CLIP_IPYNB.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0iD9FACHWYvR",
        "outputId": "61d81555-a59a-42c5-b879-fa7205ef38da"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Requirement already satisfied: kaggle in /usr/local/lib/python3.10/dist-packages (1.5.16)\n",
            "Requirement already satisfied: six>=1.10 in /usr/local/lib/python3.10/dist-packages (from kaggle) (1.16.0)\n",
            "Requirement already satisfied: certifi in /usr/local/lib/python3.10/dist-packages (from kaggle) (2023.7.22)\n",
            "Requirement already satisfied: python-dateutil in /usr/local/lib/python3.10/dist-packages (from kaggle) (2.8.2)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from kaggle) (2.27.1)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.10/dist-packages (from kaggle) (4.65.0)\n",
            "Requirement already satisfied: python-slugify in /usr/local/lib/python3.10/dist-packages (from kaggle) (8.0.1)\n",
            "Requirement already satisfied: urllib3 in /usr/local/lib/python3.10/dist-packages (from kaggle) (1.26.16)\n",
            "Requirement already satisfied: bleach in /usr/local/lib/python3.10/dist-packages (from kaggle) (6.0.0)\n",
            "Requirement already satisfied: webencodings in /usr/local/lib/python3.10/dist-packages (from bleach->kaggle) (0.5.1)\n",
            "Requirement already satisfied: text-unidecode>=1.3 in /usr/local/lib/python3.10/dist-packages (from python-slugify->kaggle) (1.3)\n",
            "Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.10/dist-packages (from requests->kaggle) (2.0.12)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.10/dist-packages (from requests->kaggle) (3.4)\n"
          ]
        }
      ],
      "source": [
        "# importing libraries\n",
        "!pip install kaggle\n",
        "from google.colab import files\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import os\n",
        "\n",
        "import requests\n",
        "from PIL import Image\n",
        "from io import BytesIO\n",
        "\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "from torchvision import transforms\n",
        "import torch.nn.functional as F\n",
        "import torch.optim as optim\n",
        "\n",
        "\n",
        "import matplotlib.pyplot as plt"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7eCrzNM4holu",
        "outputId": "e205238f-591b-41a6-980c-be6c8004b404"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n",
            "cp: cannot create regular file '/root/.kaggle/kaggle.json': No such file or directory\n",
            "chmod: cannot access '/root/.kaggle/kaggle.json': No such file or directory\n",
            "Downloading guie-laion5b-dataset.zip to /content\n",
            " 51% 25.0M/49.4M [00:00<00:00, 72.2MB/s]\n",
            "100% 49.4M/49.4M [00:00<00:00, 114MB/s] \n",
            "Archive:  guie-laion5b-dataset.zip\n",
            "  inflating: GUIE_laion5b_dataset_en.json  \n"
          ]
        }
      ],
      "source": [
        "#Bringing the dataset\n",
        "from google.colab import drive\n",
        "drive.mount('/content/drive')\n",
        "#! mkdir ~/.kaggle\n",
        "os.environ['KAGGLE_USERNAME'] = \"hanviksuruboyina\"\n",
        "os.environ['KAGGLE_KEY'] = \"b8f48ef46400e2f5a559ada21f2cc402\"\n",
        "!cp /content/drive/MyDrive/ML/kaggle.json ~/.kaggle/kaggle.json\n",
        "! chmod 600 ~/.kaggle/kaggle.json\n",
        "! kaggle datasets download vitaliykinakh/guie-laion5b-dataset\n",
        "! unzip guie-laion5b-dataset.zip"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "5rNdwWoQj-Ax",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d89a1116-e008-496f-9891-700cb0e8daa1"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "                                             caption  \\\n",
            "0                                         gladiatore   \n",
            "1  Gladiator is listed (or ranked) 4 on the list ...   \n",
            "2             are-you-not-entertained-w-text-720x396   \n",
            "3  \"Russell Crowe em \"\"Gladiador\"\" - ©Dreamworks/...   \n",
            "4  Portrait of smiling boy in the temple of Apoll...   \n",
            "\n",
            "                                                 url          id  similarity  \\\n",
            "0  https://libreriamo.it/wp-content/uploads/2020/...  4748272869    0.516147   \n",
            "1  https://imgix.ranker.com/user_node_img/54/1075...  2784821276    0.513879   \n",
            "2  https://traditionalpolytheist.files.wordpress....  2386885741    0.508544   \n",
            "3  https://conteudo.imguol.com.br/c/entreteniment...   841845904    0.507340   \n",
            "4  https://thumbs.dreamstime.com/b/portrait-smili...  1892934610    0.506108   \n",
            "\n",
            "                            prompt category  \\\n",
            "0  meme \"Are you not entertained?\"     meme   \n",
            "1  meme \"Are you not entertained?\"     meme   \n",
            "2  meme \"Are you not entertained?\"     meme   \n",
            "3  meme \"Are you not entertained?\"     meme   \n",
            "4  meme \"Are you not entertained?\"     meme   \n",
            "\n",
            "                                          caption_en  \n",
            "0                                          gladiator  \n",
            "1  Gladiator is listed (or ranked) 4 on the list ...  \n",
            "2             are-you-not-entertained-w-text-720x396  \n",
            "3  \"Russell Crowe em \"\"Gladiador\"\" - ©Dreamworks/...  \n",
            "4  Portrait of smiling boy in the temple of Apoll...  \n"
          ]
        }
      ],
      "source": [
        "#json to csv\n",
        "data_json = pd.read_json('GUIE_laion5b_dataset_en.json')\n",
        "data_json.to_csv('data.csv', index=False)\n",
        "data = pd.read_csv('data.csv')\n",
        "print(data.head(5))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "GL0rPhgZiVap",
        "outputId": "491c3d8a-ba5f-462d-cb64-717f01215a89"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "['meme', 'toys', 'furniture', 'car', 'food and packaged goods', 'clothes', 'storefronts', 'paintings', 'landmark']\n",
            "['meme \"Are you not entertained?\"', 'meme Scumbag Steve', 'meme Demotivational posters', 'meme \"All Your Base Are Belong to Us\"', 'meme Chuck Norris Facts', 'meme \"Thanks, Obama\"', 'meme Success Kid', 'meme \"Fuck it, we\\'ll do it live!\"', 'Keep Calm and Carry On', 'meme Numa Numa', 'meme The Most Interesting Man', 'meme \"Khaaaaaaaaaan!!!!\"', 'meme \"Imma let you finish\"', 'meme Michael Jackson eating popcorn', 'meme Harambe', 'meme Keyboard Cat', 'meme James Van Der Beek crying', 'meme Ceiling Cat', 'meme #Winning', 'meme woman yelling at a cat', 'meme \"Netflix and chill\"', 'meme pepe frog', 'meme \"Gangnam Style\"', 'meme Shirtless Putin', 'meme Kermit', 'meme Doge', 'meme Smash Mouth\\'s \"All Star\"', 'meme Distracted Boyfriend', 'meme Nigerian prince', 'meme The Honey Badger', 'meme trollface', 'meme Bert Is Evil', 'meme Trollface', 'meme LOLcats', 'meme Keep Calm', 'meme OBEY', 'meme \"That\\'s what she said\"', 'meme \"More cowbell!\"', 'meme Rebecca Black\\'s \"Friday\"', 'meme Derp', 'meme Dat Boi', 'meme Sad Keanu', 'meme \"Hey girl\"Sensitive Gosling', 'meme Rickrolling', 'meme Ancient Aliens', 'meme \"Don\\'t tase me, bro!\"', 'meme Grumpy Cat (RIP)', 'meme Condescending Sarcastic Willy Wonka', 'meme American Chopper argument', 'meme Dabbing', 'meme #FirstWorldProblems', 'meme Ermahgerd', 'meme Philosoraptor', 'meme Obama\\'s \"Hope\" poster', 'meme Double rainbow', 'meme Nyan Cat', 'meme \"Bye, Felicia\"', 'memes', 'memes pepe frog', 'meme Crying Michael Jordan', 'toy doll', 'toy ball', 'toy barbie', 'basketball ball', 'toy football ball', 'toy car', 'toys', 'toys blocks lego', 'toy train', 'toys blocks', 'toy football ', 'board game', 'toy teddy bear', 'toy boat', 'toy plane', 'box of toys', 'toy musical instruments', 'bookcase', 'Attic home decor', 'Drawer furniture', 'dining room', 'shelf', 'sink', 'bench furniture', 'pillow', 'window', 'Curtains furniture', 'Coffeemaker', 'kitchen', 'armchair', 'mirror furniture', 'carpet furniture', 'kitchen oven', 'washing machine', 'refrigerator', 'Cabinet furniture', 'living room', 'stool furniture', 'sheets', 'couch', 'kettle', 'chest furniture', 'lamp furniture', 'bed furniture', 'chair furniture', 'Shower bathroom', 'bedroom', 'desk furniture', 'bathtub', 'Toyota RAV4', 'car Toyota Tacoma', 'car Toyota RAV4 2020', 'car Chevrolet Silverado', 'car Ford F-150 2020', 'car Ram 1500', 'car GMC Sierra', 'car Toyota Tacoma 2020', 'car Chevrolet Silverado 2020', 'car Ford F-Series', 'car Ram 1500 2020', 'car Honda Civic', 'car Toyota Camry', 'car GMC Sierra 2020', 'car  Toyota Highlander', 'cartoon of juice', 'fried chicken', 'multipack of snacks', 'box of fruits', 'pizza', 'can of fanta', 'blister pack of medicine', 'can of pepsi', 'multipack of yogurts activia', 'can of soda', 'a bottle of milk', 'cup of tea', 'a jug of water', 'barbecue', 'kebab', 'fish disches', 'crepes', 'jar of honey', 'bar of snikers', 'bar of soap', 'piece of cheese', 'soup', 'pack of cigarettes marlboro', 'Multipack', 'pack of cigarettes lucky strike', 'sandwich', 'bottle of dr pepper', 'bottle of beer coors', 'shopping cart', 'box of candies', 'bottle of vine', 'irish breakfast', 'packet', 'pack of cigarettes winston', 'french fries', 'can of beer', 'bottle of beer budweiser', 'bottle of water', 'blister pack of candies', 'fried egg', 'hamburger', 'pack of cigarettes camel', 'sausages', 'bottle of beer corona extra', 'slice of bread', 'pack of cigarettes', 'bottle of sprite', 'box of napkins', 'ribs', 'can of beans', 'sushi', 'bottle of Coke', 'hotdog', 'bar of kitkat', 'box of food', 'a glass of water', 'packet of clothes', 'multipack of cookies', 'shopping basket', 'chicken nuggets', 'Blister pack', 'cheving gum opbit', 'multipack of yogurts', 'wrap', 'carton of milk', 'steak', 'Matchbox', 'bar of chocolate', 'rice', 'loaf of bread', 'cereals', 'bar of nuts chocolate', 'box ', 'can of Coke', 'bottle of fanta', 'blister pack of snaks', 'multipack of candies', 'can of tuna', 'bottle of pepsi', 'multipack of fruits', 'spaghetti', 'salad', 'can of sprite', 'multipack of chockolates', 'packet of food', 'multipack of chips', \"women's Necklace\", \"children's panties\", 'scarf', \"men's jumper\", \"men's socks_1\", \"women's shorts\", \"children's trousers\", \"men's coat\", \"men's derby hat\", \"children's stockings\", \"men's tee-shirt\", 'shirt', \"children's boots\", 'mittens', 'suit', 'socks_', 'Sneakers female', 'Hair band', \"women's ring\", \"children's umbrella\", \"men's socks\", \"women's purse\", 'jumper', \"children's suit\", \"women's earings\", 'accessories2', \"men's shoes\", 'female jumper', 'cap', 'children t-shirt', \"men's apron\", \"children's bra\", \"men's purse\", 'male jumper', \"women's gloves\", \"women's umbrella\", 'shoes', \"children's jeans\", \"children's shoes\", \"women's jacket\", 'running shoes', 'umbrella', 'sports suit', \"men's suit\", \"men's straw hat\", \"men's shirt\", 'sweater', 'box of clothes', \"children's skirt\", \"children's scarf\", 'child Skirt', \"men's Necklace\", \"women's trainers\", 'Pocket watch', 'hat', \"children's chinos\", 'Hoodie', 'vase', 'helmet', 'tee-shirt', \"men's ring\", 'trousers', 'male vest', 'Beret', \"women's Sunglasses\", \"children's trainers\", 'derby hat', 'jeans', \"women's trousers\", 'tie', \"children's hat\", 'jumper2', \"women's boots\", \"children's Necklace\", \"men's stockings\", 'beanie', \"men's sneakers\", \"men's jeans\", \"men's dress\", \"men's watch\", 'shorts', 'dress', \"women's mittens\", \"children's Watch\", 'male Hoodie', \"women's cap\", \"children's jacket\", \"men's cap\", 'earings', \"women's chinos\", 'male t-shirt', \"women's helmet\", \"children's earings\", 'coat', 'female t-shirt', \"women's bra\", \"men's shorts\", \"men's leather jacket\", \"men's jacket\", 'Necklace', 'female Skirt', \"women's jumper\", \"men's helmet\", 'apparel_and_accessories', \"children's mittens\", 'chinos', \"children's shirt\", 'vest', 'parts', 'Bikini', 'hoodie', 'child jumper', \"women's shoes\", \"children's pants\", \"men's pants\", 'apparel', \"women's hoodie\", 'Sneakers', 'female Hoodie', \"woman's dress\", 'Safety pin', \"children's dress\", \"women's earmuffs\", \"women's beanie\", \"women's jeans\", \"men's trainers\", \"children's shorts\", 'Sunglasses', \"women's sneakers\", \"men's skirt\", \"men's beret\", \"men's trousers\", \"women's panties\", 'gloves', 'helmet_', \"men's mittens\", 'ring', \"men's boots\", 'bowtie', 'trainers', 'female vest', \"men's gloves\", \"women's socks\", \"men's accessories\", 'child Hoodie', 'bobby pin', 'leather jacket', \"men's bra\", \"women's hat\", 'jacket', \"men's earings\", \"men's hat\", \"children's jumper\", \"children's Sunglasses\", \"children's socks\", \"women's tie\", \"men's Sunglasses\", 'purse', \"women's apron\", 'boots', \"children's tie\", \"children's earmuffs\", \"men's beanie\", \"children's gloves\", 'Sneakers male', \"women's beret\", \"men's panties\", \"women's leather jacket\", \"women's scarf\", \"men's earmuffs\", 'Apron', \"children's cap\", \"women's straw hat\", 'accessories', 'clothes', \"men's scarf\", 'straw hat', \"men's hoodie\", 'necktie', \"children's apron\", \"women's tee-shirt\", 'hoodie_kids', \"men's umbrella\", \"children's purse\", 'female Bikini', \"women's shirt\", \"children's coat\", 'Watch', 'Earmuffs', \"women's skirt\", \"women's pants\", \"children's hoodie\", \"women's derby hat\", 'Stockings', 'skirt', 'bra', \"women's Watch\", 'panties', \"women's suit\", 'Skirt', \"children's sneakers\", 'sneakers', 'child vest', \"women's coat\", 't-shirt', \"women's accessories\", \"men's tie\", \"men's chinos\", 'socks', 'storefront auchan', 'storefronts Lidl', 'storefront Aldi', 'storefront cafe', 'storefronts photos', 'storefront wolmart', 'Las Meninas Diego Velazquez painting', 'The Happy Accidents Of The Swing Jean-Honore Fragonard painting', 'The Birth Of Venus Alexandre Cabanel painting', 'The Wanderer Above The Sea Of Fog Caspar David Friedrich picture', 'mona lisa', 'artwork', 'Olympia Edouard Manet picture', 'the girl with a pearl earring painting', 'Luncheon Of The Boating Party Pierre-Auguste Renoir painting', 'the arnolfini portrait painting', 'American Gothic Grant Wood painting', 'A Cotton Office In New Orleans Edgar Degas picture', 'A Friend In Need Cassius Marcellus Coolidge painting', 'the scream artwork', 'Impression Sunrise Claude Monet painting', 'The Birth Of Venus Sandro Botticelli painting', 'The Third Of May 1808 Francisco Goya picture', 'The Creation Of Adam Michelangelo painting', 'Charles I In Three Positions Anthony van Dyck picture', 'The Grand Odalisque Jean Auguste Dominique Ingres painting', 'the starry night', 'the kiss painting', 'The Gleaners Jean-Francois Millet picture', 'Primavera Sandro Botticelli picture', 'The Last Supper Leonardo Da Vinci painting', 'The Liberty Leading The People Eugene Delacroix painting', 'The Triumph Of Galatea Raphael painting', 'The Garden Of Earthly Delights - Central Panel Hieronymus Bosch painting', 'View Of Toledo El Greco picture', 'Pollice Verso Jean-Leon Gerome painting', 'Bacchus And Ariadne Titian picture', 'Musicians Caravaggio painting', 'The Lady With The Ermine Leonardo Da Vinci painting', 'Napoleon Crossing The Alps Jacques-Louis David painting', 'The Tower Of Babel Pieter Bruegel the Elder picture', 'The Sleeping Gypsy Henri Rousseau painting', 'the night watch artwork', 'Sunday Afternoon On The Island Of La Grande Jatte Georges Seurat painting', 'landmark Golden Gate Bridge', 'landmark Bondi Beach', 'landmark Peggy’s Point Lighthouse', 'landmark The Palace of Versailles', 'toilet', 'landmark St Paul’s Cathedral', 'landmark Niagara Falls', 'landmark Cliffs of Moher', 'landmark The Shard', 'landmark Spanish Steps', 'landmark Pompeii', 'landmark The Demilitarised Zone Korea', 'landmark La Sagrada Familia', 'landmark Mount Rushmore', 'landmark Taj Mahal', 'landmark 30 St Mary Axe or The Gherkin', 'landmark Buckingham Palace', 'landmark Guggenheim Bilbao', 'landmark The Alhambra', 'landmark Burj Khalifa', 'landmark Gardens by the Bay', 'landmark Potala Palace', 'landmark The Bund', 'landmark Wembley Stadium', 'new york Statue of Liberty', 'landmark Berlin TV Tower', 'landmark The O2', 'landmark Notre Dame de Paris', 'landmark Mount Fuji', 'landmark Giant’s Causeway', 'landmark The Eiffel Tower, France', 'landmark CN Tower', 'landmark St Peter’s Basilica', 'landmark Westminster Abbey', 'landmark Seoul Tower', 'landmark Chichen Itza', 'landmark Itsukushima Shrine', 'landmark The Twelve Apostles', 'landmark Tower of London', 'landmark Petronas Towers', 'landmark Pena National Palace', 'landmark The Great Wall of China', 'landmark Marina Bay Sands', 'landmark Machu Picchu', 'landmark Hagia Sophia of Thessalonica Church', 'landmark Christ the Redeemer Statue', 'landmark Big Ben', 'landmark One Canada Square', 'landmark Bungle Bungle Range', 'landmark Leaning Tower of Pisa', 'landmark Titanic Museum and Quarter', 'landmark Hoover Dam', 'landmark Seattle Space Needle', 'landmark Arc de Triomphe', 'landmark Halong Bay', 'landmark Stonehenge', 'landmark Fairmont Le Château Frontenac', 'landmark The Berlin Wall', 'landmark Petra, Jordan', 'landmark The Forbidden City', 'landmark Empire State Building', 'landmark The Colosseum, Italy', 'landmark Tian Tan Buddha', 'landmark Sydney Harbour Bridge', 'landmark Sacré-Cœur Basilica', 'landmark Washington Monument', 'landmark Mexico City Cathedral', 'landmark Las Rambla', 'landmark  Great Pyramid of Cholula', 'landmark The London Eye', 'eiffel tower', 'landmark Shwedagon Pagoda', 'landmark Sydney Opera House', 'landmark Statue of Liberty', 'Big Ben london', 'landmark Rialto Bridge', 'landmark Saint Mark’s Basilica', 'landmark Niagara Falls_', 'landmark The Trevi Fountain', 'Gyeongbokgung and Changgyeon Palaces', 'landmark The Grand Canyon', 'landmark Hiroshima Peace Memorial Park', 'landmark The Terracotta Army Museum', 'landmark Pyramids of Giza', 'landmark Angkor Wat', 'landmark The Hollywood Sign', 'landmark Taipei 101', 'landmark The Great Mosque of Cordoba', 'landmark Ring of Kerry', 'Mont Saint-Michel', 'landmark Brandenburg Gate', 'landmark Dom Luis Bridge', 'landmark Windsor Castle', 'landmark The Louvre Museum', 'landmark Uluru', 'landmark The Imperial Palace', 'landmark Tower Bridge', 'landmark Banff National Park', 'landmark Neuschwanstein Castle', 'landmark Great Barrier Reef', 'landmark The Acropolis', 'landmark Mount Everest']\n",
            "61556\n",
            "212274\n",
            "1260\n",
            "2200\n",
            "61556\n",
            "5874\n",
            "14975\n",
            "377\n",
            "1667\n",
            "197909\n",
            "['meme', 'toys', 'furniture', 'car', 'food and packaged goods', 'clothes', 'storefronts', 'paintings', 'landmark']\n"
          ]
        }
      ],
      "source": [
        "# to find the names of categories\n",
        "def categories(category):\n",
        "  column_values = data[category]\n",
        "  unique_values = column_values.unique()\n",
        "  unique_column_names = [col.split(':')[0] for col in unique_values]\n",
        "  #print(unique_column_names)\n",
        "  #print(len(unique_column_names))\n",
        "  return unique_column_names\n",
        "\n",
        "print(categories('category'))\n",
        "print(categories('prompt'))\n",
        "\n",
        "def number_rows_category(word, category):\n",
        "  matching_rows = data[data[category].str.contains(word, case=False)]\n",
        "  num_rows = matching_rows.shape[0]\n",
