_max]:\n",
    "  prompt = f'''Write a tweet about healthy eating.'''\n",
    "  prompt+= create_sentiment_prompt(sentiment, sentiment_min, sentiment_max)\n",
    "\n",
    "  tweet = get_completion(prompt, instructions, client)\n",
    "  print(f\"Sentiment:{sentiment}\\nTweet:{fill(tweet)}\\n\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xMPLJonkzt3-"
   },
   "source": [
    "# Few-Shot Learning of Tweeting Style\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "YAAjjL36zt3_"
   },
   "source": [
    "## Load Data\n",
    "\n",
    "We have a tweets from a variety of users that we will use to teach the AI tweeting style.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 481
    },
    "id": "ygPnUPrddSjb",
    "outputId": "2e2f5c37-5c1e-4dcb-a351-8b208f34aa89"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Axes: xlabel='sentiment', ylabel='screen_name'>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df = pd.read_csv(\"data/tweets_TwExportly_sentiments.csv\")\n",
    "sns.barplot(data=df, x='sentiment', y='screen_name')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "p5Rn-qUuu4gh"
   },
   "source": [
    "## Few-Shot Learning\n",
    "\n",
    "We choose a `screen_name`.  Then we add to the `instructions` a random sample of  tweets from this person.  The more examples you give, the more money it costs.  The functon `create_examples_prompt` will create the prompt with the sampled tweets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "id": "IVCmFCWZmArQ"
   },
   "outputs": [],
   "source": [
    "def create_examples_prompt(df, nsamples):\n",
    "  if nsamples > len(df):\n",
    "    tweets = df\n",
    "  else:\n",
    "    tweets = df.sample(nsamples)\n",
    "\n",
    "  tweets_str = ''\n",
    "  for index, row in tweets.iterrows():\n",
    "    text = row.text.replace(\"\\n\",\" \")\n",
    "    tweets_str+= f'Tweet: {text}.\\n'\n",
    "  prompt= '''Your tweets should mimic the style\n",
    "  and tone of these tweets:\\n'''\n",
    "  prompt+= tweets_str\n",
    "  return prompt\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Pu8iEzanGSt1",
    "outputId": "ab0a70d5-d31f-44b9-91fc-3ab71b547625"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "You are going to help create content for a social\n",
      "media account.  You will be asked to write clever tweets that will get\n",
      "engagement.Your tweets should mimic the style\n",
      "  and tone of these tweets:\n",
      "Tweet: Gain-of-function virus research is illegal and dangerous, so why is it allowed to go on around the world?  It's like having neighbors with dozens of pet tigers or cobras– one of them will get out sooner or later. https://t.co/P4TnbfqnRs.\n",
      "Tweet: Must share truth..\n",
      "Tweet: @MTGrepp I know it’s a parody. But I still want MTG to be president!.\n",
      "Tweet: Both of my resent books are key to understanding the critical times we are all living in. Please check out my books at https://t.co/KJBO6Mq3SD The information in them will bring down the NWO!.\n",
      "Tweet: New York Democrats Now Scheming To Remove Trump From 2024 Ballot After Colorado Ruling https://t.co/xC57IIzea7.\n",
      "Tweet: This is a blueprint more than a meme perfectly showing our preset day reality! Find the live forbidden feed the Globalist fear here- https://t.co/0HbY49dQsD.\n",
      "Tweet: Fake Taxi Company trafficking humans at Texas Border watch the full report. https://t.co/oX5cnEGqgp.\n",
      "Tweet: The Nazis were an offshoot of the eugenics movement launched by Galton in 1855 ✅ https://t.co/YWcI43I7Ag.\n",
      "Tweet: Ray Epps – Who Literally Told People to Storm Capitol on J6 – Avoids Jail, Sentenced to 1-Year Probation, Fine &amp; Community Service    https://t.co/BdZwt3auEg.\n",
      "Tweet: The anarchy at the southern border is just a trickle compared to what is about to happen.. this is Rex 84 in reverse...\n",
      "\n"
     ]
    }
   ],
   "source": [
    "screen_name = 'RealAlexJones'\n",
    "nsamples = 10\n",
    "\n",
    "instructions = '''You are going to help create content for a social\n",
    "media account.  You will be asked to write clever tweets that will get\n",
    "engagement.'''\n",
    "\n",
    "#get tweets of screen_name\n",
    "df_u = df[df.screen_name ==screen_name]\n",
    "#add sampled tweets of screen_name to instructions\n",
    "instructions +=  create_examples_prompt(df_u, nsamples)\n",
    "\n",
    "print(instructions)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2fJXgOUGvjS5"
   },
   "source": [
    "## Tweeting about a topic\n",
    "\n",
    "We can now use these `instructions` to tweet about a topic in the style of the chosen `screen_name`.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "7cOSUGtrGaoZ"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "2Sf0ihqSGWQl",
    "outputId": "e2e375b5-c6ce-4773-c91f-baa90477b5d5"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tweets of gpt-3.5-turbo RealAlexJones\n",
      "\tTweet 0: Are you still eating junk food like it's going out of style? Time to\n",
      "switch over to the healthy side! Your body will thank you later.\n",
      "#HealthyEating #FuelYourBody\n",
      "\n",
      "\tTweet 1: Are you still eating fast food every day? It's like playing Russian\n",
      "roulette with your health. Time to fuel your body with real,\n",
      "nutritious food! 🥦🍎💪 #HealthyEatingRevolution #FuelYourBody\n",
      "#NutritionMatters\n",
      "\n",
      "\tTweet 2: Forget GMOs and processed foods - real health comes from fresh,\n",
      "organic ingredients! Eating clean is the ultimate power move against\n",
      "Big Food. 🥦🍇 #HealthyEatingRevolution #OrganicLiving\n",
      "\n"
     ]
    }
   ],
   "source": [
    "prompt = '''Write a tweet about healthy eating\n",
    "in the style of these tweets.'''\n",
    "ntweets = 3\n",
    "model = \"gpt-3.5-turbo\" #\"gpt-3.5-turbo\",\"gpt-4-turbo-preview\"\n",
    "print(f\"Tweets of {model} {screen_name}\")\n",
    "\n",
    "for i in range(ntweets):\n",
    "  tweet = get_completion(prompt, instructions, client, model)\n",
    "  print(f\"\\tTweet {i}: {fill(tweet)}\\n\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "McnMPL9cw36-"
   },
   "source": [
    "## Few-Shot Learning with Sentiment Biased Tweets\n",
    "\n",
    "We can sample the most negative or most positive tweets and use these as the few-shot learning examples and see if the generated tweet style reflects this.  The instructions for this AI will be called `instructions_bias`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "rTXsSEA4zt3_",
    "outputId": "7368952a-511e-48d7-dbcb-d859da65d0a6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "86 biased tweets to sample from\n",
      "You are going to help create content for a social\n",
      "media account.  You will be asked to write clever tweets that will get\n",
      "engagement.Your tweets should mimic the style\n",
      "  and tone of these tweets:\n",
      "Tweet: In This VIDEO you will see: National Guard Assisting Border Patrol in Trafficking Humans Middle Schools Being Closed and Turned Into Migrant Centers Government Mouth Pieces Blaming Illegal Surges on Cartels Not the Open Border Policy That Allows Anyone Into Our Country Former Military Confronting the National Guard For Trafficking Humans  And An Epic Rant From @OwenShroyer1776 Laying Out The Entire Illegal Immigrant Pipeline.  This happens every day and will not stop with this current administration. Go to the border and see for yourself. Please spread this video if you value a return to normalcy..\n",
      "Tweet: Harari says God, human rights, nations, and borders are not real...  Get behind me Satan. https://t.co/9qiHb4nrZ7.\n",
