TP08_introduction_a_pytorch.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "gpuType": "T4",
      "mount_file_id": "11PZffCG2uoFOwOlaG9QOou5isUgdL-KB",
      "authorship_tag": "ABX9TyPIq3yOPdq7iTbk7gMwDAzB",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/gguex/ISH_ressources_cours_ML/blob/main/TP08_introduction_a_pytorch.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# TP 8 : Introduction à PyTorch\n",
        "\n",
        "Dans ce TP, nous allons voir comment prendre en main PyTorch (https://pytorch.org/). PyTorch, à l'instar de TansorFlow, Keras, Theano et Caffe, est une librarie de Machine Learning en Python spécialement conçue pour créer des réseaux de neurones. Elle est souvent utilisée dans le monde académique car elle possède une architecture flexible et est rapide, mais son style reste néanmoins très \"pythonique\".\n",
        "\n",
        "Ce TP est divisé en trois parties :\n",
        "\n",
        "1. Les Tenseurs : Les tenseurs sont les variables utilisées par PyTorch (et par beaucoup de librairies de Machine Learning). Ce sont des tableaux de données à plusieurs dimensions, similaires aux arrays dans numpy, mais avec quelques spécificités en plus. Les opérations sur les tenseurs constituent les fondations des librairies telles que PyTorch.\n",
        "2. les Données : PyTorch permet de structurer et de manipuler les données à l'aide des classe `Datasets` et `DataLoaders`. L'utilisation de transformations permet d'isoler les opérations de pré-traitement effectuées sur le données.\n",
        "3. Les Modèles : Nous verrons ici comment charger des modèles pré-traités, faire son propre modèle, puis comment l'entrainer et l'évaluer. Nous verrons également les opérations de sauvegarde et chargement de ces derniers.\n",
        "\n",
        "N'hésitez pas à prendre le temps de faire et refaire ce TP, tout en jouant avec le code. PyTorch est une librairie qui demande un certain temps d'adaptation."
      ],
      "metadata": {
        "id": "UwghwpWkl3-w"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Les libraires et modules nécessaires pour ce TP sont les suivants."
      ],
      "metadata": {
        "id": "SGejkxb-qSdt"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "id": "emBNMQs2dEMW"
      },
      "outputs": [],
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "# La libraire pytorch, qui contient de nombreux sous-modules\n",
        "import torch\n",
        "# Il y a trois librairies complémentaires dans PyTorch : torchvision, torchtext\n",
        "# et torchaudio. Ces dernières contiennent des datasets, modèles, et fonctions\n",
        "# spécifiques, respectivement, à l'image, au texte et à l'audio.\n",
        "from torchvision import datasets\n",
        "from torchvision import transforms\n",
        "from torchvision import models\n",
        "# Ces fonctions et classes nous permettent de traiter les jeux de données\n",
        "from torch.utils.data import Dataset, DataLoader, random_split\n",
        "# Grâce au module torch.nn, on peut construire des réseaux de neurones.\n",
        "import torch.nn as nn\n",
        "# On abrège fréquemment le module torch.nn.functional aussi, qui contient les\n",
        "# fonctionnelles.\n",
        "import torch.nn.functional as F\n",
        "# Ce module permet de charger des images et de les mettres dans des objets \"PIL\"\n",
        "from PIL import Image"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Une des grandes forces des librairies telles que PyTorch est de pouvoir utiliser les GPU (Graphical Processing Units) ou les TPU (Tensor Processing Units) des machines. Ces GPU et TPU accèlèrent très fortement les opérations effectuées sur les tenseurs et donc l'entrainement des modèles.\n",
        "\n",
        "Pour activer le calcul GPU ou CPU sur Colab, vous devez tout d'abord modifier le type d'exécution dans le menu **Exécution -> Modifier le type d'éxécution**, puis choisir une option dans le menu déroulant **Accélérateur matériel**. Notez qu'avec la version gratuite de Colab, vous n'allez pas toujours avoir accès à ces cartes (les TPU sont particulièrement difficiles à avoir).\n",
        "\n",
        "Sur votre machine personnelle, il vous faut une carte GPU et TPU installée, puis suivre les instructions que vous pouvez trouver sous : https://pytorch.org/get-started/locally/. Notez que mettre tout en place pour bénéficier d'une accélération matérielle sur un ordinateur personnel peut être une opération longue et fastidieuse.\n",
        "\n",
        "Le code suivant permet d'enregistrer le dispositif (`device`) qui va effectuer les calculs en regardant automatiquement si vous avez accès à une carte graphique ou tensorielle (qui utilise alors la librairie `cuda`). La variable `device` stockant le dispositif utilisé va être réutilisée tout le long de ce code."
      ],
      "metadata": {
        "id": "Uk3Ks4vPvOS6"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
        "print(device)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JYR0XxQYrsQ7",
        "outputId": "09258183-39af-4f7b-9d3f-721864657e73"
      },
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "cuda\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "-ISFaIUJMhHO"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 1. Les Tenseurs\n",
        "\n",
        "Les tensors s'apparentent au *arrays* de numpy : ce sont également des tableaux multidimensionnels qui contiennent qu'un seul type de données. Ils ont cependant deux fonctionnalités fondamentales qui font qu'ils sont utilisés pour créer des modèles de machine learning : leurs calculs peuvent être accélérés matériellement (GPU ou TPU) et ils possèdent un méchanisme de *differentiation automatique*.  \n",
        "\n",
        "Dans cette partie, nous allons voir comment créer, manipuler et faire des calculs avec des tenseurs. On peut trouver la liste exhaustive des opérations sur les tenseurs sous https://pytorch.org/docs/stable/tensors.html#torch.Tensor."
      ],
      "metadata": {
        "id": "8jG5t27CnQUG"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 1.1 Création et Attributs"
      ],
      "metadata": {
        "id": "akdlGQKzrJRI"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "On peut créer des tenseurs avec `torch.tensor()` et à partir de liste multidimentionnelles (ou à partir d'*arrays* numpy). Ils peuvent contenir des nombres entiers, des nombres à virgule ou des booléens."
