rakuten_method(1).ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3",
      "language": "python"
    },
    "language_info": {
      "name": "python",
      "version": "3.7.6",
      "mimetype": "text/x-python",
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "pygments_lexer": "ipython3",
      "nbconvert_exporter": "python",
      "file_extension": ".py"
    },
    "colab": {
      "name": "rakuten-method.ipynb",
      "provenance": []
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "SAd3995Z7lKI",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "168400b0-8f76-4dca-deaa-6568032e6aab"
      },
      "source": [
        "!pip install transformers"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: transformers in /usr/local/lib/python3.6/dist-packages (4.0.1)\n",
            "Requirement already satisfied: tokenizers==0.9.4 in /usr/local/lib/python3.6/dist-packages (from transformers) (0.9.4)\n",
            "Requirement already satisfied: packaging in /usr/local/lib/python3.6/dist-packages (from transformers) (20.7)\n",
            "Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.6/dist-packages (from transformers) (2019.12.20)\n",
            "Requirement already satisfied: sacremoses in /usr/local/lib/python3.6/dist-packages (from transformers) (0.0.43)\n",
            "Requirement already satisfied: dataclasses; python_version < \"3.7\" in /usr/local/lib/python3.6/dist-packages (from transformers) (0.8)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from transformers) (1.18.5)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.6/dist-packages (from transformers) (3.0.12)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.6/dist-packages (from transformers) (2.23.0)\n",
            "Requirement already satisfied: tqdm>=4.27 in /usr/local/lib/python3.6/dist-packages (from transformers) (4.41.1)\n",
            "Requirement already satisfied: pyparsing>=2.0.2 in /usr/local/lib/python3.6/dist-packages (from packaging->transformers) (2.4.7)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from sacremoses->transformers) (1.15.0)\n",
            "Requirement already satisfied: click in /usr/local/lib/python3.6/dist-packages (from sacremoses->transformers) (7.1.2)\n",
            "Requirement already satisfied: joblib in /usr/local/lib/python3.6/dist-packages (from sacremoses->transformers) (0.17.0)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests->transformers) (1.24.3)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests->transformers) (2.10)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests->transformers) (3.0.4)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests->transformers) (2020.12.5)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "D2wXRv_8c9Ss",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9b046256-35f0-4611-b847-6925c5661b03"
      },
      "source": [
        "!pip install torch torchvision"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: torch in /usr/local/lib/python3.6/dist-packages (1.7.0+cu101)\n",
            "Requirement already satisfied: torchvision in /usr/local/lib/python3.6/dist-packages (0.8.1+cu101)\n",
            "Requirement already satisfied: dataclasses in /usr/local/lib/python3.6/dist-packages (from torch) (0.8)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from torch) (1.18.5)\n",
            "Requirement already satisfied: future in /usr/local/lib/python3.6/dist-packages (from torch) (0.16.0)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.6/dist-packages (from torch) (3.7.4.3)\n",
            "Requirement already satisfied: pillow>=4.1.1 in /usr/local/lib/python3.6/dist-packages (from torchvision) (7.0.0)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TWsgA5I88dZ_",
        "outputId": "34526f28-2fe3-43d3-a42b-2ee6a155710f"
      },
      "source": [
        "from google.colab import drive,files\n",
        "drive.mount(\"/content/drive\")"
      ],
      "execution_count": 71,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2-oIZrm381fi",
        "outputId": "7eb3ba23-cd1e-44f4-9377-0df146308704"
      },
      "source": [
        "!unzip /content/drive/MyDrive/DS2021/Datasets/archive\\(8\\).zip "
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Archive:  /content/drive/MyDrive/DS2021/Datasets/archive(8).zip\n",
            "replace X_test_update.csv? [y]es, [n]o, [A]ll, [N]one, [r]ename: N\n",
            "N\n",
            "N\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CpT8zTkfu9VE",
        "trusted": true
      },
      "source": [
        "import pandas as pd\n",
        "import numpy as np \n",
        "from nltk import WordPunctTokenizer\n",
        "from nltk import WordNetLemmatizer\n",
        "from nltk.corpus import stopwords\n"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "84yHAW-a7bSb"
      },
      "source": [
        "Image Path\n",
        " training :\" ../input/rakuten-france-multimodal-product-classification/images/images/image_test\"\n",
        " \n",
        "test: \"../input/rakuten-france-multimodal-product-classification/images/images/image_train\"\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RCvy4eMgsMcr",
        "trusted": true
      },
      "source": [
        "x_train = pd.read_csv(\"/content/X_train_update.csv\")\n",
        "y_train = pd.read_csv(\"/content/Y_train_CVw08PX.csv\")\n",
        "x_test = pd.read_csv(\"/content/X_test_update.csv\")"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "I9gocfpO7bSg"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bojQILBh7bSj",
        "outputId": "8620a5c2-3cbd-4ba6-e540-de9cd52fb9b1"
      },
      "source": [
        "x_train.columns"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Index(['Unnamed: 0', 'designation', 'description', 'productid', 'imageid'], dtype='object')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 310
        },
        "id": "A-q1C-hz7bSm",
        "outputId": "0bb7328e-2647-462f-9aee-94db94af83e1"
      },
      "source": [
        "x_train.head()"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Unnamed: 0</th>\n",
              "      <th>designation</th>\n",
              "      <th>description</th>\n",
              "      <th>productid</th>\n",
              "      <th>imageid</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0</td>\n",
              "      <td>Olivia: Personalisiertes Notizbuch / 150 Seite...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>3804725264</td>\n",
              "      <td>1263597046</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>1</td>\n",
              "      <td>Journal Des Arts (Le) N° 133 Du 28/09/2001 - L...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>436067568</td>\n",
              "      <td>1008141237</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>2</td>\n",