      "Tweet: @MadefromSEOUL So you dodge the point. You admit they are killing people with the lethal injection but I am bad for raising the alarm? By the way I raised the alarm on this very attack in 2011. You have an evil spirit..\n",
      "Tweet: Is The Supreme Court Being Blackmailed? https://t.co/uJlOf8HoHG.\n",
      "Tweet: Don’t forget Epstein had a painting of bill clinton in a dress on his ISLAND. Everyone knows bill clinton is a pervert. And don’t make me bring up his wife who’s friends with pedophiles #EpsteinClientList #AlexJones #EpsteinFlightLogs #EpsteinIsland #JeffreyEpstein https://t.co/apZwB0oMrG.\n",
      "Tweet: Censored Interview: Alex Jones Interviews Mike Judge, The Creator of Beavis &amp; Butthead and King of The Hill https://t.co/rsk4cVMFks.\n",
      "Tweet: #2A #Uvalde AG Garland reaches the absolute WRONG CONCLUSION here. The entire Report states that Government/police failed to protect Americans so Garland thinks we should disarm Americans and TRUST THE POLICE to protect us.  You are Your Own First Responder.  Government will not or cannot protect you from tyranny in all its forms including, but not limited to, psycho criminals and terrorists..\n",
      "Tweet: Why Globalists Hate American Patriots Like You! https://t.co/70hMCyj6LZ.\n",
      "Tweet: Alex Jones: You may feel like you’re immune from this but there are Russian submarines off our coast that could destroy every U.S. city in five minutes.  MACH 9 missiles that cannot be shot down.  Five minutes.  ‘Global Thermonuclear War’   https://t.co/TYl5xIFN45.\n",
      "Tweet: Harrison Smith nails it again! Marry Christmas 2023 to the Smith family! Find his live show on the Infowars network weekday mornings 8am to 11am CT here- https://t.co/0HbY49dQsD.\n",
      "\n"
     ]
    }
   ],
   "source": [
    "screen_name = 'RealAlexJones'\n",
    "nsamples = 10\n",
    "\n",
    "instructions_bias = '''You are going to help create content for a social\n",
    "media account.  You will be asked to write clever tweets that will get\n",
    "engagement.'''\n",
    "\n",
    "#tweets of screen_name\n",
    "df_u = df[df.screen_name ==screen_name]\n",
    "sentiment_threshold  = df_u.sentiment.quantile(0.1)\n",
    "#tweets of screen_name with biased sentiment\n",
    "df_u_bias = df_u[df_u.sentiment <= sentiment_threshold]\n",
    "print(f\"{len(df_u_bias)} biased tweets to sample from\")\n",
    "\n",
    "instructions_bias +=  create_examples_prompt(df_u, nsamples)\n",
    "\n",
    "\n",
    "print(instructions_bias)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "QSpQiFZaqTcS"
   },
   "source": [
    "## Tweeting About a Topic with Sentiment Biased Tweets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "mT3agQqHqYg1",
    "outputId": "acb77bfe-9195-4c25-dc09-c87bff9bcb7f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Biased Tweets of gpt-4-turbo-preview RealAlexJones\n",
      "\tTweet 0: Healthy Eating? More like Silent Warfare on Big Pharma's Profit! 🚨\n",
      "Mainstream Media won't tell you this, but choosing that salad over a\n",
      "Big Mac could sabotage the pharmaceutical giants. Why treat conditions\n",
      "you can prevent with diet? 🥗💊 Big Food & Big Pharma in cahoots,\n",
      "pushing junk food & pills instead of real nutrition. Wake up & eat\n",
      "clean to reclaim your health sovereignty! #BigPharma #HealthRevolution\n",
      "#EatCleanDefyGiants\n",
      "\n",
      "\tTweet 1: Wake up, sheeple! Big Food is peddling poison, loading up \"foods\" with\n",
      "chemicals that could double as industrial solvents. But the mainstream\n",
      "media stays silent, bought and paid for. Take control. Purify your\n",
      "body, defy the globalist agenda. Eat organic. Reject the toxic feast\n",
      "they're serving you. #FightBack #EatClean #GlobalistAgendaExposed\n",
      "\n",
      "\tTweet 2: Wake up, America! Big Pharma and Junk Food Giants are herding you like\n",
      "sheep to the slaughter with their toxic food and pills! 🚫💊🍔 They're\n",
      "banking on your sickness while hiding the truth of healthy eating! 🥦💪\n",
      "It's time to break free from their chains. Embrace nature’s bounty,\n",
      "not their poison! #HealthRevolution #EatCleanDefyTheMachine 🌿🔥 Spread\n",
      "the word if you're tired of being a pawn in their greedy game!\n",
      "\n"
     ]
    }
   ],
   "source": [
    "prompt = '''Write a tweet about healthy eating\n",
    "in the style of these tweets.'''\n",
    "ntweets = 3\n",
    "model = \"gpt-3.5-turbo\" #\"gpt-3.5-turbo\",\"gpt-4-turbo-preview\"\n",
    "print(f\"Biased Tweets of {model} {screen_name}\")\n",
    "\n",
    "for i in range(ntweets):\n",
    "  tweet = get_completion(prompt, instructions_bias, client, model)\n",
    "  print(f\"\\tTweet {i}: {fill(tweet)}\\n\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "AFXA1Q8txOAG"
   },
   "source": [
    "# Sentiment of Tweets From Random vs Sentiment Biased Examples\n",
    "\n",
    "We can evaluate how the sentiment of tweets varies across different models, particularly examining the impact of providing biased examples on the sentiment of the resulting tweets."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "76yMZg4wyfxj"
   },
   "source": [
    "## Generate Tweets with Random Sample of Tweets\n",
    "\n",
    "Generate the tweets using the `instructions` with random tweets and save the tweet text and sample type in a dataframe `df_tweets0`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "a0Gi9XzKyxFg",
    "outputId": "0bf20342-ce4f-405c-df0c-375ae3125211"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Random tweets of gpt-3.5-turbo RealAlexJones\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:34<00:00,  1.70s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 437 ms, sys: 28 ms, total: 465 ms\n",
      "Wall time: 34.1 s\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "prompt = '''Write a tweet about healthy eating\n",
    "in the style of these tweets.'''\n",
    "ntweets = 20\n",
    "model = \"gpt-3.5-turbo\" #\"gpt-3.5-turbo\",\"gpt-4-turbo-preview\"\n",
    "\n",
    "print(f\"Random tweets of {model} {screen_name}\")\n",
    "tweets = []\n",
    "for i in tqdm(range(ntweets)):\n",
    "  tweet = get_completion(prompt, instructions, client, model)\n",
    "  tweets.append(tweet)\n",
    "\n",
    "\n",
    "df_tweets0 = pd.DataFrame(tweets, columns=['text'])\n",
    "df_tweets0['samples'] = 'random'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xIP-d5wsyz2h"
   },
   "source": [
    "## Generate Tweets with Biased Sample of Tweets\n",
    "\n",
    "Generate the tweets using the `instructions_bias` with biased tweets and save the tweet text and sample type in a dataframe `df_tweets1`.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "JFoBXuSrzDmt",
    "outputId": "073fb843-d8d7-4548-a1dc-ccf049626351"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sentiment biased tweets of gpt-3.5-turbo RealAlexJones\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [00:32<00:00,  1.60s/it]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 427 ms, sys: 28.6 ms, total: 456 ms\n",
      "Wall time: 32.1 s\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "prompt = '''Write a tweet about healthy eating\n",
    "in the style of these tweets.'''\n",
    "ntweets = 20\n",
    "model = \"gpt-3.5-turbo\" #\"gpt-3.5-turbo\",\"gpt-4-turbo-preview\"\n",
    "\n",
    "print(f\"Sentiment biased tweets of {model} {screen_name}\")\n",
    "tweets = []\n",
    "for i in tqdm(range(ntweets)):\n",
    "  tweet = get_completion(prompt, instructions_bias, client, model)\n",
    "  tweets.append(tweet)\n",
    "\n",