      ],
      "metadata": {
        "id": "aUxvbIpX03kt"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.tensor([[1.1, 2.2, 3.3],\n",
        "                  [4.4, 5.5, 6.6],\n",
        "                  [7.7, 8.8, 9.9]])\n",
        "y = torch.tensor([[[1, 2, 3, 4],\n",
        "                   [5, 6, 7, 8]],\n",
        "                  [[9, 10, 11, 12],\n",
        "                   [13, 14, 15, 16]],\n",
        "                  [[17, 18, 19, 20],\n",
        "                   [21, 22, 23, 24]]])\n",
        "z = torch.tensor([True, False, True])\n",
        "print(x)\n",
        "print(y)\n",
        "print(z)"
      ],
      "metadata": {
        "id": "wHVyonl1_EGX",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ec5a9457-f743-49bf-b1e5-42ebd8190905"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[1.1000, 2.2000, 3.3000],\n",
            "        [4.4000, 5.5000, 6.6000],\n",
            "        [7.7000, 8.8000, 9.9000]])\n",
            "tensor([[[ 1,  2,  3,  4],\n",
            "         [ 5,  6,  7,  8]],\n",
            "\n",
            "        [[ 9, 10, 11, 12],\n",
            "         [13, 14, 15, 16]],\n",
            "\n",
            "        [[17, 18, 19, 20],\n",
            "         [21, 22, 23, 24]]])\n",
            "tensor([ True, False,  True])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Les tenseurs possèdent de nombreux attributs, les plus utilisés sont :\n",
        "* shape : la taille dans les différentes dimensions du tenseur.\n",
        "* dtype : le type de données du tenseur\n",
        "* ndim : le nombre de dimensions du tenseur"
      ],
      "metadata": {
        "id": "hicCMWrH1cdX"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(f\"x.shape = {x.shape}\")\n",
        "print(f\"x.dtype = {x.dtype}\")\n",
        "print(f\"x.ndim = {x.ndim}\")\n",
        "print(f\"y.shape = {y.shape}\")\n",
        "print(f\"y.dtype = {y.dtype}\")\n",
        "print(f\"y.ndim = {y.ndim}\")\n",
        "print(f\"x.shape = {z.shape}\")\n",
        "print(f\"x.dtype = {z.dtype}\")\n",
        "print(f\"x.ndim = {z.ndim}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YN9P7me5pMFy",
        "outputId": "3f94eb9e-2b1d-45d4-dc8c-1f4b0e768e74"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "x.shape = torch.Size([3, 3])\n",
            "x.dtype = torch.float32\n",
            "x.ndim = 2\n",
            "y.shape = torch.Size([3, 2, 4])\n",
            "y.dtype = torch.int64\n",
            "y.ndim = 3\n",
            "x.shape = torch.Size([3])\n",
            "x.dtype = torch.bool\n",
            "x.ndim = 1\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Plusieurs fonctions permettent de créer des tenseurs particuliers. Par exemple,  `torch.zeros(size)` va créer un tenseur nul avec la forme définie par `size`."
      ],
      "metadata": {
        "id": "Q1D332mg2BmQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(torch.zeros((2, 3, 4)))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rANnwPhBov4S",
        "outputId": "151440f4-83e7-4394-c241-575cb33dfbd5"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[[0., 0., 0., 0.],\n",
            "         [0., 0., 0., 0.],\n",
            "         [0., 0., 0., 0.]],\n",
            "\n",
            "        [[0., 0., 0., 0.],\n",
            "         [0., 0., 0., 0.],\n",
            "         [0., 0., 0., 0.]]])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "`torch.ones(size)` fait de même avec un tenseur contenant que des 1."
      ],
      "metadata": {
        "id": "cVwaYAIL2r7a"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(torch.ones(2))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CHla6I1vo8Db",
        "outputId": "318614d0-f8a3-4f6c-e099-5ea8c20e14d1"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([1., 1.])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "La fonction `torch.eye(n)` crée une matrice identité de taille $n \\times n$."
      ],
      "metadata": {
        "id": "CXZ_B0yb23Nr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(torch.eye(3))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "W7bMHj6Yo-_6",
        "outputId": "e528b9e6-8aae-4d8d-ce47-9f5604033f67"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[1., 0., 0.],\n",
            "        [0., 1., 0.],\n",
            "        [0., 0., 1.]])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 1.2 Indexation et Opérations\n",
        "\n",
        "L'indexation et les opérations sur des tenseurs partagent de nombreux points commun avec les *arrays* de numpy."
      ],
      "metadata": {
        "id": "2974QY-csQ31"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "a = torch.tensor([[1.1, 2.2, 3.3],\n",
        "                  [4.4, 5.5, 6.6],\n",
        "                  [7.7, 8.8, 9.9]])\n",
        "b = torch.tensor([[10, 100, 1000],\n",
        "                  [20, 200, 2000],\n",
        "                  [30, 300, 3000]])\n",
        "c = torch.tensor([1, 10, 100])"
      ],
      "metadata": {
        "id": "DQgpYD7MuWnQ"
      },
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "L'indexation se fait avec les crochets, avec un indiçage dans l'ordre des dimensions. Notez que l'on peut utiliser un *slicing* de type `start:end:step` sur n'importe quelle dimension. Lorsque l'on a un seul nombre dans un tenseur, on peut l'obtenir avec la méthode `item()`."
      ],
      "metadata": {
        "id": "OgcdJoHK4c7t"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(a[0, 0])\n",
        "print(a[0, 0].item())\n",
        "print(a[0:2, 1:3])\n",
        "print(a[0:3:2, 0])\n",
        "print(b[1]) # c'est une liste emboitée\n",
        "print(b[1, :]) # même chose\n",
        "print(b[:, 0])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cblRw_xpw6YI",
        "outputId": "5122ebd0-0ed7-4d86-96be-c72a6ed03eaa"
      },
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor(1.1000)\n",
            "1.100000023841858\n",
            "tensor([[2.2000, 3.3000],\n",
            "        [5.5000, 6.6000]])\n",
            "tensor([1.1000, 7.7000])\n",
            "tensor([  20,  200, 2000])\n",
            "tensor([  20,  200, 2000])\n",
            "tensor([10, 20, 30])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Notez que l'on peut aussi accèder aux éléments en soumettant un tenseur de booléens."
      ],
      "metadata": {
        "id": "4A9fI5Dy5Svv"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(c[torch.tensor([True, False, True])])\n",
        "print(b[:, torch.tensor([False, True, True])])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7ZcZaVNbxy2K",
        "outputId": "b83e9c0b-5e26-4f5b-f762-80513c921052"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([  1, 100])\n",
            "tensor([[ 100, 1000],\n",
            "        [ 200, 2000],\n",
            "        [ 300, 3000]])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Les opérateurs usuels ainsi que certaines fonctions de calcul s'appliquent composantes par composantes sur les tenseurs. Si un tenseur utilisé lors d'un calcul est plus petit qu'un autre, on va dupliquer le tenseur plus petit jusqu'à qu'il soit de la même taille que le grand. On appelle cette opération le *casting*."