        "  print(num_rows)\n",
        "\n",
        "number_rows_category('car', 'category')\n",
        "\n",
        "for i in range (0,9):\n",
        "  number_rows_category(categories('category')[i], 'category')\n",
        "print(categories('category'))"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "gsV2EISz1zIj"
      },
      "outputs": [],
      "source": [
        "data = data.drop('id', axis=1) # to remove the id column as it is not useful\n",
        "data = data.drop('caption', axis=1) # to remove the caption column as it is not useful\n",
        "data = data.drop('caption_en', axis=1)\n",
        "data = data.drop('similarity', axis=1)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 423
        },
        "id": "Yae65f54ATcv",
        "outputId": "d8dca3ea-7719-423d-b1c7-f76daf360773"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                                      url  \\\n",
              "0       https://libreriamo.it/wp-content/uploads/2020/...   \n",
              "1       https://imgix.ranker.com/user_node_img/54/1075...   \n",
              "2       https://traditionalpolytheist.files.wordpress....   \n",
              "3       https://conteudo.imguol.com.br/c/entreteniment...   \n",
              "4       https://thumbs.dreamstime.com/b/portrait-smili...   \n",
              "...                                                   ...   \n",
              "498087  https://static2.tripoto.com/media/filter/nt/im...   \n",
              "498088  https://images.pond5.com/peaks-near-gorak-shep...   \n",
              "498089  https://i1.vpoxod.ru/route-photo/38/c7/54/fb/4...   \n",
              "498090  https://images.centrometeoitaliano.it/wp-conte...   \n",
              "498091  https://i2.wp.com/www.vidyudabhi.com/wp-conten...   \n",
              "\n",
              "                                 prompt  category  \n",
              "0       meme \"Are you not entertained?\"      meme  \n",
              "1       meme \"Are you not entertained?\"      meme  \n",
              "2       meme \"Are you not entertained?\"      meme  \n",
              "3       meme \"Are you not entertained?\"      meme  \n",
              "4       meme \"Are you not entertained?\"      meme  \n",
              "...                                 ...       ...  \n",
              "498087           landmark Mount Everest  landmark  \n",
              "498088           landmark Mount Everest  landmark  \n",
              "498089           landmark Mount Everest  landmark  \n",
              "498090           landmark Mount Everest  landmark  \n",
              "498091           landmark Mount Everest  landmark  \n",
              "\n",
              "[498092 rows x 3 columns]"
            ],
            "text/html": [
              "\n",
              "\n",
              "  <div id=\"df-bf9496e2-35c6-445e-87e3-a390c78dd2dd\">\n",
              "    <div 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>url</th>\n",
              "      <th>prompt</th>\n",
              "      <th>category</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>https://libreriamo.it/wp-content/uploads/2020/...</td>\n",
              "      <td>meme \"Are you not entertained?\"</td>\n",
              "      <td>meme</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>https://imgix.ranker.com/user_node_img/54/1075...</td>\n",
              "      <td>meme \"Are you not entertained?\"</td>\n",
              "      <td>meme</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>https://traditionalpolytheist.files.wordpress....</td>\n",
              "      <td>meme \"Are you not entertained?\"</td>\n",
              "      <td>meme</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>https://conteudo.imguol.com.br/c/entreteniment...</td>\n",
              "      <td>meme \"Are you not entertained?\"</td>\n",
              "      <td>meme</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>https://thumbs.dreamstime.com/b/portrait-smili...</td>\n",
              "      <td>meme \"Are you not entertained?\"</td>\n",
              "      <td>meme</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>498087</th>\n",
              "      <td>https://static2.tripoto.com/media/filter/nt/im...</td>\n",
              "      <td>landmark Mount Everest</td>\n",
              "      <td>landmark</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>498088</th>\n",
              "      <td>https://images.pond5.com/peaks-near-gorak-shep...</td>\n",
              "      <td>landmark Mount Everest</td>\n",
              "      <td>landmark</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>498089</th>\n",
              "      <td>https://i1.vpoxod.ru/route-photo/38/c7/54/fb/4...</td>\n",
              "      <td>landmark Mount Everest</td>\n",
              "      <td>landmark</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>498090</th>\n",
              "      <td>https://images.centrometeoitaliano.it/wp-conte...</td>\n",
              "      <td>landmark Mount Everest</td>\n",
              "      <td>landmark</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>498091</th>\n",
              "      <td>https://i2.wp.com/www.vidyudabhi.com/wp-conten...</td>\n",
              "      <td>landmark Mount Everest</td>\n",
              "      <td>landmark</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>498092 rows × 3 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-bf9496e2-35c6-445e-87e3-a390c78dd2dd')\"\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 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "\n",
              "\n",
              "\n",
              "    <div id=\"df-69c2e4d5-52f6-484a-af87-e3995f303ae7\">\n",
              "      <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-69c2e4d5-52f6-484a-af87-e3995f303ae7')\"\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",
              "    </div>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\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-quickchart: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",
              "  [theme=dark] .colab-df-quickchart {\n",
              "    background-color: #3B4455;\n",
              "    fill: #D2E3FC;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart: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",
              "      async function quickchart(key) {\n",
              "        const containerElement = document.querySelector('#' + key);\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      }\n",
              "    </script>\n",
              "\n",
              "      <script>\n",
              "\n",
              "function displayQuickchartButton(domScope) {\n",
              "  let quickchartButtonEl =\n",
              "    domScope.querySelector('#df-69c2e4d5-52f6-484a-af87-e3995f303ae7 button.colab-df-quickchart');\n",
              "  quickchartButtonEl.style.display =\n",
              "    google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "}\n",
              "\n",
              "        displayQuickchartButton(document);\n",
              "      </script>\n",
              "      <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\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",
              "    [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-bf9496e2-35c6-445e-87e3-a390c78dd2dd 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-bf9496e2-35c6-445e-87e3-a390c78dd2dd');\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",
              "  </div>\n"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ],
      "source": [
        "data"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "CwNdwcG4Or6C"
      },
      "outputs": [],
      "source": [
        "df = pd.DataFrame()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "-ku2S62mDH5e",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2b7a1133-ef82-44f1-8462-cc5e36b9ba53"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "0\n",
            "1\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "<ipython-input-8-022ee89892df>:6: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
            "  df = df.append(items_with_specific_label)\n",
            "<ipython-input-8-022ee89892df>:6: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
            "  df = df.append(items_with_specific_label)\n",
            "<ipython-input-8-022ee89892df>:6: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
            "  df = df.append(items_with_specific_label)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2\n",
            "3\n",
            "4\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "<ipython-input-8-022ee89892df>:6: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
            "  df = df.append(items_with_specific_label)\n",
            "<ipython-input-8-022ee89892df>:6: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
            "  df = df.append(items_with_specific_label)\n",
            "<ipython-input-8-022ee89892df>:6: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
            "  df = df.append(items_with_specific_label)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "5\n",
            "6\n",
            "7\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "<ipython-input-8-022ee89892df>:6: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
            "  df = df.append(items_with_specific_label)\n",
            "<ipython-input-8-022ee89892df>:6: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
            "  df = df.append(items_with_specific_label)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "8\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "<ipython-input-8-022ee89892df>:6: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n",
            "  df = df.append(items_with_specific_label)\n"
          ]
        }
      ],
      "source": [
        "for i in range (len(categories('category'))):\n",
        "  print(i)\n",
        "  items_with_specific_label = data[data['category'] == categories('category')[i]]\n",
        "  items_with_specific_label = items_with_specific_label.sample(frac=1, random_state=42)\n",
        "  items_with_specific_label = items_with_specific_label.head(10)\n",
        "  df = df.append(items_with_specific_label)\n",
        "df = df.reset_index(drop=True)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 423
        },
        "id": "e4LAEz6DdJBL",
        "outputId": "690c0e0e-a886-45ac-b873-60aa1ddeeeef"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                                  url  \\\n",
              "0   https://static.wixstatic.com/media/039af8_8b6f...   \n",
              "1                    https://i.imgflip.com/2vo3mr.jpg   \n",
              "2   https://wiki.blogs.com/photos/uncategorized/20...   \n",
              "3   https://i.pinimg.com/200x150/36/c9/d5/36c9d5bb...   \n",
              "4   http://cdn.salemweb.net/godtube/2011/09/16/337...   \n",
              "..                                                ...   \n",
              "85  https://freedomdestinations.co.uk/wp-content/u...   \n",
              "86  https://cdn.vox-cdn.com/thumbor/ZUQmkwGVccNDMY...   \n",
              "87  https://pic.pimg.tw/jcp1223/1387283539-2192754...   \n",
              "88  https://thumbs.dreamstime.com/t/%D0%B1%D0%B0%D...   \n",
              "89  https://cdn.shopify.com/s/files/1/0071/9136/36...   \n",
              "\n",
              "                                    prompt  category  \n",
              "0                        meme Bert Is Evil      meme  \n",
              "1                 meme #FirstWorldProblems      meme  \n",
              "2                 meme \"Khaaaaaaaaaan!!!!\"      meme  \n",
              "3                           meme Keep Calm      meme  \n",
              "4        meme \"Fuck it, we'll do it live!\"      meme  \n",
              "..                                     ...       ...  \n",
              "85  landmark Fairmont Le Château Frontenac  landmark  \n",
              "86          landmark Mexico City Cathedral  landmark  \n",
              "87                     landmark Taipei 101  landmark  \n",
              "88                       landmark CN Tower  landmark  \n",
              "89              landmark Westminster Abbey  landmark  \n",
              "\n",
              "[90 rows x 3 columns]"
            ],
            "text/html": [
              "\n",
              "\n",
              "  <div id=\"df-37f6b6fc-2b74-471a-9242-adf9f62d73a3\">\n",
              "    <div 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>url</th>\n",
              "      <th>prompt</th>\n",
              "      <th>category</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>https://static.wixstatic.com/media/039af8_8b6f...</td>\n",
              "      <td>meme Bert Is Evil</td>\n",
              "      <td>meme</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>https://i.imgflip.com/2vo3mr.jpg</td>\n",
              "      <td>meme #FirstWorldProblems</td>\n",
              "      <td>meme</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>https://wiki.blogs.com/photos/uncategorized/20...</td>\n",
              "      <td>meme \"Khaaaaaaaaaan!!!!\"</td>\n",
              "      <td>meme</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>https://i.pinimg.com/200x150/36/c9/d5/36c9d5bb...</td>\n",
              "      <td>meme Keep Calm</td>\n",
              "      <td>meme</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>http://cdn.salemweb.net/godtube/2011/09/16/337...</td>\n",
              "      <td>meme \"Fuck it, we'll do it live!\"</td>\n",
              "      <td>meme</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>85</th>\n",
              "      <td>https://freedomdestinations.co.uk/wp-content/u...</td>\n",
              "      <td>landmark Fairmont Le Château Frontenac</td>\n",
              "      <td>landmark</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>86</th>\n",
              "      <td>https://cdn.vox-cdn.com/thumbor/ZUQmkwGVccNDMY...</td>\n",
              "      <td>landmark Mexico City Cathedral</td>\n",
              "      <td>landmark</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>87</th>\n",
              "      <td>https://pic.pimg.tw/jcp1223/1387283539-2192754...</td>\n",
              "      <td>landmark Taipei 101</td>\n",
              "      <td>landmark</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>88</th>\n",
              "      <td>https://thumbs.dreamstime.com/t/%D0%B1%D0%B0%D...</td>\n",
              "      <td>landmark CN Tower</td>\n",
              "      <td>landmark</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>89</th>\n",
              "      <td>https://cdn.shopify.com/s/files/1/0071/9136/36...</td>\n",
              "      <td>landmark Westminster Abbey</td>\n",
              "      <td>landmark</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>90 rows × 3 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-37f6b6fc-2b74-471a-9242-adf9f62d73a3')\"\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 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "\n",
              "\n",
              "\n",
              "    <div id=\"df-f3016a25-8c6a-4d01-9fc5-264dd335eb21\">\n",
              "      <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-f3016a25-8c6a-4d01-9fc5-264dd335eb21')\"\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",
              "    </div>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\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-quickchart: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",
              "  [theme=dark] .colab-df-quickchart {\n",
              "    background-color: #3B4455;\n",
              "    fill: #D2E3FC;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart: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",
              "      async function quickchart(key) {\n",
              "        const containerElement = document.querySelector('#' + key);\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      }\n",
              "    </script>\n",
              "\n",
              "      <script>\n",
              "\n",
              "function displayQuickchartButton(domScope) {\n",
              "  let quickchartButtonEl =\n",
              "    domScope.querySelector('#df-f3016a25-8c6a-4d01-9fc5-264dd335eb21 button.colab-df-quickchart');\n",
              "  quickchartButtonEl.style.display =\n",
              "    google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "}\n",
              "\n",
              "        displayQuickchartButton(document);\n",
              "      </script>\n",
              "      <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\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",
              "    [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-37f6b6fc-2b74-471a-9242-adf9f62d73a3 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-37f6b6fc-2b74-471a-9242-adf9f62d73a3');\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",
              "  </div>\n"
            ]
          },
          "metadata": {},
          "execution_count": 9
        }
      ],
      "source": [
        "df"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tqVqnA67ebSB",
        "outputId": "a1bf31b8-0cba-4e2b-92ac-46275cdf2bba"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(90,)"
            ]
          },
          "metadata": {},
          "execution_count": 10
        }
      ],
      "source": [
        "size = df.shape[0]\n",
        "image_array = np.zeros((size,), dtype=object)  # Use object dtype to store arrays of varying sizes\n",
        "image_array.shape"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oyV90FMNb59Z",
        "outputId": "7d50c245-b634-4ec9-b60e-93ad7a623df6"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "0\n",
            "1\n",
            "2\n",
            "Error occurred for URL https://wiki.blogs.com/photos/uncategorized/2007/04/15/cheeeese.jpg: 403 Client Error: Forbidden for url: https://wiki.blogs.com/photos/uncategorized/2007/04/15/cheeeese.jpg\n",
            "3\n",
            "4\n",
            "5\n",
            "6\n",
            "7\n",
            "8\n",
            "9\n",
            "10\n",
            "11\n",
            "Error occurred for URL https://www.darice.com/Content/Images/ProductImages/XL/Main/9163-46.jpg: HTTPSConnectionPool(host='www.darice.com', port=443): Max retries exceeded with url: /Content/Images/ProductImages/XL/Main/9163-46.jpg (Caused by NewConnectionError('<urllib3.connection.HTTPSConnection object at 0x7dcbd0970f40>: Failed to establish a new connection: [Errno -2] Name or service not known'))\n",
            "12\n",
            "13\n",
            "Error occurred for URL https://www.tobar.fr/media/catalog/product/cache/3/small_image/450x450/9df78eab33525d08d6e5fb8d27136e95/E46113_800.jpg: 404 Client Error: Not Found for url: https://www.tobar.fr/media/catalog/product/cache/3/small_image/450x450/9df78eab33525d08d6e5fb8d27136e95/E46113_800.jpg\n",
            "14\n",
            "15\n",
            "16\n",
            "Error occurred for URL https://cdn1.tedsby.com/tb/medium/storage/3/9/2/392662/collectible-teddy-bear-sam-by-taya.jpg: 403 Client Error: Forbidden for url: https://cdn1.tedsby.com/tb/medium/storage/3/9/2/392662/collectible-teddy-bear-sam-by-taya.jpg\n",
            "17\n",
            "Error occurred for URL https://tereza-shop.com.ua/assets/cache/images/462775b7cf7b901ed83a5b543951d75b.jpg: 403 Client Error: Forbidden for url: https://tereza-shop.com.ua/assets/cache/images/462775b7cf7b901ed83a5b543951d75b.jpg\n",