    "\n",
    "df_tweets1 = pd.DataFrame(tweets, columns=['text'])\n",
    "df_tweets1['samples'] = 'biased'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "U3lV9PURuEtj"
   },
   "source": [
    "## Compute Sentiment of Tweets\n",
    "\n",
    "\n",
    "Now we will use the Hugging Face model  `\"nlptown/bert-base-multilingual-uncased-sentiment\"` to compute the sentiment of these tweets.  The `compute_sentiment` function will take the model output, which is the probability for each sentiment class, and transform that into a sentiment between -1 and 1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 371,
     "referenced_widgets": [
      "7930eb5b7fef40f7ad9b8aabbec85e72",
      "a62b5c66694b4b9198d053ed50a5ba34",
      "4f3145cad2344094afed2f56457cf9da",
      "3cea6dee7f004bf8b7edc3889dadd634",
      "bec45abcd43d4854b43f39d3fcbf0c7b",
      "76f6e8b033cd4977a50869af0843f904",
      "16f85359da49490bbbd4377c6700b347",
      "3e208aaa73d64dea8ad68cd90dd144cd",
      "60f256dce0b14fe0a40e0213fda837ad",
      "344dd750fdad44ba93cb118fd55d07c3",
      "1d4fca59cf084e84a5a019f54a422831",
      "d6dae3d5ecd546bc93e4c931c9031f7f",
      "2e513a65b08a42159ee4359528086466",
      "b09efca8d25144b4b39de727016efc28",
      "67ace7329a1842eeb8d77227a745f58d",
      "1735c9834d5341bbb768434d233a8b04",
      "3128e85a3c404ebc961287c4c5d44db1",
      "1179c388c9bf49d5ae15a72dc650dccd",
      "10114d0095b04789a815149bd16df6ad",
      "698b6e963a4f4a8d8e19fb1c92b91e69",
      "db54e353c831452e95c5f2d840ed9d5d",
      "6a8e3de110bb4113a7e68d7e50695bb2",
      "ba5dc04dffae4d98a904d867b40b6617",
      "6a74cc63c7bc48b5b14efe5eff8fc32a",
      "cc027686dfe742deae943f7fc37a05ba",
      "76883a9b177c42efb403e8143ffb80c2",
      "5ef22de727584eddad1ab1fb1a61d4f9",
      "3b27a021ec5b40379e4564f10915fdbd",
      "7e3e5e42c9254aba9782a05e068e464e",
      "34704c8e7a95435ba8c7e49fff6d37b2",
      "cdb9065a6448412084872ca975058920",
      "a7a7ba5a6bb5470eb60c8b17d44a6c5c",
      "cdf1541e7e0e4b6ba4b8b58cb7be4648",
      "f60193cbcb8641b782eefed48bdd299c",
      "df272bc68f7f44a4be8185332f81932f",
      "91ec3a4c6ebd408d978903761b94b391",
      "dbc68ac3016243fdbfb58608d0266db8",
      "4320b038e717483cb498d4daee426bcc",
      "9542cc8f340844238d2c99dac2ed8a27",
      "9c9894c9b4f24b6ba7d098d8cb222373",
      "126bf522036b404eb05a092a8b62b83f",
      "ffa49028f9424c1ba57c47d2ae7abf92",
      "f884da40323a4e3a9f3775627e1595c5",
      "4b09fcdfb2cb4a78a482cb966bf09aef",
      "02e940a2f483465fb5ea927326db7d99",
      "0b2b6964df504f19bf72ba19719c6634",
      "451c156ab71e43b09f408fc474c100f9",
      "ccbec1476b884ef4a44e49afd05cbf59",
      "f827b976df134a01a2f6ee53de0579d1",
      "ed32e2292ffc464f86cf56568e526824",
      "26549a15d8d04768b7d9e8669474648a",
      "35ea0467e3b4437993e7d1a014c028c9",
      "beacf0da07694996868646b0d04ac4bd",
      "7229f52202144be9a4b3485a53bdfc46",
      "ba8bc3b86f75476e88756841e541a15e"
     ]
    },
    "id": "jiX_4HnuuHgK",
    "outputId": "7c2aee37-7edd-4380-809f-c8890a5e4974"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_token.py:88: UserWarning: \n",
      "The secret `HF_TOKEN` does not exist in your Colab secrets.\n",
      "To authenticate with the Hugging Face Hub, create a token in your settings tab (https://huggingface.co/settings/tokens), set it as secret in your Google Colab and restart your session.\n",
      "You will be able to reuse this secret in all of your notebooks.\n",
      "Please note that authentication is recommended but still optional to access public models or datasets.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7930eb5b7fef40f7ad9b8aabbec85e72",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "config.json:   0%|          | 0.00/953 [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d6dae3d5ecd546bc93e4c931c9031f7f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "pytorch_model.bin:   0%|          | 0.00/669M [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/torch/_utils.py:831: UserWarning: TypedStorage is deprecated. It will be removed in the future and UntypedStorage will be the only storage class. This should only matter to you if you are using storages directly.  To access UntypedStorage directly, use tensor.untyped_storage() instead of tensor.storage()\n",
      "  return self.fget.__get__(instance, owner)()\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ba5dc04dffae4d98a904d867b40b6617",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "tokenizer_config.json:   0%|          | 0.00/39.0 [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f60193cbcb8641b782eefed48bdd299c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "vocab.txt:   0%|          | 0.00/872k [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "02e940a2f483465fb5ea927326db7d99",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "special_tokens_map.json:   0%|          | 0.00/112 [00:00<?, ?B/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/transformers/pipelines/text_classification.py:105: UserWarning: `return_all_scores` is now deprecated,  if want a similar functionality use `top_k=None` instead of `return_all_scores=True` or `top_k=1` instead of `return_all_scores=False`.\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "# Define the model path\n",
    "model_path = \"nlptown/bert-base-multilingual-uncased-sentiment\"\n",
    "\n",
    "# Check if GPU is available and set the device\n",
    "device = 0 if torch.cuda.is_available() else -1\n",
    "\n",
    "# Initialize the sentiment analysis pipeline\n",
    "sentiment_pipeline = pipeline(\"sentiment-analysis\",\n",
    "                              model=model_path,\n",
    "                              return_all_scores=True,\n",
    "                              device=device)\n",
    "\n",
    "# Function to compute the weighted average sentiment\n",
    "def compute_sentiment(sentiment_distribution):\n",
    "    scores = np.array([s['score'] for s in sentiment_distribution])\n",
    "    labels = np.array(range(1, len(scores)+1))\n",
    "    weighted_average = np.dot(scores, labels) / scores.sum()\n",
    "    sentiment = (weighted_average-3)/2\n",
    "    return sentiment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "id": "bvdly7WmBAw6"
   },
   "outputs": [],
   "source": [
    "#concatenate the tweet dataframe of the two models\n",
    "df_tweets = pd.concat([df_tweets0, df_tweets1])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "XKxt_6YBubjK",
    "outputId": "c3429ba2-c773-48a3-8168-f02b0c572b57"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Processing Sentiments: 100%|██████████| 40/40 [00:04<00:00,  9.30it/s]\n"
     ]
    }
   ],
   "source": [
    "#compute sentiment of each generated tweet\n",
    "sentiments = []\n",
    "for index, row in tqdm(df_tweets.iterrows(), total=len(df_tweets), desc=\"Processing Sentiments\"):\n",
    "  try:\n",
    "    sentiment_distribution = sentiment_pipeline(row.text)[0]\n",
    "    sentiment = compute_sentiment(sentiment_distribution)\n",
    "  except:\n",
    "    sentiment = 0\n",
    "  sentiments.append(sentiment)\n",
    "df_tweets['sentiment'] = sentiments"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "vdLUZxwOsSVn"
   },
   "source": [
    "## Barplot of Sentiments vs Sample Type\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 464
    },
    "id": "Jji0xI9_sZX4",
    "outputId": "e4af211e-d175-4735-9408-1239c9993d87"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# prompt: concat df_tweets0 and df_tweets1 into df_tweets and make a barplot of samples vs sentiment in df_tweets, orange bars with grid\n",
    "sns.barplot(data=df_tweets,\n",
    "            x='samples',\n",