              "      <td>Grand Stylet Ergonomique Bleu Gamepad Nintendo...</td>\n",
              "      <td>PILOT STYLE Touch Pen de marque Speedlink est ...</td>\n",
              "      <td>201115110</td>\n",
              "      <td>938777978</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>3</td>\n",
              "      <td>Peluche Donald - Europe - Disneyland 2000 (Mar...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>50418756</td>\n",
              "      <td>457047496</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>4</td>\n",
              "      <td>La Guerre Des Tuques</td>\n",
              "      <td>Luc a des id&amp;eacute;es de grandeur. Il veut or...</td>\n",
              "      <td>278535884</td>\n",
              "      <td>1077757786</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   Unnamed: 0  ...     imageid\n",
              "0           0  ...  1263597046\n",
              "1           1  ...  1008141237\n",
              "2           2  ...   938777978\n",
              "3           3  ...   457047496\n",
              "4           4  ...  1077757786\n",
              "\n",
              "[5 rows x 5 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "id": "qYeSkfN17bSn"
      },
      "source": [
        "x_train.rename(columns={\"Unnamed: 0\": \"Id\"}, inplace=True)\n",
        "y_train.rename(columns={\"Unnamed: 0\": \"Id\"}, inplace=True)\n",
        "x_test.rename(columns={\"Unnamed: 0\": \"Id\"}, inplace=True)"
      ],
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "id": "vqjzgnsN7bSo",
        "outputId": "220e8a23-d6a1-4db3-8ed7-de7a929d80a7"
      },
      "source": [
        "x_train.describe()"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Id</th>\n",
              "      <th>productid</th>\n",
              "      <th>imageid</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>84916.000000</td>\n",
              "      <td>8.491600e+04</td>\n",
              "      <td>8.491600e+04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>42457.500000</td>\n",
              "      <td>2.555468e+09</td>\n",
              "      <td>1.152691e+09</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>24513.282066</td>\n",
              "      <td>1.588656e+09</td>\n",
              "      <td>1.751427e+08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>0.000000</td>\n",
              "      <td>1.839120e+05</td>\n",
              "      <td>6.728400e+04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>21228.750000</td>\n",
              "      <td>6.760519e+08</td>\n",
              "      <td>1.056269e+09</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>42457.500000</td>\n",
              "      <td>3.190506e+09</td>\n",
              "      <td>1.213354e+09</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>63686.250000</td>\n",
              "      <td>3.995599e+09</td>\n",
              "      <td>1.275646e+09</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>84915.000000</td>\n",
              "      <td>4.252012e+09</td>\n",
              "      <td>1.328824e+09</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                 Id     productid       imageid\n",
              "count  84916.000000  8.491600e+04  8.491600e+04\n",
              "mean   42457.500000  2.555468e+09  1.152691e+09\n",
              "std    24513.282066  1.588656e+09  1.751427e+08\n",
              "min        0.000000  1.839120e+05  6.728400e+04\n",
              "25%    21228.750000  6.760519e+08  1.056269e+09\n",
              "50%    42457.500000  3.190506e+09  1.213354e+09\n",
              "75%    63686.250000  3.995599e+09  1.275646e+09\n",
              "max    84915.000000  4.252012e+09  1.328824e+09"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 10
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "id": "pBCbkxnF7bSq",
        "outputId": "74b7fa58-b618-4f13-b0ef-b65e966c3ad1"
      },
      "source": [
        "y_train.describe()"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Id</th>\n",
              "      <th>prdtypecode</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>count</th>\n",
              "      <td>84916.000000</td>\n",
              "      <td>84916.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>mean</th>\n",
              "      <td>42457.500000</td>\n",
              "      <td>1773.219900</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>std</th>\n",
              "      <td>24513.282066</td>\n",
              "      <td>788.179885</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>min</th>\n",
              "      <td>0.000000</td>\n",
              "      <td>10.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25%</th>\n",
              "      <td>21228.750000</td>\n",
              "      <td>1281.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50%</th>\n",
              "      <td>42457.500000</td>\n",
              "      <td>1920.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>75%</th>\n",
              "      <td>63686.250000</td>\n",
              "      <td>2522.000000</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>max</th>\n",
              "      <td>84915.000000</td>\n",
              "      <td>2905.000000</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                 Id   prdtypecode\n",
              "count  84916.000000  84916.000000\n",
              "mean   42457.500000   1773.219900\n",
              "std    24513.282066    788.179885\n",
              "min        0.000000     10.000000\n",
              "25%    21228.750000   1281.000000\n",
              "50%    42457.500000   1920.000000\n",
              "75%    63686.250000   2522.000000\n",
              "max    84915.000000   2905.000000"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hA1ve8nq44Ag",
        "trusted": true
      },
      "source": [
        "h=y_train.prdtypecode.value_counts()"
      ],
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zGk7ylny5H8q",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "4ab82730-aadc-436c-ef1d-b4004f7a70fa"
      },
      "source": [
        "len(h)"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "27"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-Ls7qKsY5Sbs",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 351
        },
        "outputId": "56d97517-5ac5-4d57-cbbd-e610fef73ca7"
      },
      "source": [
        "import seaborn as sns\n",
        "sns.countplot(y_train.prdtypecode)"
      ],
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/seaborn/_decorators.py:43: FutureWarning: Pass the following variable as a keyword arg: x. From version 0.12, the only valid positional argument will be `data`, and passing other arguments without an explicit keyword will result in an error or misinterpretation.\n",