            "18\n",
            "19\n",
            "20\n",
            "21\n",
            "22\n",
            "23\n",
            "24\n",
            "25\n",
            "26\n",
            "27\n",
            "28\n",
            "29\n",
            "Error occurred for URL https://cdn-img-2.wanelo.com/p/2f7/74e/9d4/12d3e63fae6d2bbe67af04b/x200-q90.jpg: 404 Client Error: Not Found for url: https://cdn-img-2.wanelo.com/p/2f7/74e/9d4/12d3e63fae6d2bbe67af04b/x200-q90.jpg\n",
            "30\n",
            "31\n",
            "32\n",
            "33\n",
            "34\n",
            "35\n",
            "36\n",
            "Error occurred for URL https://derestricted.com/wp-content/uploads/2016/05/1462545410-300x225.jpg: 404 Client Error: Not Found for url: https://derestricted.com/wp-content/uploads/2016/05/1462545410-300x225.jpg\n",
            "37\n",
            "38\n",
            "39\n",
            "40\n",
            "41\n",
            "42\n",
            "43\n",
            "44\n",
            "45\n",
            "46\n",
            "47\n",
            "Error occurred for URL https://images.yaoota.com/XRUmEmuy_eXc4eEMxwkDSLwj7xg=/trim/fit-in/200x200/filters:quality(80)/yaootaweb-production-sa/media/crawledproductimages/66ee082f347a7c95600c314afbcc352ab7b0b38a.jpg: HTTPSConnectionPool(host='images.yaoota.com', port=443): Max retries exceeded with url: /XRUmEmuy_eXc4eEMxwkDSLwj7xg=/trim/fit-in/200x200/filters:quality(80)/yaootaweb-production-sa/media/crawledproductimages/66ee082f347a7c95600c314afbcc352ab7b0b38a.jpg (Caused by SSLError(SSLCertVerificationError(1, '[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate (_ssl.c:1007)')))\n",
            "48\n",
            "Error occurred for URL https://cdn.shopify.com/s/files/1/0023/7004/4004/products/clean-soap-1_1024x1024.jpg?v=1595741745: 404 Client Error: Not Found for url: https://cdn.shopify.com/s/files/1/0023/7004/4004/products/clean-soap-1_1024x1024.jpg?v=1595741745\n",
            "49\n",
            "50\n",
            "Error occurred for URL https://cdn.shopify.com/s/files/1/0373/4260/9548/products/BOTANISTS_DIARY_TIE_TLIB136_300x300.jpg?v=1590015340: 404 Client Error: Not Found for url: https://cdn.shopify.com/s/files/1/0373/4260/9548/products/BOTANISTS_DIARY_TIE_TLIB136_300x300.jpg?v=1590015340\n",
            "51\n",
            "Error occurred for URL https://imagescdn.simons.ca/images/160-51716-70-A1_2/twisted-knit-mittens.jpg?__=3: 403 Client Error: Forbidden for url: https://imagescdn.simons.ca/images/160-51716-70-A1_2/twisted-knit-mittens.jpg?__=3\n",
            "52\n",
            "53\n",
            "54\n",
            "55\n",
            "56\n",
            "57\n",
            "58\n",
            "59\n",
            "60\n",
            "61\n",
            "62\n",
            "Error occurred for URL http://img.lovpho.com/anh/width248/2015/17-12/pancakes-amp-coffee-nicole-franzen-sunday-morning-sweet-breakfast-coffee-fruit-pancakes-good-morning-breakfast-food-coffee-breakfast.jpg: HTTPConnectionPool(host='img.lovpho.com', port=80): Max retries exceeded with url: /anh/width248/2015/17-12/pancakes-amp-coffee-nicole-franzen-sunday-morning-sweet-breakfast-coffee-fruit-pancakes-good-morning-breakfast-food-coffee-breakfast.jpg (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7dcb68eff8b0>: Failed to establish a new connection: [Errno -2] Name or service not known'))\n",
            "63\n",
            "64\n",
            "65\n",
            "66\n",
            "Error occurred for URL http://d3l2rivt3pqnj2.cloudfront.net/i/prints/rw/en_easyart/sm/4/2/The-Italian-Baker-Janet-Hill-427923.jpg: HTTPConnectionPool(host='d3l2rivt3pqnj2.cloudfront.net', port=80): Max retries exceeded with url: /i/prints/rw/en_easyart/sm/4/2/The-Italian-Baker-Janet-Hill-427923.jpg (Caused by NewConnectionError('<urllib3.connection.HTTPConnection object at 0x7dcb68eff190>: Failed to establish a new connection: [Errno -5] No address associated with hostname'))\n",
            "67\n",
            "68\n",
            "69\n",
            "70\n",
            "Error occurred for URL https://smashinghub.com/wp-content/uploads/2012/01/wall-stricker-26.jpg: 403 Client Error: Forbidden for url: https://smashinghub.com/wp-content/uploads/2012/01/wall-stricker-26.jpg\n",
            "71\n",
            "72\n",
            "73\n",
            "74\n",
            "75\n",
            "76\n",
            "77\n",
            "78\n",
            "79\n",
            "80\n",
            "81\n",
            "82\n",
            "83\n",
            "84\n",
            "85\n",
            "86\n",
            "87\n",
            "88\n",
            "89\n"
          ]
        }
      ],
      "source": [
        "count = 0\n",
        "for i in range(size):\n",
        "    urls = df.loc[i, 'url']\n",
        "\n",
        "    try:\n",
        "        print(i)\n",
        "        response = requests.get(urls)\n",
        "        response.raise_for_status()  # Check for any HTTP request errors\n",
        "\n",
        "        # Read the image data from the response content\n",
        "        image_data = response.content\n",
        "\n",
        "        # Create a PIL image object from the image data\n",
        "        image = Image.open(BytesIO(image_data))\n",
        "        image = image.resize((224, 224))\n",
        "\n",
        "        # Convert the PIL image to a NumPy array\n",
        "        image_np = np.array(image)\n",
        "\n",
        "        # Assign the NumPy array to the appropriate index in image_array\n",
        "        image_array[count] = image_np\n",
        "        count = count + 1\n",
        "\n",
        "    except requests.exceptions.RequestException as err:\n",
        "        df = df.drop(i)\n",
        "        print(f\"Error occurred for URL {urls}: {err}\")\n",
        "\n",
        "    except (IOError, ValueError) as err:\n",
        "        df = df.drop(i)\n",
        "        print(f\"Error occurred while processing image {i}: {err}\")\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "LbudCBGcXu52"
      },
      "outputs": [],
      "source": [
        "df = df.reset_index(drop=True)\n",
        "image_array = image_array[:count]\n",
        "df['image_array'] = image_array"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 423
        },
        "id": "2oZAdxPQPuBV",
        "outputId": "e77cf9c2-7622-408d-e1c7-6f9d742e76a3"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                                  url  \\\n",
              "0   https://static.wixstatic.com/media/039af8_8b6f...   \n",
              "1                    https://i.imgflip.com/2vo3mr.jpg   \n",
              "2   https://i.pinimg.com/200x150/36/c9/d5/36c9d5bb...   \n",
              "3   http://cdn.salemweb.net/godtube/2011/09/16/337...   \n",
              "4   https://thumbs.dreamstime.com/t/le-mari-et-l-%...   \n",
              "..                                                ...   \n",
              "71  https://freedomdestinations.co.uk/wp-content/u...   \n",
              "72  https://cdn.vox-cdn.com/thumbor/ZUQmkwGVccNDMY...   \n",
              "73  https://pic.pimg.tw/jcp1223/1387283539-2192754...   \n",
              "74  https://thumbs.dreamstime.com/t/%D0%B1%D0%B0%D...   \n",
              "75  https://cdn.shopify.com/s/files/1/0071/9136/36...   \n",
              "\n",
              "                                    prompt  category  \\\n",
              "0                        meme Bert Is Evil      meme   \n",
              "1                 meme #FirstWorldProblems      meme   \n",
              "2                           meme Keep Calm      meme   \n",
              "3        meme \"Fuck it, we'll do it live!\"      meme   \n",
              "4        meme \"Fuck it, we'll do it live!\"      meme   \n",
              "..                                     ...       ...   \n",
              "71  landmark Fairmont Le Château Frontenac  landmark   \n",
              "72          landmark Mexico City Cathedral  landmark   \n",
              "73                     landmark Taipei 101  landmark   \n",
              "74                       landmark CN Tower  landmark   \n",
              "75              landmark Westminster Abbey  landmark   \n",
              "\n",
              "                                          image_array  \n",
              "0   [[[241, 241, 232], [245, 243, 234], [235, 231,...  \n",
              "1   [[[194, 192, 171], [194, 192, 171], [193, 192,...  \n",
              "2   [[[57, 27, 3], [63, 33, 9], [63, 32, 7], [57, ...  \n",
              "3   [[[128, 161, 178], [154, 188, 205], [156, 188,...  \n",
              "4   [[[133, 111, 97], [132, 114, 99], [131, 117, 1...  \n",
              "..                                                ...  \n",
              "71  [[[175, 156, 126], [175, 158, 129], [178, 161,...  \n",
              "72  [[[29, 19, 8], [29, 16, 8], [25, 9, 4], [31, 1...  \n",
              "73  [[[114, 76, 11], [104, 68, 9], [93, 58, 10], [...  \n",
              "74  [[[81, 159, 221], [81, 159, 221], [81, 159, 22...  \n",
              "75  [[[253, 253, 253], [253, 253, 253], [253, 253,...  \n",
              "\n",
              "[76 rows x 4 columns]"
            ],
            "text/html": [
              "\n",
              "\n",
              "  <div id=\"df-30237f3c-e821-4792-bce8-e68980fcd190\">\n",
              "    <div 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>url</th>\n",
              "      <th>prompt</th>\n",
              "      <th>category</th>\n",
              "      <th>image_array</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>https://static.wixstatic.com/media/039af8_8b6f...</td>\n",
              "      <td>meme Bert Is Evil</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[241, 241, 232], [245, 243, 234], [235, 231,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>https://i.imgflip.com/2vo3mr.jpg</td>\n",
              "      <td>meme #FirstWorldProblems</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[194, 192, 171], [194, 192, 171], [193, 192,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>https://i.pinimg.com/200x150/36/c9/d5/36c9d5bb...</td>\n",
              "      <td>meme Keep Calm</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[57, 27, 3], [63, 33, 9], [63, 32, 7], [57, ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>http://cdn.salemweb.net/godtube/2011/09/16/337...</td>\n",
              "      <td>meme \"Fuck it, we'll do it live!\"</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[128, 161, 178], [154, 188, 205], [156, 188,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>https://thumbs.dreamstime.com/t/le-mari-et-l-%...</td>\n",
              "      <td>meme \"Fuck it, we'll do it live!\"</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[133, 111, 97], [132, 114, 99], [131, 117, 1...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>71</th>\n",
              "      <td>https://freedomdestinations.co.uk/wp-content/u...</td>\n",
              "      <td>landmark Fairmont Le Château Frontenac</td>\n",
              "      <td>landmark</td>\n",
              "      <td>[[[175, 156, 126], [175, 158, 129], [178, 161,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>72</th>\n",
              "      <td>https://cdn.vox-cdn.com/thumbor/ZUQmkwGVccNDMY...</td>\n",
              "      <td>landmark Mexico City Cathedral</td>\n",
              "      <td>landmark</td>\n",
              "      <td>[[[29, 19, 8], [29, 16, 8], [25, 9, 4], [31, 1...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>73</th>\n",
              "      <td>https://pic.pimg.tw/jcp1223/1387283539-2192754...</td>\n",
              "      <td>landmark Taipei 101</td>\n",
              "      <td>landmark</td>\n",
              "      <td>[[[114, 76, 11], [104, 68, 9], [93, 58, 10], [...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>74</th>\n",
              "      <td>https://thumbs.dreamstime.com/t/%D0%B1%D0%B0%D...</td>\n",
              "      <td>landmark CN Tower</td>\n",
              "      <td>landmark</td>\n",
              "      <td>[[[81, 159, 221], [81, 159, 221], [81, 159, 22...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75</th>\n",
              "      <td>https://cdn.shopify.com/s/files/1/0071/9136/36...</td>\n",
              "      <td>landmark Westminster Abbey</td>\n",
              "      <td>landmark</td>\n",
              "      <td>[[[253, 253, 253], [253, 253, 253], [253, 253,...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>76 rows × 4 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-30237f3c-e821-4792-bce8-e68980fcd190')\"\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 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "\n",
              "\n",
              "\n",
              "    <div id=\"df-360594a9-c8f6-4539-8c2f-e5009383c3f7\">\n",
              "      <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-360594a9-c8f6-4539-8c2f-e5009383c3f7')\"\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",
              "    </div>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\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-quickchart: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",
              "  [theme=dark] .colab-df-quickchart {\n",
              "    background-color: #3B4455;\n",
              "    fill: #D2E3FC;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart: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",
              "      async function quickchart(key) {\n",
              "        const containerElement = document.querySelector('#' + key);\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      }\n",
              "    </script>\n",
              "\n",
              "      <script>\n",
              "\n",
              "function displayQuickchartButton(domScope) {\n",
              "  let quickchartButtonEl =\n",
              "    domScope.querySelector('#df-360594a9-c8f6-4539-8c2f-e5009383c3f7 button.colab-df-quickchart');\n",
              "  quickchartButtonEl.style.display =\n",
              "    google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "}\n",
              "\n",
              "        displayQuickchartButton(document);\n",
              "      </script>\n",
              "      <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\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",
              "    [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-30237f3c-e821-4792-bce8-e68980fcd190 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-30237f3c-e821-4792-bce8-e68980fcd190');\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",
              "  </div>\n"
            ]
          },
          "metadata": {},
          "execution_count": 13
        }
      ],
      "source": [
        "df"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "j2kz51sniYmv"
      },
      "outputs": [],
      "source": [
        "num_rows = df.shape[0]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 434
        },
        "id": "wmXl5sjliaf0",
        "outputId": "34003b46-fe6e-4f54-93e7-2a6d31160d78"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "\n",
        "import matplotlib.pyplot as plt\n",
        "plt.imshow(image_array[3], cmap='viridis')\n",
        "#plt.colorbar()\n",
        "\n",
        "# Display the plot\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 423
        },
        "id": "R3dXmizSA_iS",
        "outputId": "ac9edae3-8ca1-45f7-8977-1e28cb5dc079"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                                  url  \\\n",
              "0   https://static.wixstatic.com/media/039af8_8b6f...   \n",
              "1                    https://i.imgflip.com/2vo3mr.jpg   \n",
              "2   https://i.pinimg.com/200x150/36/c9/d5/36c9d5bb...   \n",
              "3   http://cdn.salemweb.net/godtube/2011/09/16/337...   \n",
              "4   https://thumbs.dreamstime.com/t/le-mari-et-l-%...   \n",
              "..                                                ...   \n",
              "71  https://freedomdestinations.co.uk/wp-content/u...   \n",
              "72  https://cdn.vox-cdn.com/thumbor/ZUQmkwGVccNDMY...   \n",
              "73  https://pic.pimg.tw/jcp1223/1387283539-2192754...   \n",
              "74  https://thumbs.dreamstime.com/t/%D0%B1%D0%B0%D...   \n",
              "75  https://cdn.shopify.com/s/files/1/0071/9136/36...   \n",
              "\n",
              "                                    prompt  category  \\\n",
              "0                        meme Bert Is Evil      meme   \n",
              "1                 meme #FirstWorldProblems      meme   \n",
              "2                           meme Keep Calm      meme   \n",
              "3        meme \"Fuck it, we'll do it live!\"      meme   \n",
              "4        meme \"Fuck it, we'll do it live!\"      meme   \n",
              "..                                     ...       ...   \n",
              "71  landmark Fairmont Le Château Frontenac  landmark   \n",
              "72          landmark Mexico City Cathedral  landmark   \n",
              "73                     landmark Taipei 101  landmark   \n",
              "74                       landmark CN Tower  landmark   \n",
              "75              landmark Westminster Abbey  landmark   \n",
              "\n",
              "                                          image_array  \n",
              "0   [[[241, 241, 232], [245, 243, 234], [235, 231,...  \n",
              "1   [[[194, 192, 171], [194, 192, 171], [193, 192,...  \n",
              "2   [[[57, 27, 3], [63, 33, 9], [63, 32, 7], [57, ...  \n",
              "3   [[[128, 161, 178], [154, 188, 205], [156, 188,...  \n",
              "4   [[[133, 111, 97], [132, 114, 99], [131, 117, 1...  \n",
              "..                                                ...  \n",
              "71  [[[175, 156, 126], [175, 158, 129], [178, 161,...  \n",
              "72  [[[29, 19, 8], [29, 16, 8], [25, 9, 4], [31, 1...  \n",
              "73  [[[114, 76, 11], [104, 68, 9], [93, 58, 10], [...  \n",
              "74  [[[81, 159, 221], [81, 159, 221], [81, 159, 22...  \n",
              "75  [[[253, 253, 253], [253, 253, 253], [253, 253,...  \n",
              "\n",
              "[76 rows x 4 columns]"
            ],
            "text/html": [
              "\n",
              "\n",
              "  <div id=\"df-bc9fe2d3-e9fb-4e8c-a08e-772629a9d287\">\n",
              "    <div 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>url</th>\n",
              "      <th>prompt</th>\n",
              "      <th>category</th>\n",
              "      <th>image_array</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>https://static.wixstatic.com/media/039af8_8b6f...</td>\n",
              "      <td>meme Bert Is Evil</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[241, 241, 232], [245, 243, 234], [235, 231,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>https://i.imgflip.com/2vo3mr.jpg</td>\n",
              "      <td>meme #FirstWorldProblems</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[194, 192, 171], [194, 192, 171], [193, 192,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>https://i.pinimg.com/200x150/36/c9/d5/36c9d5bb...</td>\n",
              "      <td>meme Keep Calm</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[57, 27, 3], [63, 33, 9], [63, 32, 7], [57, ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>http://cdn.salemweb.net/godtube/2011/09/16/337...</td>\n",