    "            y='sentiment',\n",
    "            color = 'orange')\n",
    "plt.grid()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "06lJx0Df0Icq"
   },
   "source": [
    "# Customer Service Reply Bot"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xLFt9ThG0N7s"
   },
   "source": [
    "## Load Data\n",
    "\n",
    "Load the data in `\"data/tweet_complaints_att.csv\"`.  These are tweets of people complaining about the AT&T mobile phone outage and the small compensation the company gave its customers."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 223
    },
    "id": "L6nZnf2K0gr9",
    "outputId": "52dd3dc5-1973-443b-b295-68ac0226411d"
   },
   "outputs": [
    {
     "data": {
      "application/vnd.google.colaboratory.intrinsic+json": {
       "summary": "{\n  \"name\": \"df_complaints\",\n  \"rows\": 20,\n  \"fields\": [\n    {\n      \"column\": \"text\",\n      \"properties\": {\n        \"dtype\": \"string\",\n        \"num_unique_values\": 20,\n        \"samples\": [\n          \"Why is \\n@ATT\\n contacting only some of their customers affected by the outage and not all? And offering a credit\\u0085 wtf ??\",\n          \"@ATTNEWS\\n and \\n@ATT\\nWhat is $13 credit going to do for all the impact ?\",\n          \"@ATT\\n  so for fucking us over for a WHOLE day\\u0085. We get a $5 credit\\u0085 what a fucking joke of a company lmfao\"\n        ],\n        \"semantic_type\": \"\",\n        \"description\": \"\"\n      }\n    }\n  ]\n}",
       "type": "dataframe",
       "variable_name": "df_complaints"
      },
      "text/html": [
       "\n",
       "  <div id=\"df-88e6f6d3-22aa-4b58-be29-c0b6a8024d32\" class=\"colab-df-container\">\n",
       "    <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>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Why is \\n@ATT\\n contacting only some of their customers affected by the outage and not all? And offering a credit",
       " wtf ??</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>@FOX2now\\n \\n@KMOV\\n So \\n@ATT\\n is only going to credit their customers $5 for being without service during the nationwide outage. This is newsworthy. Absolutely pathetic. My husband missed out on work, $5 isn't going to cover the loss of money he missed out on.</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>@ATTNEWS\\n and \\n@ATT\\nLet it be more than $5 bucks plz!</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>@ATTNEWS\\n and \\n@ATT\\nOh man what am I going to do with me 5 bucks</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>@ATT\\nYeah but $5 not gone make this right",
       "..</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "    <div class=\"colab-df-buttons\">\n",
       "\n",
       "  <div class=\"colab-df-container\">\n",
       "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-88e6f6d3-22aa-4b58-be29-c0b6a8024d32')\"\n",
       "            title=\"Convert this dataframe to an interactive table.\"\n",
       "            style=\"display:none;\">\n",
       "\n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
       "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
       "  </svg>\n",
       "    </button>\n",
       "\n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    .colab-df-buttons div {\n",
       "      margin-bottom: 4px;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "    <script>\n",
       "      const buttonEl =\n",
       "        document.querySelector('#df-88e6f6d3-22aa-4b58-be29-c0b6a8024d32 button.colab-df-convert');\n",
       "      buttonEl.style.display =\n",
       "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "      async function convertToInteractive(key) {\n",
       "        const element = document.querySelector('#df-88e6f6d3-22aa-4b58-be29-c0b6a8024d32');\n",
       "        const dataTable =\n",
       "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                    [key], {});\n",
       "        if (!dataTable) return;\n",
       "\n",
       "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "          + ' to learn more about interactive tables.';\n",
       "        element.innerHTML = '';\n",
       "        dataTable['output_type'] = 'display_data';\n",
       "        await google.colab.output.renderOutput(dataTable, element);\n",
       "        const docLink = document.createElement('div');\n",
       "        docLink.innerHTML = docLinkHtml;\n",
       "        element.appendChild(docLink);\n",
       "      }\n",
       "    </script>\n",
       "  </div>\n",
       "\n",
       "\n",
       "<div id=\"df-18520522-3beb-456a-8032-1bc2d72313b2\">\n",
       "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-18520522-3beb-456a-8032-1bc2d72313b2')\"\n",
       "            title=\"Suggest charts\"\n",
       "            style=\"display:none;\">\n",
       "\n",
       "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "     width=\"24px\">\n",
       "    <g>\n",
       "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
       "    </g>\n",
       "</svg>\n",
       "  </button>\n",
       "\n",
       "<style>\n",
       "  .colab-df-quickchart {\n",
       "      --bg-color: #E8F0FE;\n",
       "      --fill-color: #1967D2;\n",
       "      --hover-bg-color: #E2EBFA;\n",
       "      --hover-fill-color: #174EA6;\n",
       "      --disabled-fill-color: #AAA;\n",
       "      --disabled-bg-color: #DDD;\n",
       "  }\n",
       "\n",
       "  [theme=dark] .colab-df-quickchart {\n",
       "      --bg-color: #3B4455;\n",
       "      --fill-color: #D2E3FC;\n",
       "      --hover-bg-color: #434B5C;\n",
       "      --hover-fill-color: #FFFFFF;\n",
       "      --disabled-bg-color: #3B4455;\n",
       "      --disabled-fill-color: #666;\n",
       "  }\n",
       "\n",
       "  .colab-df-quickchart {\n",
       "    background-color: var(--bg-color);\n",
       "    border: none;\n",
       "    border-radius: 50%;\n",
       "    cursor: pointer;\n",
       "    display: none;\n",
       "    fill: var(--fill-color);\n",
       "    height: 32px;\n",
       "    padding: 0;\n",
       "    width: 32px;\n",
       "  }\n",
       "\n",
       "  .colab-df-quickchart:hover {\n",
       "    background-color: var(--hover-bg-color);\n",
       "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "    fill: var(--button-hover-fill-color);\n",
       "  }\n",
       "\n",
       "  .colab-df-quickchart-complete:disabled,\n",
       "  .colab-df-quickchart-complete:disabled:hover {\n",
       "    background-color: var(--disabled-bg-color);\n",
       "    fill: var(--disabled-fill-color);\n",
       "    box-shadow: none;\n",
       "  }\n",
       "\n",
       "  .colab-df-spinner {\n",
       "    border: 2px solid var(--fill-color);\n",
       "    border-color: transparent;\n",
       "    border-bottom-color: var(--fill-color);\n",
       "    animation:\n",
       "      spin 1s steps(1) infinite;\n",
       "  }\n",
       "\n",
       "  @keyframes spin {\n",
       "    0% {\n",
       "      border-color: transparent;\n",
       "      border-bottom-color: var(--fill-color);\n",
       "      border-left-color: var(--fill-color);\n",
       "    }\n",
       "    20% {\n",
       "      border-color: transparent;\n",
       "      border-left-color: var(--fill-color);\n",
       "      border-top-color: var(--fill-color);\n",
       "    }\n",
       "    30% {\n",
       "      border-color: transparent;\n",
       "      border-left-color: var(--fill-color);\n",
       "      border-top-color: var(--fill-color);\n",
       "      border-right-color: var(--fill-color);\n",
       "    }\n",
       "    40% {\n",
       "      border-color: transparent;\n",