            "  FutureWarning\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7fa94ec536d8>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LLzNu6QjqbQR",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ab631e91-88df-42bc-9c77-7ebec07c40c7"
      },
      "source": [
        "len(x_train)"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "84916"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6KujjabiqQ5m",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 293
        },
        "outputId": "d649d082-94cd-47ed-d0c6-b8fbe071f2ce"
      },
      "source": [
        "x_train.sample(5)"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Id</th>\n",
              "      <th>designation</th>\n",
              "      <th>description</th>\n",
              "      <th>productid</th>\n",
              "      <th>imageid</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>39694</th>\n",
              "      <td>39694</td>\n",
              "      <td>The Suffering - Ps2 - Playstation 2 - Edition ...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>1651653704</td>\n",
              "      <td>1241447361</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>42634</th>\n",
              "      <td>42634</td>\n",
              "      <td>Nintendo Gamecube - Console De Jeux - Noir - M...</td>\n",
              "      <td>NaN</td>\n",
              "      <td>1805032441</td>\n",
              "      <td>1102674573</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>10445</th>\n",
              "      <td>10445</td>\n",
              "      <td>Ultra Soft Area Moderne Tapis Shaggy Tapis Nur...</td>\n",
              "      <td>Ultra Zone moderne doux tapis shaggy Nursery T...</td>\n",
              "      <td>4209068684</td>\n",
              "      <td>1316532644</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>46035</th>\n",
              "      <td>46035</td>\n",
              "      <td>Draeger Lot De 5 Stickers Mural L'ours Bleu Mu...</td>\n",
              "      <td>Sticker mural pour enfant composé d&amp;#39;un our...</td>\n",
              "      <td>3876188970</td>\n",
              "      <td>1256969724</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>59558</th>\n",
              "      <td>59558</td>\n",
              "      <td>Surpasshobby Étanche 3650 4300kv Brushless 45a...</td>\n",
              "      <td>SurpassHobby étanche 3650 4300KV Brushless Sen...</td>\n",
              "      <td>4019800024</td>\n",
              "      <td>1277511313</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "          Id  ...     imageid\n",
              "39694  39694  ...  1241447361\n",
              "42634  42634  ...  1102674573\n",
              "10445  10445  ...  1316532644\n",
              "46035  46035  ...  1256969724\n",
              "59558  59558  ...  1277511313\n",
              "\n",
              "[5 rows x 5 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9SfPQzBZqnkU",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 206
        },
        "outputId": "55db6266-d43f-42c3-972d-f023f383a290"
      },
      "source": [
        "y_train.sample(5)"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Id</th>\n",
              "      <th>prdtypecode</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>79356</th>\n",
              "      <td>79356</td>\n",
              "      <td>50</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>47227</th>\n",
              "      <td>47227</td>\n",
              "      <td>2905</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>50147</th>\n",
              "      <td>50147</td>\n",
              "      <td>2582</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>78163</th>\n",
              "      <td>78163</td>\n",
              "      <td>1560</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>23086</th>\n",
              "      <td>23086</td>\n",
              "      <td>2280</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "          Id  prdtypecode\n",
              "79356  79356           50\n",
              "47227  47227         2905\n",
              "50147  50147         2582\n",
              "78163  78163         1560\n",
              "23086  23086         2280"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 17
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NVADGu2Jqvyh",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "fec65f78-d8c3-478e-d9a3-dcb4eaf17c2b"
      },
      "source": [
        "print(x_test.sample(5))"
      ],
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "          Id  ...     imageid\n",
            "3173   88089  ...  1100077520\n",
            "13437  98353  ...  1156656894\n",
            "676    85592  ...  1306803682\n",
            "3856   88772  ...  1324007933\n",
            "1358   86274  ...   848757863\n",
            "\n",
            "[5 rows x 5 columns]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BRPH9Yeo5Ocb"
      },
      "source": [
        "classifcation multinomiale non balance"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "quj_HSiOPl03",
        "outputId": "f23ebeb1-899b-4c47-e512-8fccbaee69e6"
      },
      "source": [
        "!pip install sentencepiece"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: sentencepiece in /usr/local/lib/python3.6/dist-packages (0.1.94)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "id": "kyIr4A4c7bS3"
      },
      "source": [
        "import torch\n",
        "from transformers import CamembertTokenizer\n",
        "\n",
        "model_name = \"camembert-base\"\n",
        "tokenizer = CamembertTokenizer.from_pretrained(model_name)"
      ],
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 476
        },
        "id": "GMjlpgyg7bS4",
        "outputId": "26f40012-d8ea-4510-fc0a-5c0391678d1e"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "reviews_len = [len(tokenizer.encode(review, max_length=512))\n",
        "                          for review in x_train.designation.values]\n",
        "print(\"Average length: {:.1f}\".format(np.mean(reviews_len)))\n",
        "print(\"Max length: {}\".format(max(reviews_len)))\n",
        "\n",
        "plt.figure(figsize=(10,5))\n",
        "ax = sns.distplot(reviews_len, bins=150, kde=False, hist_kws=dict(alpha=0.8))\n",
        "ax.set(xlabel='Number of tokens')\n",
        "\n",
        "# Finalize the plot\n",
        "sns.despine(bottom=True)\n",
        "plt.tight_layout(h_pad=2)\n",
        "\n",
        "# Saving plot\n",
        "fig = ax.get_figure()\n",