              "      <td>meme \"Fuck it, we'll do it live!\"</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[128, 161, 178], [154, 188, 205], [156, 188,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>https://thumbs.dreamstime.com/t/le-mari-et-l-%...</td>\n",
              "      <td>meme \"Fuck it, we'll do it live!\"</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[133, 111, 97], [132, 114, 99], [131, 117, 1...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>71</th>\n",
              "      <td>https://freedomdestinations.co.uk/wp-content/u...</td>\n",
              "      <td>landmark Fairmont Le Château Frontenac</td>\n",
              "      <td>landmark</td>\n",
              "      <td>[[[175, 156, 126], [175, 158, 129], [178, 161,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>72</th>\n",
              "      <td>https://cdn.vox-cdn.com/thumbor/ZUQmkwGVccNDMY...</td>\n",
              "      <td>landmark Mexico City Cathedral</td>\n",
              "      <td>landmark</td>\n",
              "      <td>[[[29, 19, 8], [29, 16, 8], [25, 9, 4], [31, 1...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>73</th>\n",
              "      <td>https://pic.pimg.tw/jcp1223/1387283539-2192754...</td>\n",
              "      <td>landmark Taipei 101</td>\n",
              "      <td>landmark</td>\n",
              "      <td>[[[114, 76, 11], [104, 68, 9], [93, 58, 10], [...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>74</th>\n",
              "      <td>https://thumbs.dreamstime.com/t/%D0%B1%D0%B0%D...</td>\n",
              "      <td>landmark CN Tower</td>\n",
              "      <td>landmark</td>\n",
              "      <td>[[[81, 159, 221], [81, 159, 221], [81, 159, 22...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75</th>\n",
              "      <td>https://cdn.shopify.com/s/files/1/0071/9136/36...</td>\n",
              "      <td>landmark Westminster Abbey</td>\n",
              "      <td>landmark</td>\n",
              "      <td>[[[253, 253, 253], [253, 253, 253], [253, 253,...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>76 rows × 4 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-bc9fe2d3-e9fb-4e8c-a08e-772629a9d287')\"\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 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "\n",
              "\n",
              "\n",
              "    <div id=\"df-7ddd2254-9a24-4487-b02f-93537a9291d0\">\n",
              "      <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-7ddd2254-9a24-4487-b02f-93537a9291d0')\"\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",
              "    </div>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\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-quickchart: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",
              "  [theme=dark] .colab-df-quickchart {\n",
              "    background-color: #3B4455;\n",
              "    fill: #D2E3FC;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart: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",
              "      async function quickchart(key) {\n",
              "        const containerElement = document.querySelector('#' + key);\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      }\n",
              "    </script>\n",
              "\n",
              "      <script>\n",
              "\n",
              "function displayQuickchartButton(domScope) {\n",
              "  let quickchartButtonEl =\n",
              "    domScope.querySelector('#df-7ddd2254-9a24-4487-b02f-93537a9291d0 button.colab-df-quickchart');\n",
              "  quickchartButtonEl.style.display =\n",
              "    google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "}\n",
              "\n",
              "        displayQuickchartButton(document);\n",
              "      </script>\n",
              "      <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\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",
              "    [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-bc9fe2d3-e9fb-4e8c-a08e-772629a9d287 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-bc9fe2d3-e9fb-4e8c-a08e-772629a9d287');\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",
              "  </div>\n"
            ]
          },
          "metadata": {},
          "execution_count": 16
        }
      ],
      "source": [
        "df"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EHKObkrNrwk5",
        "outputId": "8b77d09c-41b3-47fe-dd13-5ee2b569e1c0"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "['sierra', 'peggy’s', 'fairmont', 'paintings', 'live!\"', 'football', 'westminster', 'clothes', 'seoul', 'adam', 'toys', 'rice', 'meme', 'pillow', '1500', 'egg', 'silverado', \"men's\", 'ball', 'tower', 'beans', 'le', 'kettle', 'point', '101', 'is', 'taipei', 'with', 'cat', 'girl', 'frontenac', 'beach', 'cabinet', 'ring', 'auchan', 'lamp', 'ram', 'michelangelo', 'bondi', 'city', 'cafe', 'lego', 'do', 'chopper', 'chevrolet', '2020', 'abbey', 'square', 'picture', 'curtains', 'stool', 'this', 'canada', 'wolmart', 'it', 'tacoma', 'toy', 'snacks', 'painting', 'suit', 'keep', 'bert', 'the', 'artwork', 'olympia', 'argument', 'night', '\"fuck', 'cn', '#firstworldproblems', 'gmc', 'a', 'lisa', 'mona', 'tv', 'happy', '#winning', 'sandro', 'blocks', 'hotdog', 'arnolfini', 'of', 'lidl', 'cathedral', 'umbrella', 'pearl', 'jacket', 'packaged', 'and', 'car', 'pepsi', 'mittens', 'kitchen', 'purse', 'storefronts', 'civic', 'photos', 'creation', 'evil', 'storefront', 'wrap', 'mexico', 'primavera', 'sausages', 'landmark', 'doll', 'earring', \"we'll\", 'bed', 'goods', 'accidents', 'jean-honore', 'one', 'male', 'watch', 'swing', 'fragonard', 'bottle', 'can', 'portrait', 'botticelli', 'château', 'fried', 'american', 'toyota', 'rickrolling', 'edouard', 'honda', 'lighthouse', 'aldi', 'berlin', 'it,', 'manet', 'rav4', 'calm', 'furniture', 'ceiling', 't-shirt', 'oven', 'sneakers', 'multipack', 'food']\n"
          ]
        }
      ],
      "source": [
        "#TOKENIZER\n",
        "\n",
        "\n",
        "token_words = ['this', 'is', 'a', 'picture', 'of', 'a']\n",
        "len_token = len(token_words)\n",
        "\n",
        "def tokenizer(words):\n",
        "    unique_tokens = set(words)\n",
        "    return list(unique_tokens)\n",
        "\n",
        "for k in range (0, num_rows):\n",
        "  prompt = df.loc[k, 'prompt']\n",
        "  prompt = prompt.lower()\n",
        "  #print(prompt)\n",
        "  x = prompt.split()\n",
        "  token_words.extend(x)\n",
        "  #token_words.append(prompt)\n",
        "  token_words = tokenizer(token_words)\n",
        "\n",
        "for k in range (0, num_rows):\n",
        "  prompt = df.loc[k, 'category']\n",
        "  prompt = prompt.lower()\n",
        "  #print(prompt)\n",
        "  x = prompt.split()\n",
        "  token_words.extend(x)\n",
        "  #token_words.append(prompt)\n",
        "  token_words = tokenizer(token_words)\n",
        "\n",
        "print(token_words)"
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "zJhDE-Mj1jFr"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "1OkxWV3j1jMD"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "wRiulGJC1jQ6"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "B3ADT1MT1jUb"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BdBl_GSrsBHi",
        "outputId": "87afa755-8b20-4d47-f00b-2f928de2a791"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "142"
            ]
          },
          "metadata": {},
          "execution_count": 18
        }
      ],
      "source": [
        "len(token_words)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "UbLGZtomsfQa"
      },
      "outputs": [],
      "source": [
        "values = []\n",
        "for i in range (1, len(token_words)+1):\n",
        "  values.append(i)\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "MhQwWIzI2Lpj"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "_PgnNNZPs8gf"
      },
      "outputs": [],
      "source": [
        "def token(i):\n",
        "  prompt = df.loc[i, 'prompt']\n",
        "  prompt = prompt.lower()\n",
        "  x = prompt.split()\n",
        "  token_words.extend(x)\n",
        "  return tokenizer(token_words)"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "text_paths = df['prompt'].head(10).tolist()"
      ],
      "metadata": {
        "id": "kerqZ9NC1iBh"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "T4VHN9hG1iE8"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "(text_paths)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "5pHvwxJIUiOT",
        "outputId": "d96ab2b7-04a9-42f2-bc3f-c6fda7acd841"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "['meme Bert Is Evil',\n",
              " 'meme #FirstWorldProblems',\n",
              " 'meme Keep Calm',\n",
              " 'meme \"Fuck it, we\\'ll do it live!\"',\n",
              " 'meme \"Fuck it, we\\'ll do it live!\"',\n",
              " 'meme Rickrolling',\n",
              " 'meme American Chopper argument',\n",
              " 'meme #Winning',\n",
              " 'meme Ceiling Cat',\n",
              " 'toy doll']"
            ]
          },
          "metadata": {},
          "execution_count": 99
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "text_paths = [word.lower() for word in text_paths]"
      ],
      "metadata": {
        "id": "SPI4Fo563GRc"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "dataset = text_paths"
      ],
      "metadata": {
        "id": "_G4Fc0l5WQwi"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "Q53B-0rkOUne"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "dTD8BNgpOXNY"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Example vocabulary mapping with \"<PAD>\" token\n",
        "vocab = {}\n",
        "for i in range(len(token_words)):\n",
        "    key = token_words[i]\n",
        "    value = values[i]\n",
        "    vocab[key] = value\n",
        "\n",
        "# Add the \"<PAD>\" token to the vocab\n",
        "vocab[\"<PAD>\"] = 0\n",
        "\n",
        "sentences_list = text_paths\n",
        "\n",
        "# Initialize an empty list to store tokenized sentences\n",
        "tokenized_sentences = []\n",
        "max_sentence_length = 20\n",
        "tokenized_sentences_with_padding = []\n",
        "\n",
        "# Loop through each sentence in the sentences_list\n",
        "for sentence in sentences_list:\n",
        "    # Split the sentence into words\n",
        "    words = sentence.split()\n",
        "    # Initialize an empty list to store tokens for this sentence\n",
        "    tokenized_sentence = []\n",
        "    # Loop through each word in the words list\n",
        "    for word in words:\n",
        "        # Convert the word to lowercase\n",
        "        lowercase_word = word.lower()\n",
        "        # Check if the lowercase word exists in the vocab dictionary\n",
        "        if lowercase_word in vocab:\n",
        "            # If it exists, append its corresponding token to the tokenized_sentence list\n",
        "            tokenized_sentence.append(vocab[lowercase_word])\n",
        "    # Add padding to the tokenized_sentence to make its length 25\n",
        "    tokenized_sentence += [vocab['<PAD>']] * (max_sentence_length - len(tokenized_sentence))\n",
        "    # Truncate the sentence to the maximum length (in case it's longer than 25)\n",
        "    tokenized_sentence = tokenized_sentence[:max_sentence_length]\n",
        "    # Append the tokenized_sentence list to the tokenized_sentences_with_padding list\n",
        "    tokenized_sentences_with_padding.append(tokenized_sentence)\n",
        "\n",
        "print(tokenized_sentences_with_padding)\n",
        "input_tensor = torch.tensor(tokenized_sentences_with_padding, dtype=torch.long)\n",
        "print(input_tensor.shape)\n",
        "\n",
        "\n",
        "# Add any missing words from the dataset to the vocab\n",
        "#for sentence in text_paths:\n",
        "#    for word in sentence:\n",
        "#        if word not in vocab:\n",
        "#            vocab[word] = len(vocab)\n",
        "\n",
        "# Now convert the dataset to numerical_dataset\n",
        "#numerical_dataset = [[vocab[word] for word in sentence] for sentence in dataset]\n",
        "\n",
        "#max_sequence_length = max(len(sentence) for sentence in numerical_dataset)\n",
        "#padded_numerical_dataset = [sentence + [vocab[\"<PAD>\"]] * (max_sequence_length - len(sentence)) for sentence in numerical_dataset]\n",
        "\n",
        "#input_tensor = torch.tensor(padded_numerical_dataset, dtype=torch.long)\n",
        "#print(input_tensor.shape)\n",
        "\n",
        "mask = (input_tensor != vocab[\"<PAD>\"]).unsqueeze(1).unsqueeze(2)\n",
        "print(mask.shape)\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ROWhFnq_VXit",
        "outputId": "78d9983f-57d0-4844-e136-d62e4f9d73bf"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[[13, 62, 26, 99, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [13, 70, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [13, 61, 135, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [13, 68, 132, 108, 43, 55, 5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [13, 68, 132, 108, 43, 55, 5, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [13, 126, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [13, 124, 44, 66, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [13, 77, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [13, 137, 29, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], [57, 106, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]]\n",
            "torch.Size([10, 20])\n",
            "torch.Size([10, 1, 1, 20])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "mask = mask.float()"
      ],
      "metadata": {
        "id": "XDyORpyePZVq"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1_zR73gVPb-H",
        "outputId": "68deb576-7e1b-46fb-f694-7a6c8806277b"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[[[1., 1., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "           0., 0., 0.]]],\n",
              "\n",
              "\n",
              "        [[[1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "           0., 0., 0.]]],\n",
              "\n",
              "\n",
              "        [[[1., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "           0., 0., 0.]]],\n",
              "\n",
              "\n",
              "        [[[1., 1., 1., 1., 1., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "           0., 0., 0.]]],\n",
              "\n",
              "\n",
              "        [[[1., 1., 1., 1., 1., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "           0., 0., 0.]]],\n",
              "\n",
              "\n",
              "        [[[1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "           0., 0., 0.]]],\n",
              "\n",
              "\n",
              "        [[[1., 1., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "           0., 0., 0.]]],\n",
              "\n",
              "\n",
              "        [[[1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "           0., 0., 0.]]],\n",
              "\n",
              "\n",
              "        [[[1., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "           0., 0., 0.]]],\n",
              "\n",
              "\n",
              "        [[[1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.,\n",
              "           0., 0., 0.]]]])"
            ]
          },
          "metadata": {},
          "execution_count": 109
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "FBwHqqjGYxOQ"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "\n",
        "\n",
        "class SelfAttention(nn.Module):\n",
        "    def __init__(self, embed_size, heads):\n",
        "        super(SelfAttention, self).__init__()\n",
        "        self.embed_size = embed_size\n",
        "        self.heads = heads\n",
        "        self.head_dim = embed_size // heads\n",
        "\n",
        "        assert (\n",
        "            self.head_dim * heads == embed_size\n",
        "        ), \"Embedding size needs to be divisible by heads\"\n",
        "\n",
        "        self.values = nn.Linear(embed_size, embed_size)\n",
        "        self.keys = nn.Linear(embed_size, embed_size)\n",
        "        self.queries = nn.Linear(embed_size, embed_size)\n",
        "        self.fc_out = nn.Linear(embed_size, embed_size)\n",
        "\n",
        "    def forward(self, values, keys, query, mask):\n",
        "        # Get number of training examples\n",
        "        N = query.shape[0]\n",
        "\n",
        "        value_len, key_len, query_len = values.shape[1], keys.shape[1], query.shape[1]\n",
        "\n",
        "        values = self.values(values)  # (N, value_len, embed_size)\n",
        "        keys = self.keys(keys)  # (N, key_len, embed_size)\n",
        "        queries = self.queries(query)  # (N, query_len, embed_size)\n",
        "\n",
        "        # Split the embedding into self.heads different pieces\n",
        "        values = values.reshape(N, value_len, self.heads, self.head_dim)\n",
        "        keys = keys.reshape(N, key_len, self.heads, self.head_dim)\n",
        "        queries = queries.reshape(N, query_len, self.heads, self.head_dim)\n",
        "\n",
        "        energy = torch.einsum(\"nqhd,nkhd->nhqk\", [queries, keys])\n",
        "        # queries shape: (N, query_len, heads, heads_dim),\n",
        "        # keys shape: (N, key_len, heads, heads_dim)\n",
        "        # energy: (N, heads, query_len, key_len)\n",
        "\n",
        "        if mask is not None:\n",
        "            energy = energy.masked_fill(mask == 0, float(\"-1e20\"))\n",
        "\n",
        "        attention = torch.softmax(energy / (self.embed_size ** (1 / 2)), dim=3)\n",
        "        # attention shape: (N, heads, query_len, key_len)\n",
        "\n",
        "        out = torch.einsum(\"nhql,nlhd->nqhd\", [attention, values]).reshape(\n",
        "            N, query_len, self.heads * self.head_dim\n",
        "        )\n",
        "\n",
        "        out = self.fc_out(out)\n",
        "\n",
        "\n",
        "        return out\n",
        "\n",
        "\n",
        "class TransformerBlock(nn.Module):\n",
        "    def __init__(self, embed_size, heads, dropout, forward_expansion):\n",
        "        super(TransformerBlock, self).__init__()\n",
        "        self.attention = SelfAttention(embed_size, heads)\n",
        "        self.norm1 = nn.LayerNorm(embed_size)\n",
        "        self.norm2 = nn.LayerNorm(embed_size)\n",
        "\n",
        "        self.feed_forward = nn.Sequential(\n",
        "            nn.Linear(embed_size, forward_expansion * embed_size),\n",
        "            nn.ReLU(),\n",
        "            nn.Linear(forward_expansion * embed_size, embed_size),\n",
        "        )\n",
        "\n",
        "        self.dropout = nn.Dropout(dropout)\n",
        "\n",
        "    def forward(self, value, key, query, mask):\n",
        "        attention = self.attention(value, key, query, mask)\n",
        "\n",
        "        # Add skip connection, run through normalization and finally dropout\n",
        "        x = self.dropout(self.norm1(attention + query))\n",
        "        forward = self.feed_forward(x)\n",
        "        out = self.dropout(self.norm2(forward + x))\n",
        "        return out\n",
        "\n",
        "\n",
        "class Encoder(nn.Module):\n",
        "    def __init__(\n",
        "        self,\n",
        "        src_vocab_size,\n",
        "        embed_size,\n",
        "        num_layers,\n",