      ],
      "metadata": {
        "id": "TBWbfFj753n9"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(b / 100)\n",
        "print(a + b)\n",
        "print(a.add(b)) # correspond à a + b\n",
        "print(torch.log(a))\n",
        "print(a * c) # casting"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pe25CFdKnn8r",
        "outputId": "0c6d84f4-37a7-4b79-ddd0-388b7b9f6e0b"
      },
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[ 0.1000,  1.0000, 10.0000],\n",
            "        [ 0.2000,  2.0000, 20.0000],\n",
            "        [ 0.3000,  3.0000, 30.0000]])\n",
            "tensor([[  11.1000,  102.2000, 1003.3000],\n",
            "        [  24.4000,  205.5000, 2006.6000],\n",
            "        [  37.7000,  308.8000, 3009.8999]])\n",
            "tensor([[  11.1000,  102.2000, 1003.3000],\n",
            "        [  24.4000,  205.5000, 2006.6000],\n",
            "        [  37.7000,  308.8000, 3009.8999]])\n",
            "tensor([[0.0953, 0.7885, 1.1939],\n",
            "        [1.4816, 1.7047, 1.8871],\n",
            "        [2.0412, 2.1748, 2.2925]])\n",
            "tensor([[  1.1000,  22.0000, 330.0000],\n",
            "        [  4.4000,  55.0000, 660.0000],\n",
            "        [  7.7000,  88.0000, 989.9999]])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Les opérateurs logiques fonctionnent également sur les tenseurs, ce qui permet d'avoir une syntaxe élégante pour sélectionner des composantes vérifiant une certaine propriété."
      ],
      "metadata": {
        "id": "dRBXhhOS7Qzj"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(a < (b / 100))\n",
        "print(a[a < 5])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tquy-RJxogTM",
        "outputId": "43d8dcd2-af96-4ae1-ba96-3e3ff551de62"
      },
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[False, False,  True],\n",
            "        [False, False,  True],\n",
            "        [False, False,  True]])\n",
            "tensor([1.1000, 2.2000, 3.3000, 4.4000])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "On peut également faire des opérations sur les marges des tenseurs."
      ],
      "metadata": {
        "id": "gm09TaCn7fXE"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(b.sum())\n",
        "print(b.sum(dim=0))\n",
        "print(a.mean(dim=0))\n",
        "print(a.max(dim=1))\n",
        "print(a.argmax(dim=1))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CAPjBQKzzjT_",
        "outputId": "b6bbbd87-79dd-43fc-9ac8-5c9c1220ecee"
      },
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor(6660)\n",
            "tensor([  60,  600, 6000])\n",
            "tensor([4.4000, 5.5000, 6.6000])\n",
            "torch.return_types.max(\n",
            "values=tensor([3.3000, 6.6000, 9.9000]),\n",
            "indices=tensor([2, 2, 2]))\n",
            "tensor([2, 2, 2])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Il existe des opérations pour créer des tenseurs à partir d'autres, en les empilants ou en séparant des tenseurs."
      ],
      "metadata": {
        "id": "dgpg1x607d3y"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(torch.cat([a, b])) # empilement par ligne\n",
        "print(torch.cat([a, b], axis=1)) # empilement par colonne\n",
        "print(torch.stack([a, b])) # une nouvelle dimension contenant les deux tenseurs\n",
        "print(b.unbind(dim=1)) # séparation selon la dimension"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pFUXo3Z-zqBS",
        "outputId": "4ca82d86-bf62-4250-e0b9-fd07b29df77a"
      },
      "execution_count": 14,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[1.1000e+00, 2.2000e+00, 3.3000e+00],\n",
            "        [4.4000e+00, 5.5000e+00, 6.6000e+00],\n",
            "        [7.7000e+00, 8.8000e+00, 9.9000e+00],\n",
            "        [1.0000e+01, 1.0000e+02, 1.0000e+03],\n",
            "        [2.0000e+01, 2.0000e+02, 2.0000e+03],\n",
            "        [3.0000e+01, 3.0000e+02, 3.0000e+03]])\n",
            "tensor([[1.1000e+00, 2.2000e+00, 3.3000e+00, 1.0000e+01, 1.0000e+02, 1.0000e+03],\n",
            "        [4.4000e+00, 5.5000e+00, 6.6000e+00, 2.0000e+01, 2.0000e+02, 2.0000e+03],\n",
            "        [7.7000e+00, 8.8000e+00, 9.9000e+00, 3.0000e+01, 3.0000e+02, 3.0000e+03]])\n",
            "tensor([[[1.1000e+00, 2.2000e+00, 3.3000e+00],\n",
            "         [4.4000e+00, 5.5000e+00, 6.6000e+00],\n",
            "         [7.7000e+00, 8.8000e+00, 9.9000e+00]],\n",
            "\n",
            "        [[1.0000e+01, 1.0000e+02, 1.0000e+03],\n",
            "         [2.0000e+01, 2.0000e+02, 2.0000e+03],\n",
            "         [3.0000e+01, 3.0000e+02, 3.0000e+03]]])\n",
            "(tensor([10, 20, 30]), tensor([100, 200, 300]), tensor([1000, 2000, 3000]))\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Les tenseurs peuvent également être redimensionnés"
      ],
      "metadata": {
        "id": "79eqzekA8e2b"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "d = torch.cat([a, b], axis=1)\n",
        "print(d)\n",
        "print(d.view(2, 9)) # Nouvelle forme : (2, 9)\n",
        "print(d.view(-1, 2)) # Nouvelle forme : (?, 2) - le -1 désigne ce qu'il reste\n",
        "print(d.flatten()) # Applitissement en 1 dimension\n",
        "print(d.unsqueeze(0)) # Création d'une nouvelle dimension en 0"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TogeTqff0j--",
        "outputId": "ecba7ded-4e23-426d-a6af-dd37127cbb1d"
      },
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[1.1000e+00, 2.2000e+00, 3.3000e+00, 1.0000e+01, 1.0000e+02, 1.0000e+03],\n",
            "        [4.4000e+00, 5.5000e+00, 6.6000e+00, 2.0000e+01, 2.0000e+02, 2.0000e+03],\n",
            "        [7.7000e+00, 8.8000e+00, 9.9000e+00, 3.0000e+01, 3.0000e+02, 3.0000e+03]])\n",
            "tensor([[1.1000e+00, 2.2000e+00, 3.3000e+00, 1.0000e+01, 1.0000e+02, 1.0000e+03,\n",
            "         4.4000e+00, 5.5000e+00, 6.6000e+00],\n",
            "        [2.0000e+01, 2.0000e+02, 2.0000e+03, 7.7000e+00, 8.8000e+00, 9.9000e+00,\n",
            "         3.0000e+01, 3.0000e+02, 3.0000e+03]])\n",
            "tensor([[1.1000e+00, 2.2000e+00],\n",
            "        [3.3000e+00, 1.0000e+01],\n",