        "fig.savefig('./number_of_tokens.png', dpi=200)"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Truncation was not explicitly activated but `max_length` is provided a specific value, please use `truncation=True` to explicitly truncate examples to max length. Defaulting to 'longest_first' truncation strategy. If you encode pairs of sequences (GLUE-style) with the tokenizer you can select this strategy more precisely by providing a specific strategy to `truncation`.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Average length: 22.4\n",
            "Max length: 104\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/seaborn/distributions.py:2551: FutureWarning: `distplot` is a deprecated function and will be removed in a future version. Please adapt your code to use either `displot` (a figure-level function with similar flexibility) or `histplot` (an axes-level function for histograms).\n",
            "  warnings.warn(msg, FutureWarning)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "i70GbmlD8bKA",
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "af9704a4-4000-4cb7-a87a-e18742555552"
      },
      "source": [
        "import warnings\n",
        "warnings.filterwarnings('ignore')\n",
        "from sklearn.pipeline import FeatureUnion, Pipeline, make_pipeline\n",
        "from sklearn.base import BaseEstimator, TransformerMixin\n",
        "import spacy\n",
        "import nltk\n",
        "nltk.download('stopwords')\n",
        "!python -m spacy download en_core_web_md\n",
        "import en_core_web_md\n",
        "nlp = en_core_web_md.load()"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[nltk_data] Downloading package stopwords to /root/nltk_data...\n",
            "[nltk_data]   Package stopwords is already up-to-date!\n",
            "Requirement already satisfied: en_core_web_md==2.2.5 from https://github.com/explosion/spacy-models/releases/download/en_core_web_md-2.2.5/en_core_web_md-2.2.5.tar.gz#egg=en_core_web_md==2.2.5 in /usr/local/lib/python3.6/dist-packages (2.2.5)\n",
            "Requirement already satisfied: spacy>=2.2.2 in /usr/local/lib/python3.6/dist-packages (from en_core_web_md==2.2.5) (2.2.4)\n",
            "Requirement already satisfied: srsly<1.1.0,>=1.0.2 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (1.0.5)\n",
            "Requirement already satisfied: preshed<3.1.0,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (3.0.5)\n",
            "Requirement already satisfied: requests<3.0.0,>=2.13.0 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (2.23.0)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (50.3.2)\n",
            "Requirement already satisfied: catalogue<1.1.0,>=0.0.7 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (1.0.0)\n",
            "Requirement already satisfied: murmurhash<1.1.0,>=0.28.0 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (1.0.5)\n",
            "Requirement already satisfied: wasabi<1.1.0,>=0.4.0 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (0.8.0)\n",
            "Requirement already satisfied: plac<1.2.0,>=0.9.6 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (1.1.3)\n",
            "Requirement already satisfied: thinc==7.4.0 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (7.4.0)\n",
            "Requirement already satisfied: tqdm<5.0.0,>=4.38.0 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (4.41.1)\n",
            "Requirement already satisfied: blis<0.5.0,>=0.4.0 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (0.4.1)\n",
            "Requirement already satisfied: numpy>=1.15.0 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (1.18.5)\n",
            "Requirement already satisfied: cymem<2.1.0,>=2.0.2 in /usr/local/lib/python3.6/dist-packages (from spacy>=2.2.2->en_core_web_md==2.2.5) (2.0.5)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests<3.0.0,>=2.13.0->spacy>=2.2.2->en_core_web_md==2.2.5) (1.24.3)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests<3.0.0,>=2.13.0->spacy>=2.2.2->en_core_web_md==2.2.5) (3.0.4)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests<3.0.0,>=2.13.0->spacy>=2.2.2->en_core_web_md==2.2.5) (2.10)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests<3.0.0,>=2.13.0->spacy>=2.2.2->en_core_web_md==2.2.5) (2020.12.5)\n",
            "Requirement already satisfied: importlib-metadata>=0.20; python_version < \"3.8\" in /usr/local/lib/python3.6/dist-packages (from catalogue<1.1.0,>=0.0.7->spacy>=2.2.2->en_core_web_md==2.2.5) (3.1.1)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.6/dist-packages (from importlib-metadata>=0.20; python_version < \"3.8\"->catalogue<1.1.0,>=0.0.7->spacy>=2.2.2->en_core_web_md==2.2.5) (3.4.0)\n",
            "\u001b[38;5;2m✔ Download and installation successful\u001b[0m\n",
            "You can now load the model via spacy.load('en_core_web_md')\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "id": "-tgjhXRi7bS6"
      },
      "source": [
        "class simple_Text_cleaner(BaseEstimator, TransformerMixin):\n",
        "  # add another additional parameter, just for fun, while we are at it\n",
        "    def __init__(self, stopwords,columns,tokenizer=WordPunctTokenizer()): \n",
        "        self.columns = columns\n",
        "        self.stopwords = stopwords\n",
        "        self.tokenizer=tokenizer\n",
        "    def rm_stopwords(self,tokens):\n",
        "        tokens=self.tokenizer.tokenize(tokens)\n",
        "        return [ tk for tk in tokens if  tk not in self.stopwords ]\n",
        "    def text_clean_up(self,s=\"\"):\n",
        "        import re\n",
        "        user_pattern       = '@[^\\s]+'\n",
        "        s=re.sub(user_pattern, \"\", s)\n",
        "        remove = '\"#$%&()*+/:;<=>@[\\\\]^_`{|}~”“'\n",
        "        pattern = r\"[{}]\".format(remove)\n",
        "        s=re.sub(pattern,' ', s) \n",
        "        sequencePattern   = r\"(.)\\1\\1+\"\n",
        "        seqReplacePattern = r\"\\1\\1\"\n",
        "        s=re.sub(sequencePattern, seqReplacePattern, s)\n",
        "        s=re.sub(\"<[^>]*>\",' ', s)\n",
        "        s=re.sub(\"[\\r\\n]+\",' ', s)\n",
        "        s=re.sub(\"http\\S+\",' ', s)\n",
        "        s=re.sub(\"\\$[^>]*\\$\",' ', s)\n",
        "        s=re.sub(\"\\d+\",' ', s)\n",
        "        s=re.sub(\"\\s\\s+\",' ', s)\n",
        "        s.strip()\n",
        "        return s\n",
        "    def fit(self,X,y=None):\n",
        "        return self\n",
        "    def transform(self,X,y=None):\n",
        "        X_=X.copy()\n",
        "        X_[\"image_path\"]=(X_[\"imageid\"].apply(lambda x: str(int(x)))+\"_product_\"+X_[\"productid\"].apply(lambda x: str(int(x)))+\".jpg\").apply(lambda x: \"image_\"+x)\n",
        "        for i in self.columns:84916\n",
        "            X_[i]=X_[i].apply(lambda x: \" \".join(self.rm_stopwords(self.text_clean_up(x))))\n",
        "        return X_"
      ],
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "id": "uDVS0_ZQ7bS7"
      },