        "        heads,\n",
        "        device,\n",
        "        forward_expansion,\n",
        "        dropout,\n",
        "        max_length,\n",
        "    ):\n",
        "\n",
        "        super(Encoder, self).__init__()\n",
        "        self.embed_size = embed_size\n",
        "        self.device = device\n",
        "        self.word_embedding = nn.Embedding(src_vocab_size, embed_size)\n",
        "        self.position_embedding = nn.Embedding(max_length, embed_size)\n",
        "\n",
        "        self.layers = nn.ModuleList(\n",
        "            [\n",
        "                TransformerBlock(\n",
        "                    embed_size,\n",
        "                    heads,\n",
        "                    dropout=dropout,\n",
        "                    forward_expansion=forward_expansion,\n",
        "                )\n",
        "                for _ in range(num_layers)\n",
        "            ]\n",
        "        )\n",
        "\n",
        "        self.dropout = nn.Dropout(dropout)\n",
        "\n",
        "    def forward(self, x, mask):\n",
        "        N, seq_length = x.shape\n",
        "        positions = torch.arange(0, seq_length).expand(N, seq_length).to(self.device)\n",
        "        out = self.dropout(\n",
        "            (self.word_embedding(x) + self.position_embedding(positions))\n",
        "        )\n",
        "\n",
        "        # In the Encoder the query, key, value are all the same, it's in the\n",
        "        # decoder this will change. This might look a bit odd in this case.\n",
        "        for layer in self.layers:\n",
        "            out = layer(out, out, out, mask)\n",
        "\n",
        "        return out\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "-RT-NQoNO4nB"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "JyNo0dJpO4zD"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "HiQPZD6SYxQe"
      },
      "outputs": [],
      "source": [
        "src_vocab_size = 1000  # Assuming there are 10,000 unique words in the source vocabulary\n",
        "#embed_size = 128        # Dimension of word embeddings is 256\n",
        "embed_size = 196*2\n",
        "num_layers = 6          # Using 6 Transformer blocks (layers)\n",
        "heads = 8               # 8 attention heads\n",
        "forward_expansion = 4   # Factor by which the hidden dimension is expanded in the feed-forward network is 4\n",
        "dropout = 0.1           # Dropout rate of 0.1 (10%)\n",
        "max_length = 150\n",
        "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")  # Use GPU if available\n",
        "\n",
        "#encoder = Encoder(src_vocab_size, embed_size, num_layers, heads, device, forward_expansion, dropout, max_length)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "y_gJ8cqJMQ3h"
      },
      "outputs": [],
      "source": [
        "encoder = Encoder(src_vocab_size, embed_size, num_layers, heads, device, forward_expansion, dropout, max_length)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "RxiZTe-gcJ4K"
      },
      "outputs": [],
      "source": [
        "text_output_tensor = encoder(input_tensor, mask)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "vpYhwuGkcJ8p",
        "outputId": "6d859373-14e3-4f1a-ab24-8bf4abe95bd5"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "torch.Size([10, 20, 392])"
            ]
          },
          "metadata": {},
          "execution_count": 114
        }
      ],
      "source": [
        "text_output_tensor.shape"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "-1leq-qCUWLd",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9d101e7a-da91-49b9-819b-a5aecb34c326"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[[ 4.8373e-01, -4.0865e-01,  1.8290e-01,  ..., -9.3966e-01,\n",
              "          -6.3647e-01,  3.6896e-01],\n",
              "         [-2.2570e-01,  5.4389e-02, -9.6935e-01,  ...,  1.4189e-01,\n",
              "          -6.0111e-01,  5.5331e-02],\n",
              "         [ 2.6655e+00,  1.9713e+00, -0.0000e+00,  ..., -3.8898e-01,\n",
              "          -1.8935e-01, -2.1133e-02],\n",
              "         ...,\n",
              "         [-1.8207e-01,  2.2909e+00,  8.4090e-01,  ..., -8.2671e-01,\n",
              "           1.5485e-01, -3.7554e-01],\n",
              "         [ 9.9135e-02,  3.0946e-01,  3.5237e-01,  ..., -1.4797e-01,\n",
              "           2.9978e-01, -2.4990e-01],\n",
              "         [ 5.3351e-01,  2.9888e-01, -2.5749e-01,  ..., -8.0302e-01,\n",
              "          -1.9537e-01, -8.8437e-01]],\n",
              "\n",
              "        [[-2.9763e+00, -3.8492e-01, -3.2973e-01,  ...,  1.5482e-01,\n",
              "          -1.4199e+00,  0.0000e+00],\n",
              "         [ 9.8967e-01,  2.1981e+00,  3.1751e-01,  ..., -1.8700e+00,\n",
              "           2.5706e-02, -2.4527e-01],\n",
              "         [ 0.0000e+00,  2.8851e+00,  4.9566e-01,  ...,  3.6260e-01,\n",
              "          -1.7338e-04,  0.0000e+00],\n",
              "         ...,\n",
              "         [ 1.0050e+00, -1.9457e-01, -5.6011e-01,  ...,  3.0190e-02,\n",
              "          -1.6643e+00,  1.1085e+00],\n",
              "         [ 4.5992e+00,  0.0000e+00, -3.4926e-01,  ...,  4.0406e-01,\n",
              "          -1.9229e+00, -1.0227e-01],\n",
              "         [-0.0000e+00, -9.9209e-01, -6.6189e-02,  ..., -0.0000e+00,\n",
              "          -2.2630e+00, -1.1129e-01]],\n",
              "\n",
              "        [[ 5.5374e-02,  6.8074e-01,  5.9177e-02,  ..., -6.5803e-01,\n",
              "          -5.3337e-01,  1.5228e+00],\n",
              "         [-1.0177e+00,  3.4825e+00,  1.4915e-01,  ...,  0.0000e+00,\n",
              "          -0.0000e+00, -4.9773e-01],\n",
              "         [-1.6620e-02,  1.3955e+00, -6.3371e-01,  ..., -1.9422e-01,\n",
              "          -5.5824e-01,  2.5594e+00],\n",
              "         ...,\n",
              "         [ 1.0027e+00,  7.3793e-02,  1.0224e+00,  ...,  8.2434e-02,\n",
              "          -2.5484e-01,  2.6229e+00],\n",
              "         [ 2.4113e+00, -4.6537e-03, -3.5274e-01,  ...,  3.0970e-01,\n",
              "          -0.0000e+00,  2.0950e-01],\n",
              "         [ 0.0000e+00,  1.8318e-01, -2.6166e-01,  ..., -4.4075e-01,\n",
              "          -3.5894e-02,  8.3246e-01]],\n",
              "\n",
              "        ...,\n",
              "\n",
              "        [[ 8.1489e-01, -4.6460e-01,  9.4730e-01,  ...,  2.4888e+00,\n",
              "           2.9141e-01, -1.0535e+00],\n",
              "         [ 1.2392e+00,  2.9179e-01,  1.7834e-01,  ...,  0.0000e+00,\n",
              "          -3.3514e-01,  3.1852e-01],\n",
              "         [ 1.0798e+00, -1.1663e+00,  1.1783e+00,  ...,  1.8504e+00,\n",
              "          -4.6721e-01, -3.6731e-01],\n",
              "         ...,\n",
              "         [ 1.4057e+00, -6.3528e-01, -4.3615e-01,  ...,  1.5681e+00,\n",
              "          -5.6478e-01, -3.0669e-01],\n",
              "         [ 1.2540e-01,  2.7734e+00, -1.7021e+00,  ...,  2.6862e+00,\n",
              "          -0.0000e+00, -1.2191e+00],\n",
              "         [ 9.3399e-01,  6.4967e-01,  4.5390e-01,  ...,  3.0940e+00,\n",
              "           2.5581e-02, -2.8862e+00]],\n",
              "\n",
              "        [[ 4.6199e-01, -5.1340e-01,  3.8005e-01,  ..., -7.0135e-02,\n",
              "          -1.5212e+00,  8.6018e-02],\n",
              "         [ 5.5703e-01,  3.7414e+00,  7.5812e-01,  ...,  7.1323e-01,\n",
              "          -8.3415e-01,  5.7620e-01],\n",
              "         [ 2.8244e-02,  2.2938e-01,  9.5640e-01,  ..., -8.1638e-02,\n",
              "          -6.9071e-01, -0.0000e+00],\n",
              "         ...,\n",
              "         [ 0.0000e+00,  3.7455e-01,  4.3838e-01,  ..., -4.0311e-01,\n",
              "          -1.6781e-01, -3.8667e-02],\n",
              "         [ 1.8051e-01,  4.8411e-01, -1.6366e-01,  ...,  1.3522e-01,\n",
              "          -7.2171e-01,  1.5645e-01],\n",
              "         [-2.5865e-01,  6.2967e-01,  1.4624e+00,  ...,  5.1745e-01,\n",
              "           1.4016e-01,  0.0000e+00]],\n",
              "\n",
              "        [[ 0.0000e+00,  5.0924e-01,  4.1421e-01,  ...,  1.8980e+00,\n",
              "          -9.2620e-01, -1.2456e-01],\n",
              "         [-2.9115e-01, -2.1731e-01,  1.0119e+00,  ..., -6.7870e-02,\n",
              "           0.0000e+00,  0.0000e+00],\n",
              "         [ 3.4263e-01,  1.1998e+00,  5.8004e-01,  ...,  1.3122e-01,\n",
              "          -0.0000e+00, -1.2879e-01],\n",
              "         ...,\n",
              "         [ 8.5322e-01,  2.9436e+00,  1.0656e+00,  ...,  1.1087e+00,\n",
              "          -1.2608e+00,  8.3279e-01],\n",
              "         [ 4.4071e+00,  4.0453e+00, -2.6144e-01,  ...,  0.0000e+00,\n",
              "           4.0902e-02,  1.0696e+00],\n",
              "         [ 9.5102e-01,  5.4228e-01,  8.3723e-01,  ...,  5.1187e-01,\n",
              "          -8.8983e-01,  7.8805e-01]]], grad_fn=<MulBackward0>)"
            ]
          },
          "metadata": {},
          "execution_count": 115
        }
      ],
      "source": [
        "text_output_tensor"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "LELScRPWUWOU"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "\n",
        "class PatchEmbedding(nn.Module):\n",
        "    def __init__(self, img_size, patch_size, in_channels, embed_size):\n",
        "        super(PatchEmbedding, self).__init__()\n",
        "        self.img_size = img_size\n",
        "        self.patch_size = patch_size\n",
        "        self.in_channels = in_channels\n",
        "        self.embed_size = embed_size\n",
        "        self.num_patches = (img_size // patch_size) ** 2\n",
        "\n",
        "        self.projection = nn.Conv2d(in_channels, embed_size, kernel_size=patch_size, stride=patch_size)\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.projection(x)  # (batch_size, embed_size, num_patches, num_patches)\n",
        "        x = x.flatten(2).transpose(1, 2)  # (batch_size, num_patches*num_patches, embed_size)\n",
        "        return x\n",
        "\n",
        "\n",
        "class VisionTransformer(nn.Module):\n",
        "    def __init__(\n",
        "        self,\n",
        "        img_size,\n",
        "        patch_size,\n",
        "        in_channels,\n",
        "        embed_size,\n",
        "        num_layers,\n",
        "        heads,\n",
        "        device,\n",
        "        forward_expansion,\n",
        "        dropout,\n",
        "        num_classes,\n",
        "        max_length,\n",
        "    ):\n",
        "        super(VisionTransformer, self).__init__()\n",
        "        self.patch_embedding = PatchEmbedding(img_size, patch_size, in_channels, embed_size)\n",
        "        self.position_embedding = nn.Embedding(max_length, embed_size)\n",
        "\n",
        "        self.layers = nn.ModuleList(\n",
        "            [\n",
        "                TransformerBlock(\n",
        "                    embed_size,\n",
        "                    heads,\n",
        "                    dropout=dropout,\n",
        "                    forward_expansion=forward_expansion,\n",
        "                )\n",
        "                for _ in range(num_layers)\n",
        "            ]\n",
        "        )\n",
        "\n",
        "        self.classification_head = nn.Linear(embed_size, num_classes)\n",
        "        self.dropout = nn.Dropout(dropout)\n",
        "        self.device = device\n",
        "\n",
        "    def forward(self, x, mask=None):\n",
        "        x = self.patch_embedding(x)\n",
        "        N, num_patches, _ = x.shape\n",
        "\n",
        "        positions = torch.arange(0, num_patches).expand(N, num_patches).to(self.device)\n",
        "        x += self.position_embedding(positions)\n",
        "\n",
        "        for layer in self.layers:\n",
        "            x = layer(x, x, x, mask)\n",
        "        #print(x.shape)\n",
        "        #print(x[:, 0].shape)\n",
        "        # Take only the first token (CLS token) and pass it through the classification head\n",
        "        #cls_token = x[:, 0]\n",
        "        cls_token = x\n",
        "        cls_token = self.dropout(cls_token)\n",
        "        logits = self.classification_head(cls_token)\n",
        "\n",
        "        return logits\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "-NYZGRs_aCYB"
      },
      "outputs": [],
      "source": [
        "dftemp = df.head(10)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 363
        },
        "id": "Rlb3Oy6uaCWt",
        "outputId": "90154f3b-30ac-48f5-b1d5-9091897e90dc"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "                                                 url  \\\n",
              "0  https://static.wixstatic.com/media/039af8_8b6f...   \n",
              "1                   https://i.imgflip.com/2vo3mr.jpg   \n",
              "2  https://i.pinimg.com/200x150/36/c9/d5/36c9d5bb...   \n",
              "3  http://cdn.salemweb.net/godtube/2011/09/16/337...   \n",
              "4  https://thumbs.dreamstime.com/t/le-mari-et-l-%...   \n",
              "5  http://wheresmysammich.com/images/thumbs/47937...   \n",
              "6  https://thumbs.dreamstime.com/z/gor%C4%85ce-ro...   \n",
              "7  https://postila.ru/resize?w=460&src=%2Fdata%2F...   \n",
              "8  https://thumbs.dreamstime.com/z/toskana-k%C3%B...   \n",
              "9  https://cdn.shopify.com/s/files/1/1377/4645/pr...   \n",
              "\n",
              "                              prompt category  \\\n",
              "0                  meme Bert Is Evil     meme   \n",
              "1           meme #FirstWorldProblems     meme   \n",
              "2                     meme Keep Calm     meme   \n",
              "3  meme \"Fuck it, we'll do it live!\"     meme   \n",
              "4  meme \"Fuck it, we'll do it live!\"     meme   \n",
              "5                   meme Rickrolling     meme   \n",
              "6     meme American Chopper argument     meme   \n",
              "7                      meme #Winning     meme   \n",
              "8                   meme Ceiling Cat     meme   \n",
              "9                           toy doll     toys   \n",
              "\n",
              "                                         image_array  \n",
              "0  [[[241, 241, 232], [245, 243, 234], [235, 231,...  \n",
              "1  [[[194, 192, 171], [194, 192, 171], [193, 192,...  \n",
              "2  [[[57, 27, 3], [63, 33, 9], [63, 32, 7], [57, ...  \n",
              "3  [[[128, 161, 178], [154, 188, 205], [156, 188,...  \n",
              "4  [[[133, 111, 97], [132, 114, 99], [131, 117, 1...  \n",
              "5  [[[43, 43, 43], [42, 42, 42], [41, 41, 41], [4...  \n",
              "6  [[[186, 89, 18], [182, 85, 14], [183, 85, 14],...  \n",
              "7  [[[254, 254, 254], [254, 254, 254], [254, 254,...  \n",
              "8  [[[201, 191, 181], [199, 189, 179], [199, 189,...  \n",
              "9  [[[236, 230, 226], [237, 231, 225], [239, 231,...  "
            ],
            "text/html": [
              "\n",
              "\n",
              "  <div id=\"df-3b919d93-250a-42d9-b5e7-ea388ebee894\">\n",
              "    <div 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>url</th>\n",
              "      <th>prompt</th>\n",
              "      <th>category</th>\n",
              "      <th>image_array</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>https://static.wixstatic.com/media/039af8_8b6f...</td>\n",
              "      <td>meme Bert Is Evil</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[241, 241, 232], [245, 243, 234], [235, 231,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>https://i.imgflip.com/2vo3mr.jpg</td>\n",
              "      <td>meme #FirstWorldProblems</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[194, 192, 171], [194, 192, 171], [193, 192,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>https://i.pinimg.com/200x150/36/c9/d5/36c9d5bb...</td>\n",
              "      <td>meme Keep Calm</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[57, 27, 3], [63, 33, 9], [63, 32, 7], [57, ...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>http://cdn.salemweb.net/godtube/2011/09/16/337...</td>\n",
              "      <td>meme \"Fuck it, we'll do it live!\"</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[128, 161, 178], [154, 188, 205], [156, 188,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>https://thumbs.dreamstime.com/t/le-mari-et-l-%...</td>\n",
              "      <td>meme \"Fuck it, we'll do it live!\"</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[133, 111, 97], [132, 114, 99], [131, 117, 1...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>http://wheresmysammich.com/images/thumbs/47937...</td>\n",
              "      <td>meme Rickrolling</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[43, 43, 43], [42, 42, 42], [41, 41, 41], [4...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>6</th>\n",
              "      <td>https://thumbs.dreamstime.com/z/gor%C4%85ce-ro...</td>\n",
              "      <td>meme American Chopper argument</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[186, 89, 18], [182, 85, 14], [183, 85, 14],...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>7</th>\n",
              "      <td>https://postila.ru/resize?w=460&amp;src=%2Fdata%2F...</td>\n",
              "      <td>meme #Winning</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[254, 254, 254], [254, 254, 254], [254, 254,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>8</th>\n",
              "      <td>https://thumbs.dreamstime.com/z/toskana-k%C3%B...</td>\n",
              "      <td>meme Ceiling Cat</td>\n",
              "      <td>meme</td>\n",
              "      <td>[[[201, 191, 181], [199, 189, 179], [199, 189,...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>9</th>\n",
              "      <td>https://cdn.shopify.com/s/files/1/1377/4645/pr...</td>\n",
              "      <td>toy doll</td>\n",
              "      <td>toys</td>\n",
              "      <td>[[[236, 230, 226], [237, 231, 225], [239, 231,...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-3b919d93-250a-42d9-b5e7-ea388ebee894')\"\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 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "\n",
              "\n",
              "\n",
              "    <div id=\"df-1fcba321-8a27-48c1-a3d9-5915cad3a5ba\">\n",
              "      <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-1fcba321-8a27-48c1-a3d9-5915cad3a5ba')\"\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",
              "    </div>\n",
              "\n",
              "<style>\n",
              "  .colab-df-quickchart {\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-quickchart: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",
              "  [theme=dark] .colab-df-quickchart {\n",
              "    background-color: #3B4455;\n",
              "    fill: #D2E3FC;\n",
              "  }\n",
              "\n",
              "  [theme=dark] .colab-df-quickchart: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",
              "      async function quickchart(key) {\n",
              "        const containerElement = document.querySelector('#' + key);\n",