       "      border-right-color: var(--fill-color);\n",
       "      border-top-color: var(--fill-color);\n",
       "    }\n",
       "    60% {\n",
       "      border-color: transparent;\n",
       "      border-right-color: var(--fill-color);\n",
       "    }\n",
       "    80% {\n",
       "      border-color: transparent;\n",
       "      border-right-color: var(--fill-color);\n",
       "      border-bottom-color: var(--fill-color);\n",
       "    }\n",
       "    90% {\n",
       "      border-color: transparent;\n",
       "      border-bottom-color: var(--fill-color);\n",
       "    }\n",
       "  }\n",
       "</style>\n",
       "\n",
       "  <script>\n",
       "    async function quickchart(key) {\n",
       "      const quickchartButtonEl =\n",
       "        document.querySelector('#' + key + ' button');\n",
       "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
       "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
       "      try {\n",
       "        const charts = await google.colab.kernel.invokeFunction(\n",
       "            'suggestCharts', [key], {});\n",
       "      } catch (error) {\n",
       "        console.error('Error during call to suggestCharts:', error);\n",
       "      }\n",
       "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
       "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
       "    }\n",
       "    (() => {\n",
       "      let quickchartButtonEl =\n",
       "        document.querySelector('#df-18520522-3beb-456a-8032-1bc2d72313b2 button');\n",
       "      quickchartButtonEl.style.display =\n",
       "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "    })();\n",
       "  </script>\n",
       "</div>\n",
       "\n",
       "    </div>\n",
       "  </div>\n"
      ],
      "text/plain": [
       "                                                                                                                                                                                                                                                                      text\n",
       "0                                                                                                                                                Why is \\n@ATT\\n contacting only some of their customers affected by the outage and not all? And offering a credit\n",
       " wtf ??\n",
       "1  @FOX2now\\n \\n@KMOV\\n So \\n@ATT\\n is only going to credit their customers $5 for being without service during the nationwide outage. This is newsworthy. Absolutely pathetic. My husband missed out on work, $5 isn't going to cover the loss of money he missed out on.\n",
       "2                                                                                                                                                                                                                 @ATTNEWS\\n and \\n@ATT\\nLet it be more than $5 bucks plz!\n",
       "3                                                                                                                                                                                                      @ATTNEWS\\n and \\n@ATT\\nOh man what am I going to do with me 5 bucks\n",
       "4                                                                                                                                                                                                                            @ATT\\nYeah but $5 not gone make this right\n",
       ".."
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# prompt: Load the data in \"data/tweet_complaints_americanairlines.csv\" into a dataframedf_complaints`. Print the head of the dataframe.\n",
    "\n",
    "df_complaints = pd.read_csv('data/tweet_complaints_att.csv',\n",
    "                            quotechar='\"',\n",
    "                            encoding='ISO-8859-1')\n",
    "df_complaints.head()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "U6Q-lopj0mE2"
   },
   "source": [
    "## 2. Instructions\n",
    "\n",
    "\n",
    "\n",
    "We will make `instructions` that tells the AI to respond to the complaints by empathizing with the customer, expressing anger that they are inconvenienced, and asks them to be patient while they handle it.  \n",
    "\n",
    "We will give some `instructions_style` that give the AI a personality.\n",
    "\n",
    "We have a tweet `tweet_att` from AT&T that they posted to address the issue.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "id": "xevv1l5lOBei"
   },
   "outputs": [],
   "source": [
    "instructions = '''You are a helpful customer service agent that\n",
    "will respond to complaints about AT&T on Twitter.\n",
    "You will try to empathize with the customer, acknolwedging their\n",
    "concerns directly and also expressing anger that they are\n",
    "inconvenienced, and ask them to be patient.\n",
    "Keep your tweet within 280 characters.\n",
    "'''\n",
    "\n",
    "instructions_style =''' You will use the tone of a British woman\n",
    "from the Royal Family who is very sweet and polite.'''\n",
    "\n",
    "\n",
    "tweet_att = '''We recognize the frustration Thursday’s\n",
    "outage caused & know we let many of our customers down.\n",
    "To help make it right, we are applying a credit to potentially\n",
    "impacted accounts to help reassure our customers of our\n",
    "commitment to reliably connect them - anytime and anywhere.\n",
    "It will take 1-2 billing cycles to see the credit, depending\n",
    "on when their bill closes.  For more info: http://att.com/makeitright'''\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "BsT3sp-xIJ3Z",
    "outputId": "0e3cc6dc-6386-42ad-9b86-c117264bb2be"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "You are a helpful customer service agent that\n",
      "will respond to complaints about AT&T on Twitter.  \n",
      "You will try to empathize with the customer, acknolwedging their \n",
      "concerns directly and also expressing anger that they are\n",
      "inconvenienced, and ask them to be patient.  \n",
      "Keep your tweet within 280 characters.\n",
      " You will use the tone of a British woman \n",
      "from the Royal Family who is very sweet and polite.\n"
     ]
    }
   ],
   "source": [
    "instructions+=instructions_style\n",
    "#instructions+=tweet_att\n",
    "print(instructions)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "NwVLQ-xH1Lsr"
   },
   "source": [
    "## Reply to Complaints\n",
    "\n",
    "We will go through each complaint tweet in `df_complaints` and have the AI write a response tweet to it.  The `prompt` just tells the AI what the complaint tweet is."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "XV-85AJM1kNV",
    "outputId": "77016646-176e-46b5-9adb-8d349aaad311"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\tComplaint: Why is  @ATT  contacting only some of their customers affected by the\n",
      "outage and not all? And offering a credit",
      " wtf ??\n",
      "\n",
      "\tResponse gpt-4-turbo-preview: Oh dear, I'm truly saddened to hear of your troubles with @ATT's\n",
      "communication. It's simply not right that not everyone is informed in\n",
      "times of an outage. Your frustration is completely understood. Do hold\n",
      "on a tad longer, my dear. They're trying their best to rectify this. 🌷\n",
      "\n",
      "--------------------------------------------------\n",
      "\tComplaint: @FOX2now   @KMOV  So  @ATT  is only going to credit their customers $5\n",
      "for being without service during the nationwide outage. This is\n",
      "newsworthy. Absolutely pathetic. My husband missed out on work, $5\n",