            "        [1.0000e+02, 1.0000e+03],\n",
            "        [4.4000e+00, 5.5000e+00],\n",
            "        [6.6000e+00, 2.0000e+01],\n",
            "        [2.0000e+02, 2.0000e+03],\n",
            "        [7.7000e+00, 8.8000e+00],\n",
            "        [9.9000e+00, 3.0000e+01],\n",
            "        [3.0000e+02, 3.0000e+03]])\n",
            "tensor([1.1000e+00, 2.2000e+00, 3.3000e+00, 1.0000e+01, 1.0000e+02, 1.0000e+03,\n",
            "        4.4000e+00, 5.5000e+00, 6.6000e+00, 2.0000e+01, 2.0000e+02, 2.0000e+03,\n",
            "        7.7000e+00, 8.8000e+00, 9.9000e+00, 3.0000e+01, 3.0000e+02, 3.0000e+03])\n",
            "tensor([[[1.1000e+00, 2.2000e+00, 3.3000e+00, 1.0000e+01, 1.0000e+02,\n",
            "          1.0000e+03],\n",
            "         [4.4000e+00, 5.5000e+00, 6.6000e+00, 2.0000e+01, 2.0000e+02,\n",
            "          2.0000e+03],\n",
            "         [7.7000e+00, 8.8000e+00, 9.9000e+00, 3.0000e+01, 3.0000e+02,\n",
            "          3.0000e+03]]])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(d.shape)\n",
        "d.unsqueeze(0).shape"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "o9vZR374uvFZ",
        "outputId": "e2c76dc0-7264-4890-ab47-1c1f28cd0b22"
      },
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([3, 6])\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "torch.Size([1, 3, 6])"
            ]
          },
          "metadata": {},
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Ils existent de nombreuses opérations matricielles possible sur les tenseurs. Attention à faire en sorte que les tenseurs soient du même type de données."
      ],
      "metadata": {
        "id": "U4hDvyKf83kl"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(a.t()) # transposée\n",
        "print(a.inverse()) # inverse\n",
        "c = c.to(torch.float32) # transformation du dtype\n",
        "print(a @ c) # mutliplication matricielle"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FYgzoWHg-AEP",
        "outputId": "85a78df4-43b7-4560-d2f1-fa9e736933a1"
      },
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[1.1000, 4.4000, 7.7000],\n",
            "        [2.2000, 5.5000, 8.8000],\n",
            "        [3.3000, 6.6000, 9.9000]])\n",
            "tensor([[ -676501.1875,  1353001.5000,  -676500.6250],\n",
            "        [ 1353000.7500, -2706002.5000,  1353001.3750],\n",
            "        [ -676499.8125,  1353001.2500,  -676500.7500]])\n",
            "tensor([ 353.1000,  719.4000, 1085.7000])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 1.3 Device et Autograd"
      ],
      "metadata": {
        "id": "8NHNWJUo_wq4"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "Les tenseurs sont attribués à un dispositif de calcul particulier (CPU ou GPU). Par défaut, il s'agit du CPU (`cpu`), mais on peut les aussi les basculer sur un autre dispositif à tout moment grâce à la méthode `to(device)`. S'ils sont sur le GPU, leur attribut `device` sera alors égal à `cuda` ou `cuda:0`. Notez que le `0` de `cuda:0` désigne le numéro du GPU (cette précision n'est utile que si vous disposez de plusieurs GPU)."
      ],
      "metadata": {
        "id": "K0l1jr4O9QOc"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(a.device)\n",
        "a = a.to(device)\n",
        "print(a.device)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "S-9_xu_O_7wd",
        "outputId": "0f981083-db45-41ec-c243-7a21f87ab97a"
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "cpu\n",
            "cuda:0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "On peut préciser le type de donnée et le dispositif de calcul d'un tenseur lors de sa création."
      ],
      "metadata": {
        "id": "hrqVwNW2-UBD"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "c_gpu = torch.tensor([1, 10, 100], dtype=torch.float32, device=device)\n",
        "print(c_gpu.dtype)\n",
        "print(c_gpu.device)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JV9eTtmdApxj",
        "outputId": "3bdcdaf5-c923-4cde-d50e-8a06719a7d48"
      },
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.float32\n",
            "cuda:0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Attention, un calcul entre deux tenseurs ne peut s'effectuer que si les deux résident sur le même dispositif de calcul."
      ],
      "metadata": {
        "id": "DWaEwjRU-l70"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# print(a @ c) <- ERREUR SI UTILISATION GPU\n",
        "print(a @ c_gpu)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yh7TO8poBNL0",
        "outputId": "93ce0ece-d8b6-4d79-97e5-57df4ea25c3f"
      },
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([ 353.1000,  719.4000, 1085.7000], device='cuda:0')\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Un des énorme avantages des tenseurs sont qu'ils possèdent une méchanisme de **différentiation automatique** (voir https://pytorch.org/tutorials/beginner/basics/autogradqs_tutorial.html). Cette différentiation automatique permet de calculer automatiquement le gradient $\\frac{\\partial f(x)}{\\partial x}$ pour un tenseur $x$ donné et une fonction $f(x)$ particulière.\n",
        "\n",
        "Pour permettre la différentiation automatique, on doit créer un tenseur avec l'attribut `requires_grad=True`. Le gradient peut alors se trouver grâce à l'attribut `grad` (au début, il est vide)."
      ],
      "metadata": {
        "id": "wKvvTW2V-0RJ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x = torch.tensor([[1, 2, 3],\n",
        "                  [4, 5, 6]], dtype=torch.float32, requires_grad=True)\n",
        "print(x.requires_grad)\n",
        "print(x.grad)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hiGHmbvEBceZ",
        "outputId": "952e3ea6-5f5e-4df4-f650-b2885440e1bd"
      },
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "True\n",
            "None\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "En créant maintenant un nouveau tenseur à partir de ce premier, par exemple `f=(x**2).sum()` qui représente $f(x)$, on peut demander d'effectuer le calcul de $\\frac{\\partial f(x)}{\\partial x}$ avec la méthode `f.backward()`. Le gradient se trouvera alors stocké dans l'attribut `x.grad`."