              "        const charts = await google.colab.kernel.invokeFunction(\n",
              "            'suggestCharts', [key], {});\n",
              "      }\n",
              "    </script>\n",
              "\n",
              "      <script>\n",
              "\n",
              "function displayQuickchartButton(domScope) {\n",
              "  let quickchartButtonEl =\n",
              "    domScope.querySelector('#df-1fcba321-8a27-48c1-a3d9-5915cad3a5ba button.colab-df-quickchart');\n",
              "  quickchartButtonEl.style.display =\n",
              "    google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "}\n",
              "\n",
              "        displayQuickchartButton(document);\n",
              "      </script>\n",
              "      <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\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",
              "    [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-3b919d93-250a-42d9-b5e7-ea388ebee894 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-3b919d93-250a-42d9-b5e7-ea388ebee894');\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",
              "  </div>\n"
            ]
          },
          "metadata": {},
          "execution_count": 118
        }
      ],
      "source": [
        "dftemp"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ipp1JRuyaCVJ"
      },
      "outputs": [],
      "source": [
        "image_paths = dftemp['image_array'].tolist()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "wQJ03cvbaCRA"
      },
      "outputs": [],
      "source": [
        "preprocess = transforms.Compose([\n",
        "    #transforms.Resize((256, 256)),  # Resize images to a fixed size (e.g., 256x256)\n",
        "    transforms.ToTensor(),          # Convert images to tensors\n",
        "    transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])  # Normalize the tensor\n",
        "])\n",
        "i = 0\n",
        "preprocessed_images = []\n",
        "for image_path in image_paths:\n",
        "    img = image_paths[i]\n",
        "    img = preprocess(img)\n",
        "    preprocessed_images.append(img)\n",
        "    i = i+1\n",
        "\n",
        "\n",
        "stacked_images_tensor = torch.stack(preprocessed_images)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "B_Yh15Qgxt3q",
        "outputId": "04a12a10-7183-4826-ac7a-85b9c9e1cb32"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "10"
            ]
          },
          "metadata": {},
          "execution_count": 121
        }
      ],
      "source": [
        "i"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "aQcnx73myaCm",
        "outputId": "8383f4ba-5967-4e53-f664-62c617e91673"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "torch.Size([10, 3, 224, 224])"
            ]
          },
          "metadata": {},
          "execution_count": 40
        }
      ],
      "source": [
        "stacked_images_tensor.shape"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "stacked_images_tensor.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JFcc5YrYHQq1",
        "outputId": "4822343d-8ac9-429f-c1fb-8a95d27e2ceb"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "torch.Size([10, 3, 224, 224])"
            ]
          },
          "metadata": {},
          "execution_count": 123
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "If2XHm2myaGK",
        "outputId": "61496159-c1bd-4642-841b-9ea7b35ecb74"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([10, 196, 40])\n"
          ]
        }
      ],
      "source": [
        "\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "\n",
        "# Example input image size and patch size\n",
        "img_size = 224\n",
        "patch_size = 16\n",
        "in_channels = 3  # Assuming RGB images\n",
        "\n",
        "# Vision Transformer parameters\n",
        "embed_size = 256\n",
        "num_layers = 6\n",
        "heads = 8\n",
        "dropout = 0.1\n",
        "num_classes = 40  # Replace this with the number of classes in your classification task\n",
        "#num_classes = 142*128\n",
        "max_length = (img_size // patch_size) ** 2 + 1  # +1 for the CLS token\n",
        "\n",
        "# Create 10 random input images as a batch (batch_size=10)\n",
        "batch_size = 10\n",
        "input_images = stacked_images_tensor\n",
        "\n",
        "# Create the VisionTransformer model\n",
        "model = VisionTransformer(\n",
        "    img_size=img_size,\n",
        "    patch_size=patch_size,\n",
        "    in_channels=in_channels,\n",
        "    embed_size=embed_size,\n",
        "    num_layers=num_layers,\n",
        "    heads=heads,\n",
        "    device='cuda' if torch.cuda.is_available() else 'cpu',\n",
        "    forward_expansion=4,  # You can adjust this value if needed\n",
        "    dropout=dropout,\n",
        "    num_classes=num_classes,\n",
        "    max_length=max_length,\n",
        ")\n",
        "\n",
        "\n",
        "# Move the model to the appropriate device (CPU or GPU)\n",
        "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
        "model = model.to(device)\n",
        "\n",
        "# Forward pass to get the output tensor\n",
        "with torch.no_grad():\n",
        "    model.eval()\n",
        "    input_images = input_images.to(device)\n",
        "    image_output_tensor = model(input_images)\n",
        "\n",
        "print(image_output_tensor.shape)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Luy8sqPPBNA7",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8cda88d2-1c60-452e-9d50-c0e2bef4ce71"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[[-0.7285, -0.8557,  1.0983,  ...,  0.1058, -0.0040,  1.4328],\n",
            "         [-0.4563, -0.4228, -0.5972,  ...,  0.1845, -0.8843,  0.2663],\n",
            "         [-0.2212, -0.6337, -0.8496,  ...,  0.1360,  0.0905,  0.1873],\n",
            "         ...,\n",
            "         [ 0.2514, -0.3927, -0.2670,  ...,  0.2568,  0.3627, -0.2634],\n",
            "         [-0.1293, -0.1820,  0.5785,  ...,  1.2420, -0.3475, -0.2039],\n",
            "         [-0.3908, -0.7066, -0.2163,  ...,  0.3405,  0.4511,  0.4615]],\n",
            "\n",
            "        [[-0.8022, -0.9944,  1.1772,  ..., -0.0739, -0.0947,  1.4439],\n",
            "         [-0.5821, -0.3609, -0.5386,  ...,  0.1215, -1.1517,  0.3308],\n",
            "         [-0.5760, -0.3206, -1.0060,  ..., -0.2443, -0.0244,  0.2794],\n",
            "         ...,\n",
            "         [ 0.0846, -0.4143, -0.1419,  ...,  0.0778,  0.1169,  0.0361],\n",
            "         [-0.3477,  0.0871,  0.6260,  ...,  1.2359,  0.2867, -0.0160],\n",
            "         [-0.3261, -0.2824, -0.2798,  ...,  0.3872,  0.6690,  0.4935]],\n",
            "\n",
            "        [[-0.6840, -0.1908,  0.5949,  ..., -0.8000, -0.6878,  0.7443],\n",
            "         [-0.5703,  0.0104, -1.0118,  ..., -0.4883, -1.5771, -0.1151],\n",
            "         [-0.0498, -0.4290, -1.0793,  ..., -0.7991, -0.5368, -0.2971],\n",
            "         ...,\n",
            "         [ 0.8824, -0.7205, -0.1917,  ..., -0.2900,  0.1149,  0.1437],\n",
            "         [-0.0524,  0.5635,  0.7793,  ...,  0.6918, -0.6206, -0.1962],\n",
            "         [ 0.0800, -0.1357, -0.9479,  ..., -0.9445, -0.1770, -0.4517]],\n",
            "\n",
            "        ...,\n",
            "\n",
            "        [[-0.6945, -0.7842,  0.7919,  ...,  0.1209,  0.1987,  1.1627],\n",
            "         [-0.4986, -0.3643, -0.8122,  ...,  0.1375, -0.5808,  0.2173],\n",
            "         [-0.3713, -0.7348, -1.0411,  ...,  0.1974,  0.1958,  0.0668],\n",
            "         ...,\n",
            "         [-0.0772, -0.5992, -0.2587,  ...,  0.8778,  0.5043,  0.3678],\n",
            "         [-0.3405, -0.1635,  0.5167,  ...,  1.1335,  0.1147,  0.0775],\n",
            "         [-0.5033, -0.4600, -0.4409,  ...,  0.4733,  0.4529,  0.2762]],\n",
            "\n",
            "        [[-0.8695, -0.8859,  0.9548,  ...,  0.0706,  0.0406,  1.3336],\n",
            "         [-0.5668, -0.2623, -0.7616,  ...,  0.2724, -1.0043,  0.3002],\n",
            "         [-0.3339, -0.6232, -1.0976,  ...,  0.1835,  0.1279,  0.1775],\n",
            "         ...,\n",
            "         [-0.1183, -0.2622, -0.6470,  ...,  0.1379,  0.3461,  0.0589],\n",
            "         [-0.5918,  0.4143,  0.1834,  ...,  0.3471, -0.1891, -0.2661],\n",
            "         [-0.5913,  0.0145, -1.0276,  ..., -0.4495,  0.3975, -0.1174]],\n",
            "\n",
            "        [[-0.7747, -0.9535,  0.9363,  ...,  0.1847,  0.0326,  1.2879],\n",
            "         [-0.5638, -0.4377, -0.5190,  ...,  0.2987, -0.7916,  0.3592],\n",
            "         [-0.3248, -0.6964, -0.7852,  ...,  0.3154,  0.1632,  0.3020],\n",
            "         ...,\n",
            "         [ 0.2801, -0.6077, -0.4757,  ...,  0.7251,  0.6280, -0.0069],\n",
            "         [-0.5629,  0.3800,  0.6143,  ...,  1.4526, -0.2311,  0.0646],\n",
            "         [-0.7617, -0.5448, -0.4748,  ...,  0.2130,  0.4391,  0.5093]]])\n"
          ]
        }
      ],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Be2zT5jgyBkn"
      },
      "outputs": [],
      "source": [
        "def image_tens():\n",
        "    # Move the model to the appropriate device (CPU or GPU)\n",
        "    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
        "    model = model.to(device)\n",
        "\n",
        "    # Forward pass to get the output tensor\n",
        "    with torch.no_grad():\n",
        "        model.eval()  # Set the model to evaluation mode (important for dropout layers)\n",
        "        input_images = input_images.to(device)\n",
        "        image_output_tensor = model(input_images)\n",
        "\n",
        "    #print(image_output_tensor.shape)  # Output should be torch.Size([10, 10, 256])\n",
        "    return(image_output_tensor)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bO0fp8zt3eki"
      },
      "source": [
        "* For text = text_output_tensor\n",
        "\n",
        "\n",
        "* For images = image_output_tensor"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "-uvpvB0JaCO0"
      },
      "outputs": [],
      "source": [
        "def resize_tensor(text_output_tensor):\n",
        "  x_size = text_output_tensor.size(1)\n",
        "  y_size = text_output_tensor.size(2)\n",
        "\n",
        "  z_size = x_size * y_size\n",
        "\n",
        "  text_output_tensor = text_output_tensor.view(10, z_size)\n",
        "  return text_output_tensor"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BMTxp8x4aCLu",
        "outputId": "460899fd-133e-4d5e-edcf-a4e90f796f7e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([10, 20, 392])\n",
            "torch.Size([10, 196, 40])\n"
          ]
        }
      ],
      "source": [
        "print(text_output_tensor.shape)\n",
        "print(image_output_tensor.shape)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "YKh0KdXeIMJk"
      },
      "outputs": [],
      "source": [
        "temp = 7840\n",
        "text_output_tensor1 = text_output_tensor.view(10, temp)\n",
        "image_output_tensor1 = image_output_tensor.view(10, temp)\n",
        "#image_output_tensor = image_output_tensor.transpose(0, 1)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "RNc6N1Y_KAZM",
        "outputId": "0a8869b8-4124-432c-f78e-509d0ed3af48"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([10, 7840])\n",
            "torch.Size([10, 7840])\n"
          ]
        }
      ],
      "source": [
        "print(text_output_tensor1.shape)\n",
        "print(image_output_tensor1.shape)"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "text_output_tensor1"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qeaSgRf3-Cdl",
        "outputId": "1fba6d16-71dc-45c5-9a7d-7669cb1cdb04"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[ 0.4837, -0.4086,  0.1829,  ..., -0.8030, -0.1954, -0.8844],\n",
              "        [-2.9763, -0.3849, -0.3297,  ..., -0.0000, -2.2630, -0.1113],\n",
              "        [ 0.0554,  0.6807,  0.0592,  ..., -0.4407, -0.0359,  0.8325],\n",
              "        ...,\n",
              "        [ 0.8149, -0.4646,  0.9473,  ...,  3.0940,  0.0256, -2.8862],\n",
              "        [ 0.4620, -0.5134,  0.3801,  ...,  0.5175,  0.1402,  0.0000],\n",
              "        [ 0.0000,  0.5092,  0.4142,  ...,  0.5119, -0.8898,  0.7881]],\n",
              "       grad_fn=<ViewBackward0>)"
            ]
          },
          "metadata": {},
          "execution_count": 131
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "image_output_tensor1"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bQjuJ8yz-E0D",
        "outputId": "0b3c99be-a620-49d1-c305-cc6679563cac"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[-0.7285, -0.8557,  1.0983,  ...,  0.3405,  0.4511,  0.4615],\n",
              "        [-0.8022, -0.9944,  1.1772,  ...,  0.3872,  0.6690,  0.4935],\n",
              "        [-0.6840, -0.1908,  0.5949,  ..., -0.9445, -0.1770, -0.4517],\n",
              "        ...,\n",
              "        [-0.6945, -0.7842,  0.7919,  ...,  0.4733,  0.4529,  0.2762],\n",
              "        [-0.8695, -0.8859,  0.9548,  ..., -0.4495,  0.3975, -0.1174],\n",
              "        [-0.7747, -0.9535,  0.9363,  ...,  0.2130,  0.4391,  0.5093]])"
            ]
          },
          "metadata": {},
          "execution_count": 132
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "smqL3BjuJjEi"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "import torch.nn.functional as F"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "j577eFcDMPz3"
      },
      "outputs": [],
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "import torch.nn.functional as F\n",
        "\n",
        "# Assuming you have a custom image and text encoder models\n",
        "# Replace these with your actual models\n",
        "#image_encoder = model\n",
        "#text_encoder = encoder\n",
        "\n",
        "class CLIPModel(nn.Module):\n",
        "    def __init__(self):\n",
        "        super().__init__()\n",
        "        self.image_encoder = model\n",
        "        self.text_encoder = encoder\n",
        "\n",
        "    def forward(self, input_images, input_tensor, mask):\n",
        "        # Getting Image and Text Features\n",
        "        image_features = self.image_encoder(input_images)\n",
        "        text_features = self.text_encoder(input_tensor, mask)\n",
        "\n",
        "        batch_size = input_images.size(0)\n",
        "        temp = 7840\n",
        "        text_features_flat = text_features.view(batch_size, temp)\n",
        "        image_features_flat = image_features.view(batch_size, temp)\n",
        "\n",
        "        # to trnspose\n",
        "        image_features_flat = image_features_flat.transpose(0, 1)\n",
        "\n",
        "        # Matrix multiplication\n",
        "        #predicted_matrix = torch.matmul(text_features_flat, image_features_flat)\n",
        "        #print(predicted_matrix.shape)\n",
        "        #return predicted_matrix\n",
        "        return text_features_flat, image_features_flat\n"
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "MMmJ0_6a-exU"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "0IhBYZIHdcf3"
      },
      "outputs": [],
      "source": [
        "def pred_matrix(text_features_flat, image_features_flat):\n",
        "  predicted_matrix = torch.matmul(text_features_flat, image_features_flat)\n",
        "  #print(predicted_matrix)\n",
        "  return predicted_matrix"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "cnxVpbkRfUm8"
      },
      "outputs": [],
      "source": [
        "my_model = CLIPModel().to(device)\n",
        "#predicted_matrix = my_model(input_images, input_tensor, mask)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "qrV_dY2xerP3"
      },
      "outputs": [],
      "source": [
        "#criterion = nn.CrossEntropyLoss()\n",
        "criterion = nn.MSELoss(size_average=None, reduce=None, reduction='mean')\n",
        "optimizer = optim.Adam(my_model.parameters(), lr=0.01)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "jdNbNt6VhoP-"
      },
      "outputs": [],
      "source": [
        "epochs = 15"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "batch_size"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CirPnUzB_ImI",
        "outputId": "089e1639-a042-42ee-9001-44af1e5f4e6f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "10"
            ]
          },
          "metadata": {},
          "execution_count": 138
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "vEjtA0rMh5sv"
      },
      "outputs": [],
      "source": [
        "targets = 1000*torch.eye(batch_size)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "v--nLmI870rm",
        "outputId": "236bc81e-b899-4ce8-fa3d-737b4386db0c"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[1000.,    0.,    0.,    0.,    0.,    0.,    0.,    0.,    0.,    0.],\n",
              "        [   0., 1000.,    0.,    0.,    0.,    0.,    0.,    0.,    0.,    0.],\n",
              "        [   0.,    0., 1000.,    0.,    0.,    0.,    0.,    0.,    0.,    0.],\n",
              "        [   0.,    0.,    0., 1000.,    0.,    0.,    0.,    0.,    0.,    0.],\n",
              "        [   0.,    0.,    0.,    0., 1000.,    0.,    0.,    0.,    0.,    0.],\n",
              "        [   0.,    0.,    0.,    0.,    0., 1000.,    0.,    0.,    0.,    0.],\n",
              "        [   0.,    0.,    0.,    0.,    0.,    0., 1000.,    0.,    0.,    0.],\n",
              "        [   0.,    0.,    0.,    0.,    0.,    0.,    0., 1000.,    0.,    0.],\n",
              "        [   0.,    0.,    0.,    0.,    0.,    0.,    0.,    0., 1000.,    0.],\n",
              "        [   0.,    0.,    0.,    0.,    0.,    0.,    0.,    0.,    0., 1000.]])"
            ]
          },
          "metadata": {},
          "execution_count": 140
        }
      ],
      "source": [
        "targets"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "V2rkynKNiaaS"
      },
      "outputs": [],
      "source": [
        "my_data = (input_images, input_tensor, mask)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ePOE1EEUjH3l"
      },
      "outputs": [],
      "source": [
        "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4pUOuFqE7B2X",
        "outputId": "39ade7d1-ae14-42f4-a340-4fcd38cae5c2"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[-1.7129,  0.9559,  0.6576,  ..., -0.6016, -0.2351, -0.0040],\n",
            "        [-1.7129,  0.9559,  0.6576,  ..., -0.6016, -0.2351, -0.0040],\n",
            "        [-1.7129,  0.9559,  0.6576,  ..., -0.6016, -0.2351, -0.0040],\n",
            "        ...,\n",
            "        [-1.7129,  0.9559,  0.6576,  ..., -0.6016, -0.2351, -0.0040],\n",
            "        [-1.7129,  0.9559,  0.6576,  ..., -0.6016, -0.2351, -0.0040],\n",
            "        [-1.7129,  0.9559,  0.6576,  ..., -0.6016, -0.2351, -0.0040]],\n",
            "       grad_fn=<ViewBackward0>) tensor([[-0.5899, -0.5899, -0.5899,  ..., -0.5899, -0.5899, -0.5899],\n",
            "        [ 0.4548,  0.4548,  0.4548,  ...,  0.4548,  0.4548,  0.4548],\n",
            "        [-0.4351, -0.4351, -0.4351,  ..., -0.4351, -0.4351, -0.4351],\n",
            "        ...,\n",
            "        [-0.3597, -0.3597, -0.3597,  ..., -0.3597, -0.3597, -0.3597],\n",
            "        [ 0.3833,  0.3833,  0.3833,  ...,  0.3833,  0.3833,  0.3833],\n",
            "        [ 1.0119,  1.0119,  1.0119,  ...,  1.0119,  1.0119,  1.0119]],\n",