      "source": [
        "stop_words = stopwords.words('french')\n",
        "stop_words.extend(stopwords.words('english'))\n",
        "cleaner=simple_Text_cleaner(stopwords=stop_words,columns=[\"designation\"])\n",
        "new_data=cleaner.transform(x_train)"
      ],
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xVxnJwvc7bS8"
      },
      "source": [
        "image_imageid_product_productid.jpg."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 404
        },
        "id": "VmgyaEJj7bS9",
        "outputId": "bb6f1376-303c-45e2-c161-47b365409972"
      },
      "source": [
        "new_reviews_len = [len(tokenizer.encode(review, max_length=512))\n",
        "                          for review in new_data.designation.values]\n",
        "print(\"Average length: {:.1f}\".format(np.mean(new_reviews_len)))\n",
        "print(\"Max length: {}\".format(max(new_reviews_len)))\n",
        "\n",
        "plt.figure(figsize=(10,5))\n",
        "ax = sns.distplot(new_reviews_len, bins=150, kde=False, hist_kws=dict(alpha=0.8))\n",
        "ax.set(xlabel='Number of tokens')\n",
        "\n",
        "# Finalize the plot\n",
        "sns.despine(bottom=True)\n",
        "plt.tight_layout(h_pad=2)\n",
        "\n",
        "# Saving plot\n",
        "fig = ax.get_figure()\n",
        "fig.savefig('./number_of_tokens_clean.png', dpi=200)"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Average length: 19.7\n",
            "Max length: 114\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ibrH3xUd7bS-"
      },
      "source": [
        "On va travailler sur un max_length de 60 "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "A-ITaASYSoan"
      },
      "source": [
        "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
        "cpu=torch.device(\"cpu\")"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "NqVZ3CY3UOrD",
        "outputId": "db119385-6b92-43eb-f8ce-6fdc8addd28e"
      },
      "source": [
        "!pip install omegaconf\n",
        "!pip install hydra-core"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: omegaconf in /usr/local/lib/python3.6/dist-packages (2.0.5)\n",
            "Requirement already satisfied: PyYAML>=5.1.* in /usr/local/lib/python3.6/dist-packages (from omegaconf) (5.3.1)\n",
            "Requirement already satisfied: dataclasses; python_version == \"3.6\" in /usr/local/lib/python3.6/dist-packages (from omegaconf) (0.8)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.6/dist-packages (from omegaconf) (3.7.4.3)\n",
            "Collecting hydra-core\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/f0/1f/7f502b9e37596164111655861370b08626f46f9e4524433c354f472765d4/hydra_core-1.0.4-py3-none-any.whl (122kB)\n",
            "\u001b[K     |████████████████████████████████| 122kB 21.1MB/s \n",
            "\u001b[?25hRequirement already satisfied: omegaconf>=2.0.5 in /usr/local/lib/python3.6/dist-packages (from hydra-core) (2.0.5)\n",
            "Requirement already satisfied: importlib-resources; python_version < \"3.9\" in /usr/local/lib/python3.6/dist-packages (from hydra-core) (3.3.0)\n",
            "Collecting antlr4-python3-runtime==4.8\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/56/02/789a0bddf9c9b31b14c3e79ec22b9656185a803dc31c15f006f9855ece0d/antlr4-python3-runtime-4.8.tar.gz (112kB)\n",
            "\u001b[K     |████████████████████████████████| 112kB 40.6MB/s \n",
            "\u001b[?25hRequirement already satisfied: PyYAML>=5.1.* in /usr/local/lib/python3.6/dist-packages (from omegaconf>=2.0.5->hydra-core) (5.3.1)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.6/dist-packages (from omegaconf>=2.0.5->hydra-core) (3.7.4.3)\n",
            "Requirement already satisfied: dataclasses; python_version == \"3.6\" in /usr/local/lib/python3.6/dist-packages (from omegaconf>=2.0.5->hydra-core) (0.8)\n",
            "Requirement already satisfied: zipp>=0.4; python_version < \"3.8\" in /usr/local/lib/python3.6/dist-packages (from importlib-resources; python_version < \"3.9\"->hydra-core) (3.4.0)\n",
            "Building wheels for collected packages: antlr4-python3-runtime\n",
            "  Building wheel for antlr4-python3-runtime (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for antlr4-python3-runtime: filename=antlr4_python3_runtime-4.8-cp36-none-any.whl size=141231 sha256=89b02d603a289e22d47874bea4902c7b56f04a4ac4276fed754dd0143ab4bb74\n",
            "  Stored in directory: /root/.cache/pip/wheels/e3/e2/fa/b78480b448b8579ddf393bebd3f47ee23aa84c89b6a78285c8\n",
            "Successfully built antlr4-python3-runtime\n",
            "Installing collected packages: antlr4-python3-runtime, hydra-core\n",
            "Successfully installed antlr4-python3-runtime-4.8 hydra-core-1.0.4\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9VXzv_ia7bS_",
        "outputId": "538aaccd-b768-4a85-e07c-1dfd01731c41"
      },
      "source": [
        "from transformers import CamembertModel, CamembertTokenizer\n",
        "\n",
        "# You can replace \"camembert-base\" with any other model from the table, e.g. \"camembert/camembert-large\".\n",
        "tokenizer = CamembertTokenizer.from_pretrained(\"camembert-base\")\n",
        "camembert = CamembertModel.from_pretrained(\"camembert-base\")\n",
        "\n",
        "camembert.eval()  # disable dropout (or leave in train mode to finetune)\n",
        "\n",
        "import torch\n",
        "# Tokenize in sub-words with SentencePiece\n",
        "tokenized_sentence = tokenizer.tokenize(\"J'aime le camembert !\")\n",
        "# ['▁J', \"'\", 'aime', '▁le', '▁ca', 'member', 't', '▁!'] \n",
        "print(tokenized_sentence)\n",
        "# 1-hot encode and add special starting and end tokens \n",
        "encoded_sentence = tokenizer.encode(tokenized_sentence)\n",
        "# [5, 121, 11, 660, 16, 730, 25543, 110, 83, 6] \n",
        "# NB: Can be done in one step : tokenize.encode(\"J'aime le camembert !\")\n",
        "print(encoded_sentence)\n",
        "# Feed tokens to Camembert as a torch tensor (batch dim 1)BERT\n",
        "encoded_sentence = torch.tensor(encoded_sentence).unsqueeze(0)\n",
        "embeddings= camembert(encoded_sentence)[\"last_hidden_state\"]\n",
        "#print(embeddings.size())\n",
        "# embeddings.detach()\n",
        "# embeddings.size torch.Size([1, 10, 768])\n",
        "# tensor([[[-0.0254,  0.0235,  0.1027,  ..., -0.1459, -0.0205, -0.0116],\n",
        "#         [ 0.0606, -0.1811, -0.0418,  ..., -0.1815,  0.0880, -0.0766],\n",
        "#         [-0.1561, -0.1127,  0.2687,  ..., -0.0648,  0.0249,  0.0446],\n",
        "#         ...,\n",
        "\n",
        "\n",
        "\n",
        "from transformers import CamembertConfig\n",
        "# (Need to reload the model with new config)\n",