      ],
      "metadata": {
        "id": "GP82Mc0N_uo_"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "f = (x ** 2).sum()\n",
        "print(f)\n",
        "f.backward()\n",
        "print(x.grad)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "q21YKNmrCLyv",
        "outputId": "ab6816ea-09f0-4e08-cd1e-2d056773e52c"
      },
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor(91., grad_fn=<SumBackward0>)\n",
            "tensor([[ 2.,  4.,  6.],\n",
            "        [ 8., 10., 12.]])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Attention cependant, l'attibut `x.grad` va accumuler tous les gradients demandés par les méthodes `backward()`."
      ],
      "metadata": {
        "id": "F_u8SN6aAZNW"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "for i in range(3):\n",
        "  f = (x ** 2).sum()\n",
        "  f.backward()\n",
        "  print(x.grad)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OT_7tZ-2CbpB",
        "outputId": "a8c35f8c-ca7d-4c79-e4b7-1fdfd94535e9"
      },
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[ 4.,  8., 12.],\n",
            "        [16., 20., 24.]])\n",
            "tensor([[ 6., 12., 18.],\n",
            "        [24., 30., 36.]])\n",
            "tensor([[ 8., 16., 24.],\n",
            "        [32., 40., 48.]])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Pour réinitialiser un gradient, on doit utiliser la méthode `x.grad.zero_()` (avec le suffixe `_` qui modifie directement l'objet). On peut alors calculer un nouveau gradient."
      ],
      "metadata": {
        "id": "_VBzd5BqAxSY"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "x.grad.zero_()\n",
        "print(x.grad)\n",
        "f2 = torch.log(x ** 3).sum()\n",
        "print(f2)\n",
        "f2.backward()\n",
        "print(x.grad)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ilO3S_teDQt5",
        "outputId": "078aa2e5-ef2e-47e8-a1a9-4a4bb217e6f7"
      },
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[0., 0., 0.],\n",
            "        [0., 0., 0.]])\n",
            "tensor(19.7378, grad_fn=<SumBackward0>)\n",
            "tensor([[3.0000, 1.5000, 1.0000],\n",
            "        [0.7500, 0.6000, 0.5000]])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "C'est ce méchanisme d'autodifférentiation qui rend la librairie PyTorch si puissante : on peut faire des calculs très complexes avec des données et de nombreux paramètres, mais retrouver très facilement les gradients qui nous permettrons d'optimiser une fonctionnelle particulière."
      ],
      "metadata": {
        "id": "aQ73_ooWBGE6"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "f2YSVoTOMeZ8"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 2. Les Données\n",
        "\n",
        "PyTorch possède deux classes qui sont utilisés pour gérer nos jeux de données :  `Dataset` et `DataLoader`.\n",
        "\n",
        "La classe `Dataset`, dont nous sommes invités à créer une classe fille, permet de gérer un jeu de donnée. Elle est particulièrement utile pour les gros jeux de données ou les jeux de données distants, où les individus du jeu de données sont chargés dans la mémoire uniquement lorsqu'ils sont utilisés. La classe `Dataset` va également opérer certaines transformations sur les données, afin de pouvoir les utiliser dans les modèles.\n",
        "\n",
        "La classe `Dataloader` permet de créer un **conteneur itérable** au dessus (surcouche) d'un jeu de données de type `Dataset`. Ce conteneur itérable permet de tirer les données par **batch**, selon un schéma d'échantillonnage prédifini, et sera vide lorsque toutes les données auront été tirées. On utilisera ces `Dataloader` lors des étapes d'entrainement, de validation et de test."
      ],
      "metadata": {
        "id": "rDiUXOlbMSMd"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 2.1 Jeux de Données Intégrés\n",
        "\n",
        "PyTorch possède de nombreux jeux de données intégrés (par exemple, pour la vision https://pytorch.org/vision/stable/datasets.html). Ces jeux de données sont déjà sous la forme d'un objet de classe `Dataset`, ce qui nous permettera, dans un premier temps, de voir comment les utiliser.\n",
        "\n",
        "Nous téléchargeons ici le jeu d'entrainement de la base de données *CIFAR10* (https://www.cs.toronto.edu/~kriz/cifar.html), qui contient 50'000 32x32 images en couleur, classées dans 10 classes. Dans la fonction suivante, le paramètre `root=\"./cifar10_train/\"` va sauvegarder le jeu de données sous le dossier `\"./cifar10_train/\"`, `train=True` précise que nous voulons le jeu d'entrainement, et `download=True` permet de télécharger le dataset."
      ],
      "metadata": {
        "id": "5yWs9sXvnzhY"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "cifar10_train = datasets.CIFAR10(root=\"./cifar10_train/\",\n",
        "                                 train=True,\n",
        "                                 download=True)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xdhbYoKNf3ix",
        "outputId": "ad1912bc-d6fb-4776-99c3-a9211fbcbce0"
      },
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Files already downloaded and verified\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "La variable `cifar10_train` contient maintenant une instance de la classe `Dataset` (en réalité une instance d'une classe fille). On peut obtenir plusieurs information de cet objet."
      ],
      "metadata": {
        "id": "Wsp2q8UXK4VC"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(cifar10_train) # Afficher des informations\n",
        "print(len(cifar10_train)) # Le nombre d'individus\n",
        "print(cifar10_train.data.shape) # La forme des entrées\n",
        "print(cifar10_train.targets[:20]) # Les sorties\n",
        "print(cifar10_train.classes) # Le noms des différentes classes"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "seWNaEr2k9eb",
        "outputId": "07d6063b-502d-4a23-b283-e77fb99609dd"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Dataset CIFAR10\n",
            "    Number of datapoints: 50000\n",
            "    Root location: ./cifar10_train/\n",
            "    Split: Train\n",
            "50000\n",
            "(50000, 32, 32, 3)\n",
            "[6, 9, 9, 4, 1, 1, 2, 7, 8, 3, 4, 7, 7, 2, 9, 9, 9, 3, 2, 6]\n",
            "['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Cet objet est également une liste : nous pouvons accéder aux différents éléments avec les crochets. Chaque élément est consistué de deux objets, l'entrée et la sortie d'un individu du jeu de données. Comme l'entrée est une image de type `PIL`, il nous est possinle de l'afficher avec `plt.imshow()`."