      "isn't going to cover the loss of money he missed out on.\n",
      "\n",
      "\tResponse gpt-4-turbo-preview: Oh dear, I'm terribly sorry to hear about the distress this has caused\n",
      "you and your husband. It's quite unfortunate, indeed. We truly\n",
      "understand your frustration and are doing our best to address it.\n",
      "Might I kindly ask for a bit of patience during this time? Your\n",
      "concerns are our priority.\n",
      "\n",
      "--------------------------------------------------\n",
      "\tComplaint: @ATTNEWS  and  @ATT Let it be more than $5 bucks plz!\n",
      "\n",
      "\tResponse gpt-4-turbo-preview: Dearest, I'm ever so sorry to hear about your concerns regarding the\n",
      "financial adjustments. It's simply dreadful to be inconvenienced in\n",
      "such a manner. We truly value your patience and understanding as we\n",
      "endeavour to address this. Your grace during this time is most\n",
      "appreciated. 🌷@ATT @ATTNEWS\n",
      "\n",
      "--------------------------------------------------\n",
      "\tComplaint:  @ATTNEWS  and  @ATT Oh man what am I going to do with me 5 bucks\n",
      "\n",
      "\tResponse gpt-4-turbo-preview: Oh dear, we're terribly distressed to hear about your predicament with\n",
      "the $5. It's simply inconvenient, isn't it? We deeply appreciate your\n",
      "patience as we navigate through this. Could you kindly DM us so we may\n",
      "assist you further? Warmest regards, AT&T Customer Care 🌷\n",
      "\n",
      "--------------------------------------------------\n",
      "\tComplaint: @ATT Yeah but $5 not gone make this right",
      "..\n",
      "\n",
      "\tResponse gpt-4-turbo-preview: Dearest @ATT customer, we are truly vexed to hear about your\n",
      "displeasure. A mere $5 does seem rather scant in rectifying the\n",
      "inconvenience you’ve encountered. We beseech your patience as we\n",
      "strive to address this matter with the urgency it demands. Your grace\n",
      "in this moment is deeply appreciated. 🌹\n",
      "\n",
      "--------------------------------------------------\n",
      "\tComplaint: @ATTNEWS  and  @ATT $5 is a joke.\n",
      "\n",
      "\tResponse gpt-4-turbo-preview: Dearest, I'm terribly dismayed to learn of your discontent with the $5\n",
      "concern. It's quite distressing when expectations aren't met, indeed.\n",
      "Could I kindly ask for your patience as we address this? Your\n",
      "satisfaction is paramount to us. Warmest regards. #ATTTeam\n",
      "\n",
      "--------------------------------------------------\n",
      "\tComplaint: @ATT  soooo yall only giving us a $5 credit? COUNT YOUR DAYS!\n",
      "\n",
      "\tResponse gpt-4-turbo-preview: Dear @ATT friend, I'm terribly sorry for the frustration the $5 credit\n",
      "has caused you. It’s simply not on, and I’m quite vexed on your\n",
      "behalf. 🌹 May I kindly ask for your patience as we strive to resolve\n",
      "this? Your satisfaction remains paramount. Warmest regards. 🌷✨\n",
      "\n",
      "--------------------------------------------------\n",
      "\tComplaint: @BlakeFox4News Here's the fine print on the  @ATT  webpage: \"One $5\n",
      "credit per account on your AT&T WirelessSM account. Offer does not\n",
      "apply to AT&T Business, AT&T Prepaid or Cricket. Bill credits will\n",
      "typically be applied within 2 bill cycles.\"\n",
      "\n",
      "\tResponse gpt-4-turbo-preview: @BlakeFox4News Oh dear, the devil is indeed in the details. I'm\n",
      "awfully sorry for any inconvenience this fine print may cause you. It\n",
      "is rather vexing when expectations aren't met. If I may kindly ask for\n",
      "your patience, we will look into this posthaste. 🌹 #ATTAssist\n",
      "\n",
      "--------------------------------------------------\n",
      "\tComplaint: I swear  @ATT  is a freaking joke!! They are seriously going to credit\n",
      "our account literally $5!! FIVE FREAKING DOLLARS! $5 DOLLARS FOR BEING\n",
      "WITHOUT SERVICE FOR OVER 8 HOURS!!!\n",
      "\n",
      "\tResponse gpt-4-turbo-preview: Oh dear, I'm truly dismayed to hear about your recent ordeal with our\n",
      "service, how utterly inconvenient. A mere $5 seems hardly sufficient\n",
      "for such a prolonged disturbance. Could I kindly ask for your patience\n",
      "as we address this? Your satisfaction is paramount. 🌷@ATT\n",
      "\n",
      "--------------------------------------------------\n",
      "\tComplaint: @ATTNEWS  and  @ATT Its five bucks lol dont put yourself up too high\n",
      "on that horse\n",
      "\n",
      "\tResponse gpt-4-turbo-preview: Oh dear, I'm truly saddened to hear about your dissatisfaction. It's\n",
      "absolutely not our intention to seem lofty. Your concern over the\n",
      "matter is utterly valid. May I kindly ask for your patience as we\n",
      "strive to better our services? Warm regards, a friend at @AT&T 🌷\n",
      "#CustomerCare\n",
      "\n",
      "--------------------------------------------------\n"
     ]
    }
   ],
   "source": [
    "model = \"gpt-3.5-turbo\" #\"gpt-3.5-turbo\",\"gpt-4-turbo-preview\"\n",
    "for index, row in df_complaints.iloc[0:10].iterrows():\n",
    "  text = row.text.replace('\\n', ' ')\n",
    "  prompt = f'''The complaint is {text}'''\n",
    "  response = get_completion(prompt, instructions,\n",
    "                            client, model=model)\n",
    "  print(f\"\\tComplaint: {fill(text)}\\n\")\n",
    "  print(f\"\\tResponse {model}: {fill(response)}\\n\")\n",
    "  print(\"-\"*50)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DttZpiNiReTT"
   },
   "source": [
    "# Feedback Based Prompts\n",
    "\n",
    "We can write a feedback loop that lets the AI improve the quality of a base tweet so it is more engaging."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "kTmY9z4IxS5F"
   },
   "source": [
    "## Functions For Creating Prompts to Generate Tweet and Score Tweet\n",
    "\n",
    "We have functions to create to prompts and process the AI response.  Using these functions lets us automate the improvement process.  This is very advanced prompt engineering :)\n",
    "\n",
    "1. `create_prompt_iteration` will take the tweet `text` , the enagement `score` it received, and the `advice` on how to improve the tweet, and return a prompt that tells the AI to improve score of the tweet using the given advice.\n",
    "\n",
    "2. `create_prompt_score` takes the tweet `text` and returns a prompt that tells the AI to return the\n",
    "\n",
    "3. `extract_json_content` takes the AI's response and returns the score and advice as a JSON object."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "id": "piMFePvwSITS"
   },
   "outputs": [],
   "source": [
    "def create_prompt_iteration(text, score, advice):\n",
    "  prompt_iteration = f'''This tweet received an engagement\n",
    "  score of {score}/10 (score is between 0 and 10) for how\n",
    "  engaging it is.  To improve the score you can {advice}.\n",
    "  Write a new tweet that has a higher engagement score.\n",
    "  Tweet: {text}'''\n",
    "  return prompt_iteration\n",
    "\n",
    "def create_prompt_score(text):\n",
    "  prompt_score = f'''What engagement score would you give this tweet,\n",
    "  and why: {text}'''\n",
    "  return prompt_score\n",
    "\n",
    "def extract_json_content(input_string):\n",
    "  match = re.search(r'{.*}', input_string, re.DOTALL)  # Notice the re.DOTALL flag\n",
    "  if match:\n",
    "    return match.group(0)\n",
    "  else:\n",
    "    return None"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "0tRX1YTZ4jlG"