        "config = CamembertConfig.from_pretrained(\"camembert-base\", output_hidden_states=True)\n",
        "camembert = CamembertModel.from_pretrained(\"camembert-base\", config=config)\n",
        "camembert.eval()  # disable dropout (or leave in train mode to finetune)\n",
        "\n",
        "all_layer_embeddings = camembert(encoded_sentence)[2]\n",
        "#  all_layer_embeddings list of len(all_layer_embeddings) == 13 (input embedding layer + 12 self attention layers)\n",
        "all_layer_embeddings[5]\n",
        "print(len(all_layer_embeddings))\n",
        "# layer 5 contextual embedding : size torch.Size([1, 10, 768])\n",
        "#tensor([[[-0.0032,  0.0075,  0.0040,  ..., -0.0025, -0.0178, -0.0210],\n",
        "#         [-0.0996, -0.1474,  0.1057,  ..., -0.0278,  0.1690, -0.2982],\n",
        "#         [ 0.0557,"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "['▁J', \"'\", 'aime', '▁le', '▁ca', 'member', 't', '▁!']\n",
            "[5, 121, 11, 660, 16, 730, 25543, 110, 83, 6]\n",
            "13\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "v2tlWkqXQ8MS",
        "outputId": "b5611d4e-bab1-4d63-b56d-791988cd38d4"
      },
      "source": [
        "all_layer_embeddings[0].size()"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "torch.Size([1, 10, 768])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 40
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "id": "6-I208Qw7bTE"
      },
      "source": [
        "def encode_reviews(tokenizer, reviews, max_length):\n",
        "    token_ids = np.zeros(shape=(len(reviews), max_length),\n",
        "                         dtype=np.int32)\n",
        "    for i, review in enumerate(reviews):\n",
        "        encoded = tokenizer.encode(review, max_length=max_length)\n",
        "        token_ids[i, 0:len(encoded)] = encoded\n",
        "    attention_mask = (token_ids != 0).astype(np.int32)\n",
        "    return {\"input_ids\": token_ids, \"attention_mask\": attention_mask}\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "id": "S-wCL-Yt7bTF"
      },
      "source": [
        "class CamembertPreprocessor(BaseEstimator, TransformerMixin):\n",
        "    def __init__(self, tokenizer, max_seq_length,column):\n",
        "        self.tokenizer = tokenizer\n",
        "        self.max_seq_length = max_seq_length\n",
        "        self.column=column\n",
        "    def fit(self, X=None):\n",
        "        return self \n",
        "    \n",
        "    def transform(self, X, y):\n",
        "        # 1. Tokenize\n",
        "        X_encoded=X.copy()\n",
        "        X_encoded[self.column] = encode_reviews(self.tokenizer, X[columns], self.max_seq_length)\n",
        "        return X_encoded     \n",
        "    \n",
        "    def fit_transform(self, X, y):        \n",
        "        return self.transform(X, y)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "id": "kzl-z9mg7bTF"
      },
      "source": [
        "MAX_SEQ_LEN = 60 # in terms of generated tokens (not words)\n",
        "\n",
        "short_reviews = sum(np.array(reviews_len) <= MAX_SEQ_LEN)\n",
        "long_reviews = sum(np.array(reviews_len) > MAX_SEQ_LEN)\n",
        "\n",
        "print(\"{} reviews with LEN > {} ({:.2f} % of total data)\".format(\n",
        "    long_reviews,\n",
        "    MAX_SEQ_LEN,\n",
        "    100 * long_reviews / len(reviews_len)\n",
        "))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "REP17u487bTG"
      },
      "source": [
        "Simple sentence embedding strategy: mean of second to last hidden layer"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "id": "mub-gzsf7bTL"
      },
      "source": [
        "import torch.nn as nn \n",
        "import transformers as tt\n",
        "from transformers import CamembertModel, CamembertTokenizer\n",
        "class Second_to_last_SentenceEmbedding(nn.Module):\n",
        "    def __init__(self):\n",
        "        #super(, self).__init__()\n",
        "        super().__init__()\n",
        "        self.config = CamembertConfig.from_pretrained(\"camembert-base\", output_hidden_states=True)\n",
        "        self.camembert =CamembertModel.from_pretrained(\"camembert-base\", config=self.config)\n",
        "        self.fc1=nn.Linear(768,100)\n",
        "        self.relu=nn.ReLU()\n",
        "        self.fc2=nn.Linear(100,27)\n",
        "        for p in self.camembert.parameters():\n",
        "            p.requires_grad_(False)\n",
        "    def forward(self, input,attention_mask=None):\n",
        "        hidden = self.camembert(input,attention_mask=attention_mask)[2]\n",
        "        token_vecs=hidden[-2]\n",
        "        x=torch.mean(token_vecs,dim=1)\n",
        "      \n",
        "        return torch.mean(token_vecs,dim=1)"
      ],
      "execution_count": 43,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "id": "iwQWOBI77bTM"
      },
      "source": [
        "model=Second_to_last_SentenceEmbedding().to(device)"
      ],
      "execution_count": 44,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cUuuzAxx7bTM",
        "outputId": "9c287c8c-d454-411b-b76c-a4302209efbd"
      },
      "source": [
        "model.eval()"
      ],
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Second_to_last_SentenceEmbedding(\n",
              "  (camembert): CamembertModel(\n",
              "    (embeddings): RobertaEmbeddings(\n",
              "      (word_embeddings): Embedding(32005, 768, padding_idx=1)\n",
              "      (position_embeddings): Embedding(514, 768, padding_idx=1)\n",
              "      (token_type_embeddings): Embedding(1, 768)\n",
              "      (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "      (dropout): Dropout(p=0.1, inplace=False)\n",
              "    )\n",
              "    (encoder): RobertaEncoder(\n",
              "      (layer): ModuleList(\n",
              "        (0): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (1): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (2): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (3): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (4): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (5): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (6): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (7): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (8): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (9): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (10): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "        (11): RobertaLayer(\n",
              "          (attention): RobertaAttention(\n",
              "            (self): RobertaSelfAttention(\n",
              "              (query): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (key): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (value): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "            (output): RobertaSelfOutput(\n",