            "       grad_fn=<TransposeBackward0>)\n",
            "Epoch: 0\n",
            "tensor([[-1.9772,  0.8829,  0.8450,  ..., -0.5467, -0.1995,  0.2872],\n",
            "        [-1.9759,  0.9029,  0.8407,  ..., -0.5387, -0.1453,  0.3035],\n",
            "        [-1.8761,  0.7453,  0.8999,  ..., -0.5410, -0.1437,  0.2978],\n",
            "        ...,\n",
            "        [-1.9671,  0.7679,  0.8841,  ..., -0.5182, -0.1312,  0.0000],\n",
            "        [-1.9467,  0.8279,  0.8867,  ..., -0.5439, -0.1548,  0.2938],\n",
            "        [-0.0000,  0.8173,  0.9184,  ..., -0.5611, -0.0000,  0.3134]],\n",
            "       grad_fn=<ViewBackward0>) tensor([[-0.3324, -0.3054, -0.5975,  ..., -0.7000, -0.2578, -0.5354],\n",
            "        [ 0.4729,  0.1847,  0.2308,  ...,  0.4854, -0.2446,  0.0072],\n",
            "        [-0.2859,  0.0438,  0.0517,  ...,  0.0696,  0.1965,  0.0461],\n",
            "        ...,\n",
            "        [-1.2666, -1.2716, -1.1010,  ..., -0.7629, -1.0061, -1.2891],\n",
            "        [ 0.6134,  0.2179,  0.4977,  ...,  0.2082,  0.3126,  0.0831],\n",
            "        [ 1.0468,  0.6482,  0.5078,  ...,  1.4015,  0.8400,  0.8436]],\n",
            "       grad_fn=<TransposeBackward0>)\n",
            "Epoch: 1\n",
            "tensor([[-1.9624,  0.6552,  0.8912,  ..., -0.4410, -0.1202,  0.4408],\n",
            "        [-1.9500,  0.6271,  0.9618,  ..., -0.4407, -0.0728,  0.4347],\n",
            "        [-1.9703,  0.6065,  0.9268,  ..., -0.0000, -0.1071,  0.4532],\n",
            "        ...,\n",
            "        [-1.8803,  0.5477,  0.9907,  ..., -0.4238, -0.0000,  0.4456],\n",
            "        [-1.9628,  0.6870,  0.9156,  ..., -0.4362, -0.0815,  0.4385],\n",
            "        [-0.0000,  0.6152,  0.9279,  ..., -0.4706, -0.0000,  0.4329]],\n",
            "       grad_fn=<ViewBackward0>) tensor([[-0.3108,  0.1227, -0.2912,  ..., -0.0965,  0.1244, -0.2214],\n",
            "        [-0.2898,  0.1152,  0.4675,  ...,  0.3808, -0.1050, -0.2870],\n",
            "        [ 0.6202,  0.2165,  0.4444,  ...,  0.8221,  1.0291,  0.7482],\n",
            "        ...,\n",
            "        [-1.7159, -1.9982, -1.5464,  ..., -1.7222, -2.0766, -1.6951],\n",
            "        [ 0.0953,  0.5903, -0.2361,  ..., -0.0181,  0.4208,  0.1612],\n",
            "        [ 0.8641,  0.9553,  0.8030,  ...,  0.9970,  0.2234,  0.2214]],\n",
            "       grad_fn=<TransposeBackward0>)\n",
            "Epoch: 2\n",
            "tensor([[-1.9511,  0.4620,  0.9559,  ..., -0.3328, -0.0450,  0.5444],\n",
            "        [-1.9469,  0.4988,  0.0000,  ..., -0.3205, -0.0071,  0.5535],\n",
            "        [-1.9582,  0.0000,  0.0000,  ..., -0.3455, -0.0142,  0.5433],\n",
            "        ...,\n",
            "        [-1.9269,  0.4304,  1.0212,  ..., -0.3372, -0.0058,  0.5461],\n",
            "        [-0.0000,  0.5145,  0.9295,  ..., -0.3459, -0.0000,  0.5509],\n",
            "        [-1.9476,  0.5392,  0.9383,  ..., -0.3598, -0.0103,  0.5125]],\n",
            "       grad_fn=<ViewBackward0>) tensor([[-0.0589,  0.5889,  0.0905,  ..., -0.4095,  0.2501, -0.0045],\n",
            "        [-0.4518, -0.1765, -0.4647,  ...,  0.0382, -0.3794, -0.0624],\n",
            "        [ 1.0657,  1.0830,  1.3371,  ...,  0.6857,  1.2311,  1.1095],\n",
            "        ...,\n",
            "        [-1.5814, -2.5676, -2.6577,  ..., -2.6072, -2.2711, -2.1041],\n",
            "        [ 0.2494,  0.0879,  0.0858,  ..., -0.0584,  0.2649,  0.6214],\n",
            "        [ 0.1037,  0.2424,  0.8352,  ...,  0.3654,  0.5801,  0.5145]],\n",
            "       grad_fn=<TransposeBackward0>)\n",
            "Epoch: 3\n",
            "tensor([[-0.0000,  0.3919,  0.9866,  ..., -0.2944, -0.0291,  0.6070],\n",
            "        [-1.9413,  0.4098,  0.9748,  ..., -0.2719, -0.0515,  0.5909],\n",
            "        [-1.9503,  0.4180,  0.9664,  ..., -0.3094, -0.0323,  0.0000],\n",
            "        ...,\n",
            "        [-1.9520,  0.4010,  0.9747,  ..., -0.2960, -0.0396,  0.6078],\n",
            "        [-1.9469,  0.4087,  0.9568,  ..., -0.2853, -0.0339,  0.6103],\n",
            "        [-1.9297,  0.3715,  1.0006,  ..., -0.3123, -0.0496,  0.5979]],\n",
            "       grad_fn=<ViewBackward0>) tensor([[ 0.9668,  0.6606,  0.4049,  ...,  0.8070,  0.7679,  0.4364],\n",
            "        [-0.3137, -0.6445, -0.2910,  ..., -0.5469, -0.7036, -0.4613],\n",
            "        [ 1.1822,  1.9461,  1.7659,  ...,  1.3941,  1.3502,  1.5955],\n",
            "        ...,\n",
            "        [-3.3620, -3.2270, -3.1706,  ..., -2.9620, -3.0218, -3.0914],\n",
            "        [ 0.1657,  0.0747,  0.2016,  ...,  0.2781, -0.3038, -0.0130],\n",
            "        [ 0.4270, -0.0436,  0.3201,  ...,  0.8487,  0.9307, -0.0737]],\n",
            "       grad_fn=<TransposeBackward0>)\n",
            "Epoch: 4\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "CLIPModel(\n",
              "  (image_encoder): VisionTransformer(\n",
              "    (patch_embedding): PatchEmbedding(\n",
              "      (projection): Conv2d(3, 256, kernel_size=(16, 16), stride=(16, 16))\n",
              "    )\n",
              "    (position_embedding): Embedding(197, 256)\n",
              "    (layers): ModuleList(\n",
              "      (0-5): 6 x TransformerBlock(\n",
              "        (attention): SelfAttention(\n",
              "          (values): Linear(in_features=256, out_features=256, bias=True)\n",
              "          (keys): Linear(in_features=256, out_features=256, bias=True)\n",
              "          (queries): Linear(in_features=256, out_features=256, bias=True)\n",
              "          (fc_out): Linear(in_features=256, out_features=256, bias=True)\n",
              "        )\n",
              "        (norm1): LayerNorm((256,), eps=1e-05, elementwise_affine=True)\n",
              "        (norm2): LayerNorm((256,), eps=1e-05, elementwise_affine=True)\n",
              "        (feed_forward): Sequential(\n",
              "          (0): Linear(in_features=256, out_features=1024, bias=True)\n",
              "          (1): ReLU()\n",
              "          (2): Linear(in_features=1024, out_features=256, bias=True)\n",
              "        )\n",
              "        (dropout): Dropout(p=0.1, inplace=False)\n",
              "      )\n",
              "    )\n",
              "    (classification_head): Linear(in_features=256, out_features=40, bias=True)\n",
              "    (dropout): Dropout(p=0.1, inplace=False)\n",
              "  )\n",
              "  (text_encoder): Encoder(\n",
              "    (word_embedding): Embedding(1000, 392)\n",
              "    (position_embedding): Embedding(150, 392)\n",
              "    (layers): ModuleList(\n",
              "      (0-5): 6 x TransformerBlock(\n",
              "        (attention): SelfAttention(\n",
              "          (values): Linear(in_features=392, out_features=392, bias=True)\n",
              "          (keys): Linear(in_features=392, out_features=392, bias=True)\n",
              "          (queries): Linear(in_features=392, out_features=392, bias=True)\n",
              "          (fc_out): Linear(in_features=392, out_features=392, bias=True)\n",
              "        )\n",
              "        (norm1): LayerNorm((392,), eps=1e-05, elementwise_affine=True)\n",
              "        (norm2): LayerNorm((392,), eps=1e-05, elementwise_affine=True)\n",
              "        (feed_forward): Sequential(\n",
              "          (0): Linear(in_features=392, out_features=1568, bias=True)\n",
              "          (1): ReLU()\n",
              "          (2): Linear(in_features=1568, out_features=392, bias=True)\n",
              "        )\n",
              "        (dropout): Dropout(p=0.1, inplace=False)\n",
              "      )\n",
              "    )\n",
              "    (dropout): Dropout(p=0.1, inplace=False)\n",
              "  )\n",
              ")"
            ]
          },
          "metadata": {},
          "execution_count": 145
        }
      ],
      "source": [
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "\n",
        "\n",
        "# Define your cross_entropy function\n",
        "def cross_entropy(preds, targets, reduction='mean'):\n",
        "    log_softmax = nn.LogSoftmax(dim=-1)\n",
        "    loss = (-targets * log_softmax(preds)).sum(1)\n",
        "    if reduction == \"mean\":\n",
        "        return loss.mean()\n",
        "    elif reduction == \"sum\":\n",
        "        return loss.sum()\n",
        "    else:\n",
        "        return loss\n",
        "\n",
        "criterion = nn.MSELoss()\n",
        "\n",
        "# Assuming you have a training loop\n",
        "epochs = 5\n",
        "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
        "my_model.to(device)\n",
        "\n",
        "for epoch in range(epochs):\n",
        "        #my_model.train()  # Set the model to training mode\n",
        "    #for batch_idx, (input_images, input_tensor, mask, targets) in enumerate(train_loader):\n",
        "        # Move data to the device\n",
        "        input_images = input_images.to(device)\n",
        "        input_tensor = input_tensor.to(device)\n",
        "        mask = mask.to(device)\n",
        "        targets = targets.to(device)\n",
        "\n",
        "        # Forward pass to get predictions\n",
        "        a, b = my_model(input_images, input_tensor, mask)\n",
        "        print(a,b)\n",
        "        scores = pred_matrix(a, b)\n",
        "\n",
        "        ground_truth_matrix = 1000*torch.eye(10)  # Identity matrix of size (10, 10)\n",
        "\n",
        "        # Calculate the mean squared error loss\n",
        "        #loss = F.mse_loss(scores, ground_truth_matrix)\n",
        "        # Compute the cross-entropy loss\n",
        "        loss = cross_entropy(scores, targets, reduction='mean')\n",
        "        #loss = criterion(scores, targets)\n",
        "        #print(loss.item())  # Print the loss for each batch\n",
        "\n",
        "\n",
        "        # Perform backpropagation and update model parameters\n",
        "        optimizer.zero_grad()\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "\n",
        "        print('Epoch:', epoch)\n",
        "        #print(scores)\n",
        "        my_model.train()\n",
        "\n",
        "# After training, you can switch the model to evaluation mode before making predictions.\n",
        "my_model.eval()\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WyCEq-r67v62",
        "outputId": "62fa0e81-9349-40ac-92e9-0a6857afc454"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[1330.4524, 1274.8289, 1288.9905, 1289.8621, 1301.0552, 1252.7659,\n",
            "         1291.7332, 1276.0105, 1264.9371, 1336.3562],\n",
            "        [1344.9308, 1281.1897, 1283.5940, 1287.3844, 1304.9342, 1247.2356,\n",
            "         1301.3933, 1285.5803, 1267.9680, 1366.5996],\n",
            "        [1319.8713, 1259.2062, 1264.8359, 1249.7379, 1276.5583, 1212.7511,\n",
            "         1263.3159, 1259.3708, 1257.9832, 1323.8003],\n",
            "        [1335.0276, 1250.8412, 1257.6436, 1256.4061, 1280.3801, 1199.5854,\n",
            "         1277.3486, 1236.6224, 1244.0259, 1319.3960],\n",
            "        [1344.1985, 1272.7249, 1294.4536, 1287.6489, 1313.3770, 1252.3737,\n",
            "         1305.6685, 1269.3167, 1273.0503, 1351.0890],\n",
            "        [1332.3577, 1269.3467, 1283.8257, 1277.3861, 1306.1401, 1251.7676,\n",
            "         1314.3220, 1266.4905, 1270.1614, 1347.3773],\n",
            "        [1346.6340, 1284.2849, 1295.5767, 1301.6548, 1312.4468, 1268.9785,\n",
            "         1302.6438, 1293.2473, 1283.9990, 1361.3018],\n",
            "        [1345.8276, 1286.3767, 1293.3661, 1291.6174, 1302.1572, 1259.0327,\n",
            "         1283.1123, 1273.8168, 1281.8311, 1359.6604],\n",
            "        [1289.0317, 1240.5302, 1256.6151, 1232.6497, 1267.3722, 1194.9106,\n",
            "         1268.1046, 1224.6877, 1238.5276, 1315.7744],\n",
            "        [1389.7498, 1316.6180, 1325.9254, 1327.0950, 1332.3433, 1284.7424,\n",
            "         1326.1130, 1300.9832, 1297.7642, 1381.2085]], grad_fn=<MmBackward0>)\n"
          ]
        }
      ],
      "source": [
        "print(scores)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "DFQPlnky6XZ0"
      },
      "outputs": [],
      "source": [
        "def cross_entropy(preds, targets, reduction='none'):\n",
        "    log_softmax = nn.LogSoftmax(dim=-1)\n",
        "    loss = (-targets * log_softmax(preds)).sum(1)\n",
        "    if reduction == \"none\":\n",
        "        return loss\n",
        "    elif reduction == \"mean\":\n",
        "        return loss.mean()"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "mask.shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "v28fKwV-RGYf",
        "outputId": "411fbc57-d2d3-42fe-a7e9-e93d0709ca62"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "torch.Size([10, 1, 1, 20])"
            ]
          },
          "metadata": {},
          "execution_count": 65
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "a"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "RCZBqu-PFVYj",
        "outputId": "6066938d-f37a-443a-a7a4-41c4d7c701fe"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[-2.1907, -0.1324,  0.0000,  ..., -1.8535, -0.2541,  2.1599],\n",
              "        [-2.1869, -0.1375,  0.4212,  ..., -1.8523, -0.2542,  2.1615],\n",
              "        [-2.1863, -0.1328,  0.4226,  ..., -1.8506, -0.2522,  2.1628],\n",
              "        ...,\n",
              "        [-2.1891, -0.1362,  0.4219,  ..., -1.8543, -0.2541,  2.1592],\n",
              "        [-2.1907, -0.1296,  0.0000,  ..., -0.0000, -0.2437,  2.1611],\n",
              "        [-2.1868, -0.1295,  0.4134,  ..., -1.8474, -0.2528,  2.1578]],\n",
              "       grad_fn=<ViewBackward0>)"
            ]
          },
          "metadata": {},
          "execution_count": 66
        }
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "3DmCPfPPhkBX"
      },
      "outputs": [],
      "source": [
        "a, b = my_model(input_images, input_tensor, mask)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "B6NUaE00hkEB",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "eaaed593-a8a2-4cfc-c1e9-35c5f7fb9b9d"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[-1.9789, -0.3080,  0.5120,  ..., -1.5675, -0.0826,  1.9187],\n",
              "        [-1.9789, -0.3080,  0.5120,  ..., -1.5675, -0.0826,  1.9187],\n",
              "        [-1.9789, -0.3080,  0.5120,  ..., -1.5675, -0.0826,  1.9187],\n",
              "        ...,\n",
              "        [-1.9789, -0.3080,  0.5120,  ..., -1.5675, -0.0826,  1.9187],\n",
              "        [-1.9789, -0.3080,  0.5120,  ..., -1.5675, -0.0826,  1.9187],\n",
              "        [-1.9789, -0.3080,  0.5120,  ..., -1.5675, -0.0826,  1.9187]],\n",
              "       grad_fn=<ViewBackward0>)"
            ]
          },
          "metadata": {},
          "execution_count": 77
        }
      ],
      "source": [
        "a"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(c[1])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 166
        },
        "id": "sM1dWUqTEkN0",
        "outputId": "1f050794-a8ae-4701-9f70-b22e05e2c0f2"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "error",
          "ename": "NameError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-69-8a21c57dcc0d>\u001b[0m in \u001b[0;36m<cell line: 1>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mc\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m: name 'c' is not defined"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "input_images[0]"
      ],
      "metadata": {
        "id": "EYALnd1T8Juy"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "SimN2wTghkKo"
      },
      "outputs": [],
      "source": [
        " temp"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "IOqE2G_HhaHQ"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "gLsH9TKDhaLB"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "OCTAO5EZVOl6"
      },
      "outputs": [],
      "source": [
        "first_batch = input_images[:1, :, :, :]"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ciTcZQBXVOoU"
      },
      "outputs": [],
      "source": [
        "first_batch.shape"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "DLZin-LmVOvu",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "c207f1e6-aedb-4db5-84f4-08e0f8d7c313"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[[[ 2.0092,  2.0777,  1.9064,  ...,  1.3070,  1.2385,  1.2728],\n",
              "          [ 1.8037,  1.9064,  1.8379,  ...,  1.1872,  1.2385,  1.2385],\n",
              "          [ 1.6495,  1.7865,  1.7865,  ...,  1.2214,  1.2385,  1.2043],\n",
              "          ...,\n",
              "          [ 1.6838,  1.5125,  1.3755,  ...,  2.0605,  2.0605,  2.0605],\n",
              "          [ 1.7523,  1.6838,  1.6495,  ...,  2.0605,  2.0605,  2.0605],\n",
              "          [ 1.8893,  1.8379,  1.7865,  ...,  2.0605,  2.0605,  2.0605]],\n",
              "\n",
              "         [[ 2.1835,  2.2185,  2.0084,  ...,  1.3081,  1.2381,  1.2556],\n",
              "          [ 1.9909,  2.0434,  1.9384,  ...,  1.1681,  1.2381,  1.2381],\n",
              "          [ 1.7983,  1.9034,  1.8683,  ...,  1.1856,  1.2031,  1.1506],\n",
              "          ...,\n",
              "          [ 1.9209,  1.7458,  1.6232,  ...,  2.2360,  2.2360,  2.2360],\n",
              "          [ 2.0084,  1.9384,  1.9034,  ...,  2.2360,  2.2360,  2.2360],\n",
              "          [ 2.1310,  2.0784,  2.0259,  ...,  2.2360,  2.2360,  2.2360]],\n",
              "\n",
              "         [[ 2.2391,  2.2740,  2.0474,  ...,  1.2457,  1.1934,  1.2282],\n",
              "          [ 2.0474,  2.0997,  1.9951,  ...,  1.0888,  1.1759,  1.1934],\n",
              "          [ 1.8905,  1.9951,  1.9254,  ...,  1.1062,  1.1237,  1.0888],\n",
              "          ...,\n",
              "          [ 2.2914,  2.1171,  1.9777,  ...,  2.4483,  2.4483,  2.4483],\n",
              "          [ 2.3088,  2.2566,  2.2217,  ...,  2.4483,  2.4483,  2.4483],\n",
              "          [ 2.4134,  2.3786,  2.3088,  ...,  2.4483,  2.4483,  2.4483]]],\n",
              "\n",
              "\n",
              "        [[[ 1.2043,  1.2043,  1.1872,  ...,  0.9988,  1.0331,  0.9988],\n",
              "          [ 1.1358,  1.1358,  1.1358,  ...,  0.9988,  1.0502,  1.0673],\n",
              "          [ 1.1187,  1.1358,  1.1529,  ...,  0.9817,  1.0159,  1.0331],\n",
              "          ...,\n",
              "          [ 1.0159,  0.1426,  0.4508,  ...,  1.8379,  1.4783,  0.9474],\n",
              "          [ 1.3242,  0.3481,  0.1426,  ...,  1.7523,  1.4612,  1.0502],\n",
              "          [ 1.6838,  1.3584,  0.8961,  ...,  1.5297,  1.5297,  1.4783]],\n",
              "\n",
              "         [[ 1.3256,  1.3256,  1.3256,  ...,  1.1331,  1.1856,  1.1681],\n",
              "          [ 1.3081,  1.3081,  1.3081,  ...,  1.1331,  1.1506,  1.1506],\n",
              "          [ 1.3606,  1.3606,  1.3606,  ...,  1.1155,  1.1331,  1.1681],\n",
              "          ...,\n",
              "          [ 1.2206,  0.3102,  0.6078,  ...,  1.7108,  1.4132,  0.8880],\n",
              "          [ 1.4482,  0.4678,  0.2577,  ...,  1.5532,  1.2906,  0.9055],\n",
              "          [ 1.6583,  1.3431,  0.8704,  ...,  1.2381,  1.2381,  1.2031]],\n",
              "\n",