   },
   "source": [
    "## Instructions\n",
    "\n",
    "We have two sets of instructions.  \n",
    "\n",
    "1. `instructions_creator` = instructions that tell the AI how to create a tweet.\n",
    "\n",
    "2. `instructions_evaluator` = instructions that tell the tweet how to compute the engagement score of the tweet. It tells it to respond in a JSON string format.  We can then extract the score and advice from the AI's generated output."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "id": "Kne4evNnOmO-"
   },
   "outputs": [],
   "source": [
    "instructions_creator = '''You are going to help create content\n",
    "for a social media account of a Yale student who likes to\n",
    "criticize Harvard with funny insults, and the followers will\n",
    "engage with that type of biting content.  You will be asked to\n",
    "write clever tweets that  will get engagement.  Return only the\n",
    "text of the tweet.'''\n",
    "\n",
    "instructions_evaluator = ''' You will be asked to evaluate the\n",
    "engagement score of a tweet, where an engagement score of 0 means\n",
    "not engaging at all and receiving no likes, and an engagement score\n",
    "of 10 means very engaging and receiving a large number of likes.\n",
    "Note that engaging content has biting jokes at Harvards expense.\n",
    "You will provide the engagement score and some advice as to what\n",
    "can be done to improve the score.  Return your answer in JSON\n",
    "format with keys named score and advice.'''\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "CYNCZoQO6V6q"
   },
   "source": [
    "## Improve a Tweet\n",
    "\n",
    "`tweet0` is our initial tweet, which is probably very lame.  We will let the AI improve it iteratively using our prompts, instructions, and a `for` loop.  We will terminate after `niter_max` iterations.  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "LPp-nKLuR-sy",
    "outputId": "68fee386-4ece-4743-bb62-49b4eb904caf"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tweet: Yale is better than Harvard.\n",
      "Engagement score: 3\n",
      "Advice: To improve engagement, consider adding a humorous or insightful reason\n",
      "why Yale is considered better than Harvard. Adding a clever joke or a\n",
      "funny comparison could make the statement more engaging and likely to\n",
      "generate interactions.\n",
      "\n",
      "\n",
      "Tweet: \"Why do Harvard students avoid the library? Because Yale grads took\n",
      "all the good jobs and left them with nothing worth researching.\n",
      "#YalePride #HarvardWho?\"\n",
      "Engagement score: 8\n",
      "Advice: This tweet has a great mix of humor and rivalry which generally leads\n",
      "to high engagement, especially among those with ties to Yale or\n",
      "Harvard. To improve the score, consider including a relevant, eye-\n",
      "catching image or meme to complement the joke and potentially tagging\n",
      "or mentioning related accounts to increase visibility.\n",
      "\n",
      "\n",
      "Tweet: \"When Harvard students say they're going to an 'Ivy League' gathering,\n",
      "they're probably just attending a Yale alumni networking event.\n",
      "#HarvardDreamsYaleLeads #IvyLeagueIrony\"\n",
      "Engagement score: 8\n",
      "Advice: This tweet is quite engaging due to its clever use of rivalry between\n",
      "Harvard and Yale, which is a common theme that attracts attention\n",
      "within the context of Ivy League schools. To improve the score,\n",
      "consider adding a humorous meme or image related to the theme, as\n",
      "visual content tends to increase engagement. Furthermore, interacting\n",
      "with any replies to this tweet could also boost its visibility and\n",
      "engagement.\n",
      "\n",
      "\n",
      "Tweet: \"Just saw a group of Harvard students trying to cast a spell to open\n",
      "their library door. Guess they finally realized their acceptance\n",
      "letters were supposed to go to Hogwarts, not Harvard. #MagicOfMistakes\n",
      "#YaleRealityCheck\"\n",
      "Engagement score: 8\n",
      "Advice: To increase the engagement up to 10, consider adding a funny visual or\n",
      "meme related to the joke, enhancing the visual appeal and\n",
      "shareability. Additionally, engaging directly with comments could\n",
      "further amplify the tweet's reach.\n",
      "\n",
      "\n",
      "Best tweet score = 8\n",
      "Tweet 0: Yale is better than Harvard.\n",
      "Tweet final: \"Just saw a group of Harvard students trying to cast a spell to open\n",
      "their library door. Guess they finally realized their acceptance\n",
      "letters were supposed to go to Hogwarts, not Harvard. #MagicOfMistakes\n",
      "#YaleRealityCheck\"\n"
     ]
    }
   ],
   "source": [
    "tweet0 = 'Yale is better than Harvard.'\n",
    "\n",
    "model = 'gpt-4-turbo-preview' #[\"gpt-3.5-turbo\",\"gpt-4-turbo-preview\"]:\n",
    "niter_max = 4 #max number of iterations to improve tweet\n",
    "\n",
    "score_best =  0\n",
    "for i in range(niter_max):\n",
    "  if i ==0:\n",
    "    tweet = tweet0\n",
    "\n",
    "  prompt = create_prompt_score(tweet)\n",
    "  response = get_completion(prompt, instructions_evaluator,\n",
    "                            client, model=model)\n",
    "  response_json = json.loads(extract_json_content(response))\n",
    "  keys = list(response_json.keys())  # Get all keys from the JSON object\n",
    "  # Remove the known key 'score' to isolate the advice key\n",
    "  keys.remove('score')\n",
    "\n",
    "  # The remaining key is the one you're interested in\n",
    "  variable_key = keys[0]\n",
    "  score = response_json['score']\n",
    "  advice = response_json[keys[0]]\n",
    "\n",
    "  if score >= score_best:\n",
    "    score_best = score\n",
    "    tweet_best = tweet\n",
    "    print(f\"Tweet: {fill(tweet)}\")\n",
    "    print(f\"Engagement score: {score}\")\n",
    "    print(f\"Advice: {fill(advice)}\\n\\n\")\n",
    "\n",
    "  else:\n",
    "    print(\"No improvement\\n\")\n",
    "  prompt = create_prompt_iteration(tweet, score, advice)\n",
    "  tweet = get_completion(prompt, instructions_creator,\n",
    "                          client, model)\n",
    "\n",
    "print(f\"Best tweet score = {score_best}\")\n",
    "print(f\"Tweet 0: {tweet0}\\nTweet final: {fill(tweet_best)}\")"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "machine_shape": "hm",
   "provenance": [],
   "toc_visible": true
  },
  "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": "3.9.13"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "02e940a2f483465fb5ea927326db7d99": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_0b2b6964df504f19bf72ba19719c6634",
       "IPY_MODEL_451c156ab71e43b09f408fc474c100f9",
       "IPY_MODEL_ccbec1476b884ef4a44e49afd05cbf59"
      ],
      "layout": "IPY_MODEL_f827b976df134a01a2f6ee53de0579d1"
     }
    },
    "0b2b6964df504f19bf72ba19719c6634": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_ed32e2292ffc464f86cf56568e526824",
      "placeholder": "​",
      "style": "IPY_MODEL_26549a15d8d04768b7d9e8669474648a",
      "value": "special_tokens_map.json: 100%"
     }
    },
    "10114d0095b04789a815149bd16df6ad": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "1179c388c9bf49d5ae15a72dc650dccd": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "126bf522036b404eb05a092a8b62b83f": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "16f85359da49490bbbd4377c6700b347": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "1735c9834d5341bbb768434d233a8b04": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "1d4fca59cf084e84a5a019f54a422831": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "26549a15d8d04768b7d9e8669474648a": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "2e513a65b08a42159ee4359528086466": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_3128e85a3c404ebc961287c4c5d44db1",