      ],
      "metadata": {
        "id": "Ar1EfQ7pLmO2"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "32 * 32 * 3"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pfokVig5zVzS",
        "outputId": "5d5fb8ba-6d02-416c-f471-973c925ff3f5"
      },
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "3072"
            ]
          },
          "metadata": {},
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "print(cifar10_train[0])\n",
        "sample_in, sample_out = cifar10_train[0]\n",
        "print(cifar10_train.classes[sample_out])\n",
        "plt.imshow(sample_in)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 482
        },
        "id": "dEYCacP2mx-8",
        "outputId": "2feb904d-2470-4b80-f354-a9bbd0b85316"
      },
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(<PIL.Image.Image image mode=RGB size=32x32 at 0x7AA9FA9DD330>, 6)\n",
            "frog\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7aa9fa9dde40>"
            ]
          },
          "metadata": {},
          "execution_count": 28
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 2.2 Les Transformations\n",
        "\n",
        "Particulièrement utiles sur les entrées complexes, telles que des images, du texte ou des données sonores, des *transformations* peuvent être construites afin d'être appliquées sur nos données. Ces transformations vous permettre de rendre lisibles ces données par un modèle statistique, mais elles permettent également d'ajouter de l'aléatoire dans la lecture des données, afin d'*augmenter* artificiellement la taille de notre échantillon (voir https://en.wikipedia.org/wiki/Data_augmentation)\n",
        "\n",
        "Plusieurs transformations sont possibles pour les images (voir https://pytorch.org/vision/stable/transforms.html) et il est possible d'enchaîner ces transformations grâce à la classe `transforms.Compose`."
      ],
      "metadata": {
        "id": "Vhdvyh-XyVWY"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "cifar10_train_transforms = transforms.Compose([\n",
        "    # Coupe l'image en une image 32x32, avec un débordement possible de 4 pixels\n",
        "    transforms.RandomCrop(32, padding=4),\n",
        "    # Peut faire un miroir horizontal sur l'image, avec une chance d'1/2\n",
        "    transforms.RandomHorizontalFlip(),\n",
        "    # Transforme l'image en tenseur\n",
        "    transforms.ToTensor(),\n",
        "    # Standardise chaque couleur avec les moyennes et deviations standards\n",
        "    # obtenues sur le jeu de données\n",
        "    transforms.Normalize(\n",
        "        mean=(0.4914, 0.4822, 0.4465),\n",
        "        std=(0.2023, 0.1994, 0.2010)\n",
        "    )\n",
        "])"
      ],
      "metadata": {
        "id": "0VilwOqJ1YRU"
      },
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "On peut maintenant affecter cette transformation lors de l'instanciation du jeu de données. A noter que cette transformation va s'appliquer à chaque fois que nous allons tirer des individus du jeu de données, i.e. en utilisant par exemple les crochets (par contre, les individus seront dans le format original dans l'attribut `cifar10_train_trs.data`)."
      ],
      "metadata": {
        "id": "9NnzzWXMOSzu"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "cifar10_train_trs = datasets.CIFAR10(root=\"./cifar10_train/\",\n",
        "                                 train=True,\n",
        "                                 download=True,\n",
        "                                 transform=cifar10_train_transforms)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "k6-SWo4X2mOa",
        "outputId": "c74c56a2-1eb6-4412-f8ef-c9cd6ae15d86"
      },
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Files already downloaded and verified\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "On effectue ici un tirage, et on regarde le résultat."
      ],
      "metadata": {
        "id": "4kwwPosaPPMo"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "sample_in, sample_out = cifar10_train_trs[0]\n",
        "plt.imshow(sample_in.permute(1, 2, 0))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 465