              "              (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "              (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "              (dropout): Dropout(p=0.1, inplace=False)\n",
              "            )\n",
              "          )\n",
              "          (intermediate): RobertaIntermediate(\n",
              "            (dense): Linear(in_features=768, out_features=3072, bias=True)\n",
              "          )\n",
              "          (output): RobertaOutput(\n",
              "            (dense): Linear(in_features=3072, out_features=768, bias=True)\n",
              "            (LayerNorm): LayerNorm((768,), eps=1e-05, elementwise_affine=True)\n",
              "            (dropout): Dropout(p=0.1, inplace=False)\n",
              "          )\n",
              "        )\n",
              "      )\n",
              "    )\n",
              "    (pooler): RobertaPooler(\n",
              "      (dense): Linear(in_features=768, out_features=768, bias=True)\n",
              "      (activation): Tanh()\n",
              "    )\n",
              "  )\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 45
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7-JsWGQLcSQx",
        "outputId": "f21cb05f-fafc-4ce8-a495-632e51de2729"
      },
      "source": [
        "x_train.shape"
      ],
      "execution_count": 48,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(84916, 5)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 48
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bjsw2LHuQq6P",
        "outputId": "86fa41f7-7205-424f-ce93-e774074302a1"
      },
      "source": [
        "model(encoded_sentence.to(device)).size()"
      ],
      "execution_count": 51,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "torch.Size([1, 768])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 51
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "__b07R5AjIim",
        "outputId": "5e493501-becb-4e71-db3b-9da5ea4afb83"
      },
      "source": [
        "tokenizer.encode_plus(\n",
        "                        \"J'aime le camembert !\",                      # Sentence to encode.\n",
        "                        add_special_tokens = True, # Add '[CLS]' and '[SEP]'\n",
        "                        max_length = 64,           # Pad & truncate all sentences.\n",
        "                        pad_to_max_length = True,\n",
        "                        return_attention_mask = True,   # Construct attn. masks.\n",
        "                        return_tensors = 'pt',     # Return pytorch tensors.\n",
        "                   )"
      ],
      "execution_count": 49,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Truncation was not explicitly activated but `max_length` is provided a specific value, please use `truncation=True` to explicitly truncate examples to max length. Defaulting to 'longest_first' truncation strategy. If you encode pairs of sequences (GLUE-style) with the tokenizer you can select this strategy more precisely by providing a specific strategy to `truncation`.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'input_ids': tensor([[    5,   121,    11,   660,    16,   730, 25543,   110,    83,     6,\n",
              "             1,     1,     1,     1,     1,     1,     1,     1,     1,     1,\n",
              "             1,     1,     1,     1,     1,     1,     1,     1,     1,     1,\n",
              "             1,     1,     1,     1,     1,     1,     1,     1,     1,     1,\n",
              "             1,     1,     1,     1,     1,     1,     1,     1,     1,     1,\n",
              "             1,     1,     1,     1,     1,     1,     1,     1,     1,     1,\n",
              "             1,     1,     1,     1]]), 'attention_mask': tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
              "         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,\n",
              "         0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]])}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 49
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bFjl06PflSkH"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3iLQDL9AQrWe"
      },
      "source": [
        "- Loading Image \n",
        "First we will build a custom Dataset uitilitaries and a Data Loader"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "trusted": true,
        "id": "b3Cr6YdV7bTN"
      },
      "source": [
        "from skimage import io, transform\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from torch.utils.data import Dataset, DataLoader\n",
        "from torchvision import transforms, utils\n"
      ],
      "execution_count": 42,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4Mmi-lXCanpS"
      },
      "source": [
        "class MultimodalDataset(Dataset):\n",
        "    \"\"\"Face Landmarks dataset.\"\"\"\n",
        "\n",
        "    def __init__(self,tokenizer, input_csv_file, root_dir,max_length=60 , transform=None,Y_csv_file=None):\n",
        "        \"\"\"\n",
        "        Args:\n",
        "            csv_file (string): Path to the csv file with annotations.\n",
        "            root_dir (string): Directory with all the images.\n",
        "            transform (callable, optional): Optional transform to be applied\n",
        "                on a sample.\n",
        "        \"\"\"\n",
        "        self.input_file = pd.read_csv(input_csv_file)\n",
        "        self.root_dir = root_dir\n",
        "        self.transform = transform\n",
        "        self.max_length=max_length\n",
        "        self.input_file[\"image_path\"]=(self.input_file[\"imageid\"].apply(lambda x: str(int(x)))+\"_product_\"+self.input_file[\"productid\"].apply(lambda x: str(int(x)))+\".jpg\").apply(lambda x: \"image_\"+x)\n",
        "        if Y_csv_file !=None:\n",
        "            self.output = pd.read_csv(Y_csv_file)\n",
        "            self.classes = list(set(self.output[\"prdtypecode\"].values))\n",
        "        else:\n",
        "            self.ouptut=None\n",
        "    def __len__(self):\n",
        "        return len(self.input_file)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        if torch.is_tensor(idx):\n",
        "            idx = idx.tolist()\n",
        "        \n",
        "        img_name = os.path.join(self.root_dir,\n",
        "                                self.input_file[\"image_path\"].iloc[idx])\n",
        "        image = io.imread(img_name)\n",
        "        text = self.input_file[\"designation\"].iloc[idx]\n",
        "        encoded_text = tokenizer.encode_plus(\n",