              "         [[ 1.1759,  1.1759,  1.1759,  ...,  0.8274,  0.8622,  0.8274],\n",
              "          [ 1.1759,  1.1759,  1.1585,  ...,  0.8448,  0.8797,  0.8797],\n",
              "          [ 1.2108,  1.2108,  1.2108,  ...,  0.8622,  0.8971,  0.9319],\n",
              "          ...,\n",
              "          [ 1.2805,  0.4614,  0.8797,  ...,  0.8622,  0.6879,  0.2522],\n",
              "          [ 1.4200,  0.4265,  0.3045,  ...,  0.5311,  0.3045, -0.0441],\n",
              "          [ 1.3328,  0.9145,  0.4962,  ...,  0.1825,  0.1651,  0.1825]]],\n",
              "\n",
              "\n",
              "        [[[-1.1418, -1.0390, -1.0390,  ..., -1.1760, -1.2617, -1.4500],\n",
              "          [-1.0219, -0.9534, -0.9534,  ..., -1.1589, -1.2274, -1.3987],\n",
              "          [-0.9020, -0.8507, -0.8335,  ..., -1.1418, -1.2103, -1.3473],\n",
              "          ...,\n",
              "          [-0.7137, -0.5596, -0.4397,  ..., -0.7993, -0.9877, -1.2959],\n",
              "          [-0.7650, -0.5767, -0.4054,  ..., -0.7137, -0.9192, -1.2959],\n",
              "          [-0.8164, -0.6109, -0.3712,  ..., -0.6281, -0.8335, -1.2445]],\n",
              "\n",
              "         [[-1.5630, -1.4580, -1.4755,  ..., -1.6681, -1.7031, -1.8782],\n",
              "          [-1.4405, -1.3704, -1.3880,  ..., -1.6331, -1.6681, -1.8256],\n",
              "          [-1.3179, -1.2654, -1.2829,  ..., -1.6155, -1.6681, -1.7731],\n",
              "          ...,\n",
              "          [-1.3354, -1.1604, -1.0728,  ..., -1.3704, -1.5280, -1.8081],\n",
              "          [-1.3880, -1.1779, -1.0378,  ..., -1.2654, -1.4580, -1.8081],\n",
              "          [-1.4405, -1.2129, -1.0203,  ..., -1.1604, -1.3704, -1.7731]],\n",
              "\n",
              "         [[-1.7522, -1.6476, -1.6824,  ..., -1.6302, -1.6650, -1.8044],\n",
              "          [-1.6302, -1.5604, -1.5953,  ..., -1.6127, -1.6302, -1.7522],\n",
              "          [-1.5081, -1.4559, -1.4733,  ..., -1.5953, -1.6302, -1.7173],\n",
              "          ...,\n",
              "          [-1.5430, -1.4036, -1.3164,  ..., -1.4907, -1.5953, -1.7870],\n",
              "          [-1.5953, -1.4210, -1.2816,  ..., -1.3687, -1.5256, -1.8044],\n",
              "          [-1.6476, -1.4559, -1.2641,  ..., -1.2641, -1.4384, -1.8044]]],\n",
              "\n",
              "\n",
              "        ...,\n",
              "\n",
              "\n",
              "        [[[ 2.2318,  2.2318,  2.2318,  ...,  1.2557,  1.2557,  1.2214],\n",
              "          [ 2.2318,  2.2318,  2.2318,  ...,  1.2557,  1.2557,  1.2214],\n",
              "          [ 2.2318,  2.2318,  2.2318,  ...,  1.2557,  1.2557,  1.2385],\n",
              "          ...,\n",
              "          [ 2.2147,  2.2147,  2.2318,  ...,  1.3584,  1.3584,  1.3242],\n",
              "          [ 2.2147,  2.2147,  2.2318,  ...,  1.3755,  1.3584,  1.3242],\n",
              "          [ 2.2147,  2.2147,  2.2318,  ...,  1.3755,  1.3584,  1.3242]],\n",
              "\n",
              "         [[ 2.4111,  2.4111,  2.4111,  ...,  2.0084,  2.0609,  2.0959],\n",
              "          [ 2.4111,  2.4111,  2.4111,  ...,  2.0084,  2.0609,  2.0959],\n",
              "          [ 2.4111,  2.4111,  2.4111,  ...,  2.0084,  2.0609,  2.0959],\n",
              "          ...,\n",
              "          [ 2.3936,  2.4111,  2.3936,  ...,  2.0434,  2.0959,  2.1134],\n",
              "          [ 2.3936,  2.4111,  2.3936,  ...,  2.0609,  2.0959,  2.1134],\n",
              "          [ 2.3936,  2.4111,  2.3936,  ...,  2.0609,  2.0959,  2.1134]],\n",
              "\n",
              "         [[ 2.6226,  2.6226,  2.6226,  ...,  2.6400,  2.6400,  2.6400],\n",
              "          [ 2.6226,  2.6226,  2.6226,  ...,  2.6400,  2.6400,  2.6400],\n",
              "          [ 2.6226,  2.6226,  2.6226,  ...,  2.6400,  2.6400,  2.6400],\n",
              "          ...,\n",
              "          [ 2.6400,  2.6400,  2.6400,  ...,  2.5180,  2.5529,  2.5703],\n",
              "          [ 2.6400,  2.6400,  2.6400,  ...,  2.5354,  2.5703,  2.5703],\n",
              "          [ 2.6400,  2.6400,  2.6400,  ...,  2.5529,  2.5703,  2.5703]]],\n",
              "\n",
              "\n",
              "        [[[ 1.3242,  1.2899,  1.2899,  ...,  1.3584,  1.3584,  1.3755],\n",
              "          [ 1.2728,  1.2728,  1.2728,  ...,  1.3413,  1.3413,  1.3413],\n",
              "          [ 1.2728,  1.2728,  1.2557,  ...,  1.3242,  1.3413,  1.3413],\n",
              "          ...,\n",
              "          [-1.8610, -1.8610, -1.8610,  ..., -1.8610, -1.8610, -1.8610],\n",
              "          [-1.8610, -1.8610, -1.8610,  ..., -1.8610, -1.8610, -1.8610],\n",
              "          [-1.8610, -1.8610, -1.8610,  ..., -1.8610, -1.8610, -1.8610]],\n",
              "\n",
              "         [[ 1.3081,  1.2731,  1.2731,  ...,  1.5182,  1.5182,  1.5357],\n",
              "          [ 1.2556,  1.2556,  1.2556,  ...,  1.5007,  1.5007,  1.5007],\n",
              "          [ 1.2556,  1.2556,  1.2381,  ...,  1.4832,  1.5007,  1.5007],\n",
              "          ...,\n",
              "          [ 0.0476,  0.0476,  0.0476,  ...,  0.0476,  0.0476,  0.0476],\n",
              "          [ 0.0476,  0.0476,  0.0476,  ...,  0.0476,  0.0476,  0.0476],\n",
              "          [ 0.0476,  0.0476,  0.0476,  ...,  0.0476,  0.0476,  0.0476]],\n",
              "\n",
              "         [[ 1.3502,  1.3154,  1.3154,  ...,  1.5942,  1.5942,  1.6117],\n",
              "          [ 1.2980,  1.2980,  1.2980,  ...,  1.5768,  1.5768,  1.5768],\n",
              "          [ 1.2980,  1.2980,  1.2805,  ...,  1.5594,  1.5768,  1.5768],\n",
              "          ...,\n",
              "          [ 1.2282,  1.2282,  1.2282,  ...,  1.2282,  1.2282,  1.2282],\n",
              "          [ 1.2282,  1.2282,  1.2282,  ...,  1.2282,  1.2282,  1.2282],\n",
              "          [ 1.2282,  1.2282,  1.2282,  ...,  1.2282,  1.2282,  1.2282]]],\n",
              "\n",
              "\n",
              "        [[[ 1.9235,  1.9407,  1.9749,  ...,  1.6667,  1.7009,  1.7009],\n",
              "          [ 1.9920,  1.8893,  2.0092,  ...,  1.7352,  1.6667,  1.7523],\n",
              "          [ 1.9407,  1.8037,  1.9578,  ...,  1.7694,  1.6324,  1.6838],\n",
              "          ...,\n",
              "          [ 1.9749,  1.9749,  1.9749,  ...,  1.8893,  1.9064,  1.9064],\n",
              "          [ 1.9407,  1.9407,  1.9407,  ...,  1.8893,  1.9064,  1.9064],\n",
              "          [ 1.9578,  1.9578,  1.9578,  ...,  1.8893,  1.9064,  1.9064]],\n",
              "\n",
              "         [[ 1.9909,  2.0084,  2.0084,  ...,  1.7108,  1.6232,  1.7633],\n",
              "          [ 2.0609,  1.8859,  2.0434,  ...,  1.7458,  1.6057,  1.8333],\n",
              "          [ 1.9384,  1.7458,  2.1134,  ...,  1.6933,  1.6758,  1.7458],\n",
              "          ...,\n",
              "          [ 2.1134,  2.1134,  2.1134,  ...,  2.0259,  2.0434,  2.0434],\n",
              "          [ 2.0784,  2.0784,  2.0784,  ...,  2.0259,  2.0434,  2.0434],\n",
              "          [ 2.0959,  2.0959,  2.0959,  ...,  2.0259,  2.0434,  2.0434]],\n",
              "\n",
              "         [[ 2.1346,  2.1171,  2.2566,  ...,  1.8034,  1.6117,  1.8034],\n",
              "          [ 2.1868,  2.0300,  2.2914,  ...,  1.8731,  1.6640,  1.9603],\n",
              "          [ 2.1171,  1.9080,  2.3263,  ...,  1.7860,  1.8034,  1.8731],\n",
              "          ...,\n",
              "          [ 2.3437,  2.3437,  2.3437,  ...,  2.2566,  2.2740,  2.2740],\n",
              "          [ 2.3088,  2.3088,  2.3088,  ...,  2.2566,  2.2740,  2.2740],\n",
              "          [ 2.3263,  2.3263,  2.3263,  ...,  2.2566,  2.2740,  2.2740]]]])"
            ]
          },
          "metadata": {},
          "execution_count": 148
        }
      ],
      "source": [
        "input_images"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "NK6k4W4SVOxk"
      },
      "outputs": [],
      "source": [
        "x = model(input_images)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "rqY5Q1GbVOz3",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4ff03d41-9bec-4a18-f0ba-9a5cd4716ab8"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "tensor([[[ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         ...,\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758]],\n",
              "\n",
              "        [[ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         ...,\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758]],\n",
              "\n",
              "        [[ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         ...,\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758]],\n",
              "\n",
              "        ...,\n",
              "\n",
              "        [[ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         ...,\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758]],\n",
              "\n",
              "        [[ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         ...,\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758]],\n",
              "\n",
              "        [[ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         ...,\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758],\n",
              "         [ 0.9269, -0.8635,  2.1012,  ..., -3.6619, -0.0741,  0.1758]]],\n",
              "       grad_fn=<ViewBackward0>)"
            ]
          },
          "metadata": {},
          "execution_count": 152
        }
      ],
      "source": [
        "x"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "PEeRMYKTVO3R"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "XeIZVpBlVO5b"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "zyPPP32QVO8b"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "FXQpxl_Mew9o"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ChndVNfDMP32"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "LzZEQoTDMP-A"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "aSFwek4BNocs"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "zcZj8h9INofV"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "RtmKQ-qaNoiI"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Eh-C6jXHNoke"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "eld0KpPxNonX"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "jUJuWxn8NoqH"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ix0cAMSeNosK"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "aBy1GDEXNo0V"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "cyADl1lsNo4X"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "poVpp3EDNo6o"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "w-SLDrV0No9X"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "GnFVCpD5No_b"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "eisOfRzANpBu"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "iihgOmvbTFxw"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "MlsPmSu2TFz4"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "RbafvPnMTF12"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "-X_O9tuJTF5a"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "CFYhaij3TGHR"
      },
      "outputs": [],
      "source": [
        "len(input_tensor)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "aZ44vkZE0Pxo"
      },
      "outputs": [],
      "source": [
        "c, d = my_model(input_images, input_tensor, mask)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "WNUxwnnOKj3A"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "-CtKwrREKkil"
      },
      "outputs": [],
      "source": []
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Bht2sKyj0P1Z",
        "outputId": "72ff6466-254c-4b4d-a94c-7c237c8347f6"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Parameter 'patch_embedding.projection.weight' is unfrozen.\n",
            "Parameter 'patch_embedding.projection.bias' is unfrozen.\n",
            "Parameter 'position_embedding.weight' is unfrozen.\n",
            "Parameter 'layers.0.attention.values.weight' is unfrozen.\n",
            "Parameter 'layers.0.attention.values.bias' is unfrozen.\n",
            "Parameter 'layers.0.attention.keys.weight' is unfrozen.\n",
            "Parameter 'layers.0.attention.keys.bias' is unfrozen.\n",
            "Parameter 'layers.0.attention.queries.weight' is unfrozen.\n",
            "Parameter 'layers.0.attention.queries.bias' is unfrozen.\n",
            "Parameter 'layers.0.attention.fc_out.weight' is unfrozen.\n",
            "Parameter 'layers.0.attention.fc_out.bias' is unfrozen.\n",
            "Parameter 'layers.0.norm1.weight' is unfrozen.\n",
            "Parameter 'layers.0.norm1.bias' is unfrozen.\n",
            "Parameter 'layers.0.norm2.weight' is unfrozen.\n",
            "Parameter 'layers.0.norm2.bias' is unfrozen.\n",
            "Parameter 'layers.0.feed_forward.0.weight' is unfrozen.\n",
            "Parameter 'layers.0.feed_forward.0.bias' is unfrozen.\n",
            "Parameter 'layers.0.feed_forward.2.weight' is unfrozen.\n",
            "Parameter 'layers.0.feed_forward.2.bias' is unfrozen.\n",
            "Parameter 'layers.1.attention.values.weight' is unfrozen.\n",
            "Parameter 'layers.1.attention.values.bias' is unfrozen.\n",
            "Parameter 'layers.1.attention.keys.weight' is unfrozen.\n",
            "Parameter 'layers.1.attention.keys.bias' is unfrozen.\n",
            "Parameter 'layers.1.attention.queries.weight' is unfrozen.\n",
            "Parameter 'layers.1.attention.queries.bias' is unfrozen.\n",
            "Parameter 'layers.1.attention.fc_out.weight' is unfrozen.\n",
            "Parameter 'layers.1.attention.fc_out.bias' is unfrozen.\n",
            "Parameter 'layers.1.norm1.weight' is unfrozen.\n",
            "Parameter 'layers.1.norm1.bias' is unfrozen.\n",
            "Parameter 'layers.1.norm2.weight' is unfrozen.\n",
            "Parameter 'layers.1.norm2.bias' is unfrozen.\n",
            "Parameter 'layers.1.feed_forward.0.weight' is unfrozen.\n",
            "Parameter 'layers.1.feed_forward.0.bias' is unfrozen.\n",
            "Parameter 'layers.1.feed_forward.2.weight' is unfrozen.\n",
            "Parameter 'layers.1.feed_forward.2.bias' is unfrozen.\n",
            "Parameter 'layers.2.attention.values.weight' is unfrozen.\n",
            "Parameter 'layers.2.attention.values.bias' is unfrozen.\n",
            "Parameter 'layers.2.attention.keys.weight' is unfrozen.\n",
            "Parameter 'layers.2.attention.keys.bias' is unfrozen.\n",
            "Parameter 'layers.2.attention.queries.weight' is unfrozen.\n",
            "Parameter 'layers.2.attention.queries.bias' is unfrozen.\n",
            "Parameter 'layers.2.attention.fc_out.weight' is unfrozen.\n",
            "Parameter 'layers.2.attention.fc_out.bias' is unfrozen.\n",
            "Parameter 'layers.2.norm1.weight' is unfrozen.\n",
            "Parameter 'layers.2.norm1.bias' is unfrozen.\n",
            "Parameter 'layers.2.norm2.weight' is unfrozen.\n",
            "Parameter 'layers.2.norm2.bias' is unfrozen.\n",
            "Parameter 'layers.2.feed_forward.0.weight' is unfrozen.\n",
            "Parameter 'layers.2.feed_forward.0.bias' is unfrozen.\n",
            "Parameter 'layers.2.feed_forward.2.weight' is unfrozen.\n",
            "Parameter 'layers.2.feed_forward.2.bias' is unfrozen.\n",
            "Parameter 'layers.3.attention.values.weight' is unfrozen.\n",
            "Parameter 'layers.3.attention.values.bias' is unfrozen.\n",
            "Parameter 'layers.3.attention.keys.weight' is unfrozen.\n",
            "Parameter 'layers.3.attention.keys.bias' is unfrozen.\n",
            "Parameter 'layers.3.attention.queries.weight' is unfrozen.\n",
            "Parameter 'layers.3.attention.queries.bias' is unfrozen.\n",
            "Parameter 'layers.3.attention.fc_out.weight' is unfrozen.\n",
            "Parameter 'layers.3.attention.fc_out.bias' is unfrozen.\n",
            "Parameter 'layers.3.norm1.weight' is unfrozen.\n",
            "Parameter 'layers.3.norm1.bias' is unfrozen.\n",
            "Parameter 'layers.3.norm2.weight' is unfrozen.\n",
            "Parameter 'layers.3.norm2.bias' is unfrozen.\n",
            "Parameter 'layers.3.feed_forward.0.weight' is unfrozen.\n",
            "Parameter 'layers.3.feed_forward.0.bias' is unfrozen.\n",
            "Parameter 'layers.3.feed_forward.2.weight' is unfrozen.\n",
            "Parameter 'layers.3.feed_forward.2.bias' is unfrozen.\n",
            "Parameter 'layers.4.attention.values.weight' is unfrozen.\n",
            "Parameter 'layers.4.attention.values.bias' is unfrozen.\n",
            "Parameter 'layers.4.attention.keys.weight' is unfrozen.\n",
            "Parameter 'layers.4.attention.keys.bias' is unfrozen.\n",
            "Parameter 'layers.4.attention.queries.weight' is unfrozen.\n",
            "Parameter 'layers.4.attention.queries.bias' is unfrozen.\n",
            "Parameter 'layers.4.attention.fc_out.weight' is unfrozen.\n",
            "Parameter 'layers.4.attention.fc_out.bias' is unfrozen.\n",
            "Parameter 'layers.4.norm1.weight' is unfrozen.\n",
            "Parameter 'layers.4.norm1.bias' is unfrozen.\n",
            "Parameter 'layers.4.norm2.weight' is unfrozen.\n",
            "Parameter 'layers.4.norm2.bias' is unfrozen.\n",
            "Parameter 'layers.4.feed_forward.0.weight' is unfrozen.\n",
            "Parameter 'layers.4.feed_forward.0.bias' is unfrozen.\n",
            "Parameter 'layers.4.feed_forward.2.weight' is unfrozen.\n",
            "Parameter 'layers.4.feed_forward.2.bias' is unfrozen.\n",
            "Parameter 'layers.5.attention.values.weight' is unfrozen.\n",
            "Parameter 'layers.5.attention.values.bias' is unfrozen.\n",
            "Parameter 'layers.5.attention.keys.weight' is unfrozen.\n",
            "Parameter 'layers.5.attention.keys.bias' is unfrozen.\n",
            "Parameter 'layers.5.attention.queries.weight' is unfrozen.\n",
            "Parameter 'layers.5.attention.queries.bias' is unfrozen.\n",
            "Parameter 'layers.5.attention.fc_out.weight' is unfrozen.\n",
            "Parameter 'layers.5.attention.fc_out.bias' is unfrozen.\n",
            "Parameter 'layers.5.norm1.weight' is unfrozen.\n",
            "Parameter 'layers.5.norm1.bias' is unfrozen.\n",
            "Parameter 'layers.5.norm2.weight' is unfrozen.\n",
            "Parameter 'layers.5.norm2.bias' is unfrozen.\n",
            "Parameter 'layers.5.feed_forward.0.weight' is unfrozen.\n",
            "Parameter 'layers.5.feed_forward.0.bias' is unfrozen.\n",
            "Parameter 'layers.5.feed_forward.2.weight' is unfrozen.\n",
            "Parameter 'layers.5.feed_forward.2.bias' is unfrozen.\n",
            "Parameter 'classification_head.weight' is unfrozen.\n",
            "Parameter 'classification_head.bias' is unfrozen.\n"
          ]
        }
      ],
      "source": [
        "import torch\n",
        "import torchvision.models as models\n",
        "\n",
        "#model =my_model\n",
        "def are_params_unfrozen(model):\n",
        "    for name, param in model.named_parameters():\n",
        "        if param.requires_grad:\n",
        "            print(f\"Parameter '{name}' is unfrozen.\")\n",
        "        else:\n",
        "            print(f\"Parameter '{name}' is frozen.\")\n",
        "\n",
        "are_params_unfrozen(model)"
      ]
    },
    {
      "cell_type": "code",
      "source": [],
      "metadata": {
        "id": "bDNQhyC5L0Sp"
      },
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}