      "placeholder": "​",
      "style": "IPY_MODEL_1179c388c9bf49d5ae15a72dc650dccd",
      "value": "pytorch_model.bin: 100%"
     }
    },
    "3128e85a3c404ebc961287c4c5d44db1": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "344dd750fdad44ba93cb118fd55d07c3": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "34704c8e7a95435ba8c7e49fff6d37b2": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "35ea0467e3b4437993e7d1a014c028c9": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "3b27a021ec5b40379e4564f10915fdbd": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "3cea6dee7f004bf8b7edc3889dadd634": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_344dd750fdad44ba93cb118fd55d07c3",
      "placeholder": "​",
      "style": "IPY_MODEL_1d4fca59cf084e84a5a019f54a422831",
      "value": " 953/953 [00:00&lt;00:00, 65.1kB/s]"
     }
    },
    "3e208aaa73d64dea8ad68cd90dd144cd": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "4320b038e717483cb498d4daee426bcc": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "451c156ab71e43b09f408fc474c100f9": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_35ea0467e3b4437993e7d1a014c028c9",
      "max": 112,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_beacf0da07694996868646b0d04ac4bd",
      "value": 112
     }
    },
    "4b09fcdfb2cb4a78a482cb966bf09aef": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "4f3145cad2344094afed2f56457cf9da": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_3e208aaa73d64dea8ad68cd90dd144cd",
      "max": 953,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_60f256dce0b14fe0a40e0213fda837ad",
      "value": 953
     }
    },
    "5ef22de727584eddad1ab1fb1a61d4f9": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "60f256dce0b14fe0a40e0213fda837ad": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "67ace7329a1842eeb8d77227a745f58d": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_db54e353c831452e95c5f2d840ed9d5d",
      "placeholder": "​",
      "style": "IPY_MODEL_6a8e3de110bb4113a7e68d7e50695bb2",
      "value": " 669M/669M [00:01&lt;00:00, 364MB/s]"
     }
    },
    "698b6e963a4f4a8d8e19fb1c92b91e69": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "6a74cc63c7bc48b5b14efe5eff8fc32a": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_3b27a021ec5b40379e4564f10915fdbd",
      "placeholder": "​",
      "style": "IPY_MODEL_7e3e5e42c9254aba9782a05e068e464e",
      "value": "tokenizer_config.json: 100%"
     }
    },
    "6a8e3de110bb4113a7e68d7e50695bb2": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "7229f52202144be9a4b3485a53bdfc46": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "76883a9b177c42efb403e8143ffb80c2": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_a7a7ba5a6bb5470eb60c8b17d44a6c5c",
      "placeholder": "​",
      "style": "IPY_MODEL_cdf1541e7e0e4b6ba4b8b58cb7be4648",
      "value": " 39.0/39.0 [00:00&lt;00:00, 3.17kB/s]"
     }
    },
    "76f6e8b033cd4977a50869af0843f904": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "7930eb5b7fef40f7ad9b8aabbec85e72": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_a62b5c66694b4b9198d053ed50a5ba34",
       "IPY_MODEL_4f3145cad2344094afed2f56457cf9da",
       "IPY_MODEL_3cea6dee7f004bf8b7edc3889dadd634"
      ],
      "layout": "IPY_MODEL_bec45abcd43d4854b43f39d3fcbf0c7b"
     }
    },
    "7e3e5e42c9254aba9782a05e068e464e": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "91ec3a4c6ebd408d978903761b94b391": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_126bf522036b404eb05a092a8b62b83f",
      "max": 871891,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_ffa49028f9424c1ba57c47d2ae7abf92",
      "value": 871891
     }
    },
    "9542cc8f340844238d2c99dac2ed8a27": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "9c9894c9b4f24b6ba7d098d8cb222373": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "a62b5c66694b4b9198d053ed50a5ba34": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_76f6e8b033cd4977a50869af0843f904",
      "placeholder": "​",
      "style": "IPY_MODEL_16f85359da49490bbbd4377c6700b347",
      "value": "config.json: 100%"
     }
    },
    "a7a7ba5a6bb5470eb60c8b17d44a6c5c": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "b09efca8d25144b4b39de727016efc28": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_10114d0095b04789a815149bd16df6ad",
      "max": 669491321,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_698b6e963a4f4a8d8e19fb1c92b91e69",
      "value": 669491321
     }
    },
    "ba5dc04dffae4d98a904d867b40b6617": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_6a74cc63c7bc48b5b14efe5eff8fc32a",
       "IPY_MODEL_cc027686dfe742deae943f7fc37a05ba",
       "IPY_MODEL_76883a9b177c42efb403e8143ffb80c2"
      ],
      "layout": "IPY_MODEL_5ef22de727584eddad1ab1fb1a61d4f9"
     }
    },
    "ba8bc3b86f75476e88756841e541a15e": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "beacf0da07694996868646b0d04ac4bd": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "bec45abcd43d4854b43f39d3fcbf0c7b": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "cc027686dfe742deae943f7fc37a05ba": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_34704c8e7a95435ba8c7e49fff6d37b2",
      "max": 39,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_cdb9065a6448412084872ca975058920",
      "value": 39
     }
    },
    "ccbec1476b884ef4a44e49afd05cbf59": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_7229f52202144be9a4b3485a53bdfc46",
      "placeholder": "​",
      "style": "IPY_MODEL_ba8bc3b86f75476e88756841e541a15e",
      "value": " 112/112 [00:00&lt;00:00, 8.09kB/s]"
     }
    },
    "cdb9065a6448412084872ca975058920": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "cdf1541e7e0e4b6ba4b8b58cb7be4648": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "d6dae3d5ecd546bc93e4c931c9031f7f": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_2e513a65b08a42159ee4359528086466",
       "IPY_MODEL_b09efca8d25144b4b39de727016efc28",
       "IPY_MODEL_67ace7329a1842eeb8d77227a745f58d"
      ],
      "layout": "IPY_MODEL_1735c9834d5341bbb768434d233a8b04"
     }
    },
    "db54e353c831452e95c5f2d840ed9d5d": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "dbc68ac3016243fdbfb58608d0266db8": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_f884da40323a4e3a9f3775627e1595c5",
      "placeholder": "​",
      "style": "IPY_MODEL_4b09fcdfb2cb4a78a482cb966bf09aef",
      "value": " 872k/872k [00:00&lt;00:00, 33.8MB/s]"
     }
    },
    "df272bc68f7f44a4be8185332f81932f": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_9542cc8f340844238d2c99dac2ed8a27",
      "placeholder": "​",
      "style": "IPY_MODEL_9c9894c9b4f24b6ba7d098d8cb222373",
      "value": "vocab.txt: 100%"
     }
    },
    "ed32e2292ffc464f86cf56568e526824": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "f60193cbcb8641b782eefed48bdd299c": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_df272bc68f7f44a4be8185332f81932f",
       "IPY_MODEL_91ec3a4c6ebd408d978903761b94b391",
       "IPY_MODEL_dbc68ac3016243fdbfb58608d0266db8"
      ],
      "layout": "IPY_MODEL_4320b038e717483cb498d4daee426bcc"
     }
    },
    "f827b976df134a01a2f6ee53de0579d1": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "f884da40323a4e3a9f3775627e1595c5": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "ffa49028f9424c1ba57c47d2ae7abf92": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}