        "                        text,                      # Sentence to encode.\n",
        "                        add_special_tokens = True, # Add '[CLS]' and '[SEP]'\n",
        "                        max_length = 64,           # Pad & truncate all sentences.\n",
        "                        pad_to_max_length = True,\n",
        "                        return_attention_mask = True,   # Construct attn. masks.\n",
        "                        return_tensors = 'pt',     # Return pytorch tensors.\n",
        "                   )\n",
        "        sample = {'image': image, 'input_ids':encoded_text['input_ids'] ,'attention_mask':encoded_text['attention_mask']}\n",
        "\n",
        "        if self.transform:\n",
        "            sample['image'] = self.transform(sample['image'])\n",
        "        if self.output is None:\n",
        "            return sample\n",
        "        sample[\"label\"]=torch.tensor(self.classes.index(self.output[\"prdtypecode\"].iloc[idx]))\n",
        "        return sample"
      ],
      "execution_count": 79,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 269
        },
        "id": "j3crHdoKjB5l",
        "outputId": "89a755bb-a20e-4ab6-8100-d06fdc2e8d83"
      },
      "source": [
        "from __future__ import print_function, division\n",
        "import os\n",
        "import torch\n",
        "import pandas as pd\n",
        "from skimage import io, transform\n",
        "import numpy as np\n",
        "def show_landmarks(image):\n",
        "    \"\"\"Show image with landmarks\"\"\"\n",
        "    plt.imshow(image)\n",
        "    plt.pause(0.001)  # pause a bit so that plots are updated\n",
        "\n",
        "plt.figure()\n",
        "show_landmarks(io.imread(os.path.join('/content/images/images/image_train', new_data[\"image_path\"].iloc[25])))\n",
        "plt.show()"
      ],
      "execution_count": 60,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a7_C0VvE1Bfm"
      },
      "source": [
        "?np.random.random_integers"
      ],
      "execution_count": 70,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 554
        },
        "id": "mCHu41yq06Gw",
        "outputId": "f5b08e7c-560f-47c2-e4da-27862405a47a"
      },
      "source": [
        "face_dataset = MultimodalDataset(tokenizer=tokenizer, input_csv_file=\"/content/X_train_update.csv\", root_dir=\"/content/images/images/image_train\",max_length=60 , transform=None,Y_csv_file=\"/content/Y_train_CVw08PX.csv\")\n",
        "\n",
        "fig = plt.figure()\n",
        "l=np.random.random_integers(0,84916,6)\n",
        "for i in range(len(l)):\n",
        "    sample = face_dataset[l[i]]\n",
        "\n",
        "    print(i, sample['image'].shape)\n",
        "\n",
        "    ax = plt.subplot(1, 6, i + 1)\n",
        "    plt.tight_layout()\n",
        "    ax.set_title('Sample #{}'.format(face_dataset.classes[sample[\"label\"]]))\n",
        "    ax.axis('off')\n",
        "    show_landmarks(sample[\"image\"])\n",
        "    plt.show()"
      ],
      "execution_count": 93,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0 (500, 500, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "1 (500, 500, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "2 (500, 500, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "3 (500, 500, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "4 (500, 500, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "5 (500, 500, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UherlV3_6eNn",
        "outputId": "5d167466-4d31-4677-9300-f06aa0a1a01b"
      },
      "source": [
        "4%3"
      ],
      "execution_count": 91,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 91
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1u5tLJSMsNqy"
      },
      "source": [
        "height=[]\n",
        "width=[]\n",
        "for i in range(84916):\n",
        "  image=io.imread(os.path.join('/content/images/images/image_train', new_data[\"image_path\"].iloc[i]))\n",
        "  height.append(image.shape[0])\n",
        "  width.append(image.shape[1])"
      ],
      "execution_count": 63,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bCAA7s2zxF8L"
      },
      "source": [
        "Training Dataset"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qTqW3cFv7FTx"
      },
      "source": [
        "Image Transformation (Data Augmentation)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OT2wJ51s7Ejk"
      },
      "source": [
        "class Microscope:\n",
        "    \"\"\"\n",
        "    Cutting out the edges around the center circle of the image\n",
        "    Imitating a picture, taken through the microscope\n",
        "\n",
        "    Args:\n",
        "        p (float): probability of applying an augmentation\n",
        "    \"\"\"\n",
        "\n",
        "    def __init__(self, p: float = 0.5):\n",
        "        self.p = p\n",
        "\n",
        "    def __call__(self, img):\n",
        "        \"\"\"\n",
        "        Args:\n",
        "            img (PIL Image): Image to apply transformation to.\n",
        "\n",
        "        Returns:\n",
        "            PIL Image: Image with transformation.\n",
        "        \"\"\"\n",
        "        if random.random() < self.p:\n",
        "            circle = cv2.circle((np.ones(img.shape) * 255).astype(np.uint8), # image placeholder\n",
        "                        (img.shape[0]//2, img.shape[1]//2), # center point of circle\n",
        "                        random.randint(img.shape[0]//2 - 3, img.shape[0]//2 + 15), # radius\n",
        "                        (0, 0, 0), # color\n",
        "                        -1)\n",
        "\n",
        "            mask = circle - 255\n",
        "            img = np.multiply(img, mask)\n",
        "        \n",
        "        return img\n",
        "\n",
        "    def __repr__(self):\n",
        "        return f'{self.__class__.__name__}(p={self.p})'"
      ],
      "execution_count": 94,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3BeUZDJ98U-b"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b_CLxxMX7sT9"
      },
      "source": [
        "train_transform = transforms.Compose([\n",
        "    Microscope(p=0.5),\n",
        "    transforms.ToPILImage(),\n",
        "    transforms.RandomResizedCrop(size=384, scale=(0.8, 1.0)),\n",
        "    transforms.RandomHorizontalFlip(),\n",
        "    transforms.RandomVerticalFlip(),\n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize(mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225])\n",
        "])\n",
        "test_transform = transforms.Compose([\n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize(mean=[0.485, 0.456, 0.406],std=[0.229, 0.224, 0.225])\n",
        "])"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}