        },
        "id": "qEeKiGHx28tc",
        "outputId": "919d4922-861f-4360-d56b-f65197ffa7e3"
      },
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7aa9f01ca8f0>"
            ]
          },
          "metadata": {},
          "execution_count": 31
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaAAAAGdCAYAAABU0qcqAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjguMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy81sbWrAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAl30lEQVR4nO3df3DV9Z3v8VfA5AAlOTQE8mNJaIAKWn50pZJmUERI+dG9XBCc8dfcQsvoQINbSF1rulXUbieU3bXqTop3p16oe0VcegVGe0UxmnDtEigRitSalWx2A0MSCr2cA0FCTD73D27P9kiQ8w7n8MlJno+ZMyPJK+98Dt8kLw85eSfFOecEAMA1NsD3AQAA/RMFBADwggICAHhBAQEAvKCAAABeUEAAAC8oIACAFxQQAMCL63wf4NO6urp0/PhxpaenKyUlxfdxAABGzjmdOXNGeXl5GjDg8o9zel0BHT9+XPn5+b6PAQC4SkePHtWoUaMu+/qE/RNcZWWlvvCFL2jQoEEqKirSvn37Ynq79PT0RB0JAHANXenreUIK6OWXX1ZZWZnWrl2r9957T1OmTNHcuXN14sSJK74t/+wGAH3DFb+euwSYNm2aKy0tjfy5s7PT5eXluYqKiiu+bSgUcpK4cePGjVuS30Kh0Gd+vY/7I6ALFy6orq5OJSUlkZcNGDBAJSUl2rNnzyX59vZ2hcPhqBsAoO+LewGdPHlSnZ2dys7Ojnp5dna2WlpaLslXVFQoGAxGbjwBAQD6B+8/B1ReXq5QKBS5HT161PeRAADXQNyfhp2VlaWBAweqtbU16uWtra3Kycm5JB8IBBQIBOJ9DABALxf3R0BpaWmaOnWqqqqqIi/r6upSVVWViouL4/3uAABJKiE/iFpWVqalS5fqK1/5iqZNm6ann35abW1t+uY3v5mIdwcASEIJKaC77rpLv//97/XYY4+ppaVFX/7yl7Vz585LnpgAAOi/Upxzzvch/lQ4HFYwGEzY/NWptvxwQ75grG22zsUeTe2wjZ5kOEtzs2323g9t+WpDNmQbLcuT9hP5/MqqK0ei3GjMWz4j6o2zxxuyCfwQ1yvG2Zdf8HKpe4yzrU4ask3G2daPrd4kFAopIyPjsq/3/iw4AED/RAEBALyggAAAXlBAAAAvKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALzod6t4epPrEzjbsrnHuOVHx4x5C8t6Fcl2lkHG2eeNefQdA415y4Yv4zYwnTHmexNW8QAAeiUKCADgBQUEAPCCAgIAeEEBAQC8oIAAAF5QQAAALyggAIAXFBAAwAsKCADgBQUEAPDiOt8H6M9Chux04+xfJegcUmL3ZI03zraw7rzrTbvgLHvsrJsUWw3ZdOPsZN1j1pnAvPX6WD5//mCc7RuPgAAAXlBAAAAvKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlBAAAAvWMWTJBqM+SEJOcVFltUtkpRryFrXAuUbsi3G2RaWVTmSfc1PItcCXW/IzjTO/kdjvj+wfv70ZTwCAgB4QQEBALyggAAAXlBAAAAvKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXrALziPLTqhE7o+y7jGbYswXGLIHjLMnGrLWPXMWidzVZmU9y78mKJtomYbsHxJ2iuRm+dxPxMc4j4AAAF7EvYAef/xxpaSkRN0mTJgQ73cDAEhyCfknuC996Ut66623/vOdXMe/9AEAoiWkGa677jrl5OQkYjQAoI9IyPeAPvroI+Xl5WnMmDG677771NTUdNlse3u7wuFw1A0A0PfFvYCKioq0adMm7dy5Uxs2bFBjY6NuvfVWnTlzptt8RUWFgsFg5Jafb/kdlwCAZJXinHOJfAenT5/W6NGj9dRTT2n58uWXvL69vV3t7e2RP4fDYUroGrM+DXu8Md9bnoZ9+cfh3fvAmMe1xdOwr16in4YdCoWUkZFx2dcn/NkBw4YN0/XXX68jR450+/pAIKBAIJDoYwAAepmE/xzQ2bNn1dDQoNzc3ES/KwBAEol7AT300EOqqanRv//7v+tf/uVfdMcdd2jgwIG655574v2uAABJLO7/BHfs2DHdc889OnXqlEaMGKFbbrlFtbW1GjFiRLzfFeLE8n0UScoy5l815i2OJXA2erdk/b6O5XtXkjTEkLX+8Mu/GbKJWMUT9wLasmVLvEcCAPogdsEBALyggAAAXlBAAAAvKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXiT81zHAD8u+qb8wzn7XmLf4ijG/PyGnQF8zw5jfZ8xb9qRZd9gFjXmL1ATOjgWPgAAAXlBAAAAvKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlBAAAAvWMVzBQMN2c6EncJufAJnn0zg7P6yWsdlGd/A8JeeYhydrG40ZFuMs62XJ9+Qta6/sawFss7+c0P2gCHbJen3MeR4BAQA8IICAgB4QQEBALyggAAAXlBAAAAvKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALzod7vgLPujJOmDhJziokGG7BDj7HOG7Hbj7N8Y8/3BL61vkMiFev1E2JC17GqTpH815jsMWcvnpiSdN2St5242ZC378bpizPEICADgBQUEAPCCAgIAeEEBAQC8oIAAAF5QQAAALyggAIAXFBAAwAsKCADgBQUEAPCCAgIAeNHvdsF9YsynG7KWXUmSlGrIWi+UZTau3nPG/C5jfrgx3x8cM2Stnw/XG/NNhqxlb1yiTTdkFxqyH0sqiyHHIyAAgBfmAtq9e7cWLFigvLw8paSkaPv27VGvd87pscceU25urgYPHqySkhJ99NFH8TovAKCPMBdQW1ubpkyZosrKym5fv379ej377LN67rnntHfvXn3uc5/T3Llzdf68Zak4AKCvM38PaP78+Zo/f363r3PO6emnn9YPfvADLVx48V8MX3jhBWVnZ2v79u26++67r+60AIA+I67fA2psbFRLS4tKSkoiLwsGgyoqKtKePXu6fZv29naFw+GoGwCg74trAbW0tEiSsrOzo16enZ0ded2nVVRUKBgMRm75+dbfXQgASEbenwVXXl6uUCgUuR09etT3kQAA10BcCygnJ0eS1NraGvXy1tbWyOs+LRAIKCMjI+oGAOj74lpAhYWFysnJUVVVVeRl4XBYe/fuVXFxcTzfFQAgyZmfBXf27FkdOXIk8ufGxkYdPHhQmZmZKigo0OrVq/U3f/M3+uIXv6jCwkI9+uijysvL06JFi+J5bgBAkjMX0P79+3X77bdH/lxWdnHhwtKlS7Vp0yY9/PDDamtr0wMPPKDTp0/rlltu0c6dOzVo0KD4nfoqWO/wEEPWui7FMrv7p3DEJ9+b1rwsNub/qSD27BLLvhTZ1quEbKP1pjE/x5CdbZx9zpDt/rmsvV+jMb/AmLd8Xck1zrZcn1PG2WON+XgzF9DMmTPlnLvs61NSUvTkk0/qySefvKqDAQD6Nu/PggMA9E8UEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlBAAAAvKCAAgBcUEADAC/MqnmT3iTFv2deWyN/l2mHMW36tX9A4+zfGvMX/+q7xDVZOijn6+tb3TaN3lMeeXWiaLK0x5mcastbdfncastbf1vV9Q/Ym4+x/NOYt9hnzrVeORHxgnJ1pyFr2xknSu4as5WvQhRhzPAICAHhBAQEAvKCAAABeUEAAAC8oIACAFxQQAMALCggA4AUFBADwggICAHhBAQEAvEhxzjnfh/hT4XBYwaB1OUzsrjfmMwxZ67ocS9665scyO9U4+5gxn23I/nfj7IV/ZwiPNQ7/pSG7Mss2+1zIlh9iuKLbbaO1ckLs2Z9/aJttWGekCtvo2wyzd9tGJ61SY97yUXjSkP1E0luSQqGQMjIu/1WUR0AAAC8oIACAFxQQAMALCggA4AUFBADwggICAHhBAQEAvKCAAABeUEAAAC8oIACAFxQQAMCLfrcLboYxf86QLTDO3mfINhtnjzdkc42zHzTmZ86JPRt81Djcst9tiHF24j4MZf9oMexr0wrj7Dtij25PsY3eYMi+YRv9K8NRbrGNTlqFxvw0Q9ayX7JD0qtiFxwAoJeigAAAXlBAAAAvKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlzn+wDXWr0xb1nFEzbOtrBuhTlqyFpX6yx0vzG+xWRD9oJxdoMh+5Jxdo4hu9Q427oXyLgCJ1GyLLuPpG+9Gfv1+R+P2Y4y/ZnYs5nfsc3+gy3eaww35rMScorYP4t5BAQA8IICAgB4YS6g3bt3a8GCBcrLy1NKSoq2b98e9fply5YpJSUl6jZv3rx4nRcA0EeYC6itrU1TpkxRZWXlZTPz5s1Tc3Nz5PbSS9Z/ewcA9HXmJyHMnz9f8+fP/8xMIBBQTo7lG7gAgP4mId8Dqq6u1siRIzV+/HitXLlSp06dumy2vb1d4XA46gYA6PviXkDz5s3TCy+8oKqqKv34xz9WTU2N5s+fr87Ozm7zFRUVCgaDkVt+fn68jwQA6IXi/nNAd999d+S/J02apMmTJ2vs2LGqrq7W7NmzL8mXl5errKws8udwOEwJAUA/kPCnYY8ZM0ZZWVk6cuRIt68PBALKyMiIugEA+r6EF9CxY8d06tQp5ebmJvpdAQCSiPmf4M6ePRv1aKaxsVEHDx5UZmamMjMz9cQTT2jJkiXKyclRQ0ODHn74YY0bN05z586N68EBAMnNXED79+/X7bffHvnzH79/s3TpUm3YsEGHDh3Sz3/+c50+fVp5eXmaM2eOfvjDHyoQCMTv1Feh1ZgfaMj+q3G2heUcktT9Uz66N2SScbhpt5tVmjFv2ZL3pHF2P/Hht2PPNll270l3GrK3/dA0WjXOxZw91WTbpZfy97az9BaJfA6xZS9mrLvgzAU0c+ZMuc+48G+88YZ1JACgH2IXHADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlBAAAAvKCAAgBcUEADACwoIAOBF3H8fUF9j2an2FePsLEM2WGCbPcSwuCn3JtvsxKo1pf96bnHM2R+9UWU8yyxjPklNWBF79v0NptFfN3xs/eI902hJv4s9+nex742TpDf/3rY7bo4pnTgdxrzla5BlC2Cs5+AREADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlBAAAAvKCAAgBcUEADACwoIAOBFinPOtqMiwcLhsILBYMLmpxvzXzNkraduMmRTjbMtm3sWGlfxfL3uhO0NNCL2aINtBUqDYX3L2Dl/aZqtdz6MPWvdZ1RUYcv3GheM+cUxJ8+l/NI2+pHYPyuGVFjP/UNT+r+lPBZz9n8aT2IxxZi/3ZA1fDboE0lvSQqFQsrIyLhsjkdAAAAvKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlBAAAAvKCAAgBcUEADAi363C25tri3/ZnPs2QO20abdccONs4cYspWWsKRpbYn8kHnbmN8Xe3RruWny8/fFnq3vMI1WyBbX8u/Gvvds2pxFtuGTimLP5n7XNttisnEPoOHjdmxtgr/M/Sz2j62U+9cl7BgDjfnOhJziP7ELDgDQK1FAAAAvKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlBAAAAv+t0qnnnG/M6EnKJ32WDMr+hVHzJ/H3Py8WEPmSa/YNiX809ZptHaddKW/7EhO942Wr8xZN3xlbbhubF/Ln+YYltRk3W7Ift27/mYbfqObeXQf3029mzYeJbYFzxJhq1kcpLOilU8AIBeylRAFRUVuvnmm5Wenq6RI0dq0aJFqq+vj8qcP39epaWlGj58uIYOHaolS5aotbU1rocGACQ/UwHV1NSotLRUtbW12rVrlzo6OjRnzhy1tbVFMmvWrNGrr76qrVu3qqamRsePH9fixYvjfnAAQHK7zhLeuTP6OyKbNm3SyJEjVVdXpxkzZigUCun555/X5s2bNWvWLEnSxo0bdcMNN6i2tlZf/epX43dyAEBSu6rvAYVCF79Lm5mZKUmqq6tTR0eHSkpKIpkJEyaooKBAe/bs6XZGe3u7wuFw1A0A0Pf1uIC6urq0evVqTZ8+XRMnTpQktbS0KC0tTcOGDYvKZmdnq6Wlpds5FRUVCgaDkVt+fn5PjwQASCI9LqDS0lIdPnxYW7ZsuaoDlJeXKxQKRW5Hjx69qnkAgORg+h7QH61atUqvvfaadu/erVGjRkVenpOTowsXLuj06dNRj4JaW1uVk5PT7axAIKBAINCTYwAAkpjpEZBzTqtWrdK2bdv09ttvq7CwMOr1U6dOVWpqqqqqqiIvq6+vV1NTk4qLi+NzYgBAn2B6BFRaWqrNmzdrx44dSk9Pj3xfJxgMavDgwQoGg1q+fLnKysqUmZmpjIwMPfjggyouLuYZcACAKKYC2rDh4tKWmTNnRr1848aNWrZsmSTpJz/5iQYMGKAlS5aovb1dc+fO1U9/+tO4HBYA0HeYCiiWtXGDBg1SZWWlKisre3yoRDKu4OoXTloWQkmSNtniIcM7CBoWfEk6uS72/W7vGXa7SbaPlZeMH1h/PcGWP/ehIWsbrX+zhNcZNwcaLv0O22Td3hB7NqvjDdvw1Lm2vEHBMydM+dxnR8ac7f65xpfXYchavkzEunmPXXAAAC8oIACAFxQQAMALCggA4AUFBADwggICAHhBAQEAvKCAAABeUEAAAC8oIACAFz36dQzJ7Nf/x5a/7dbYs4dto/UHYz5Rdln2cUjal/LNxBxE0lhj/l1D1ng3NdOQbTbOfsGwWkeSMgzZJttoPWwJG1cIWf5ibjGOzsk1hLe+YBt+k2230r4fro45+9Jm2+yZhuxO02Qp3ZANGrJdkv5vDDkeAQEAvKCAAABeUEAAAC8oIACAFxQQAMALCggA4AUFBADwggICAHhBAQEAvKCAAABeUEAAAC9SnHPO9yH+VDgcVjBo2TpkY723a2bFnn36Hdts3WTIDjHOft+QDdlGZ9ri+poha72b1v1uFpZVY8ONs08Z85ZLZP07KTBk7zTugpv4pCEctJxE0vvnYo6e/KVt/5r14/A2w+d+vXF2mSFr/RKUaszH6hNJNZJCoZAyMi6/yZBHQAAALyggAIAXFBAAwAsKCADgBQUEAPCCAgIAeEEBAQC8oIAAAF5QQAAALyggAIAX1/k+QG837XZD2LoH4z1D9l7b6MKxsWcbf26b/QdbXC8b8xbzDNnpxtmWhVDW1TrvGvOGy6lm4+wsQ3bi7ypsw5sMO6EKjKt4fr4u5uhR4+fmYFtcf2HIFhlnW1YrPWicvc+Q3WvIfhJjjkdAAAAvKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlBAAAAvKCAAgBcUEADAi363C+6pybZ82YuxZ194zDZ7pyW82TY71bBwaqBttDqN+USy/B3ONM5ONWQta/0k2w4uSZpkyL610jh85VJDeIxtdsEjsWeblplG79sae3baX5pG69ybtvy0D2PPWnf1nTRkDxhnn0tQNtavETwCAgB4YSqgiooK3XzzzUpPT9fIkSO1aNEi1dfXR2VmzpyplJSUqNuKFSviemgAQPIzFVBNTY1KS0tVW1urXbt2qaOjQ3PmzFFbW1tU7v7771dzc3Pktn79+rgeGgCQ/EzfA9q5M/pf3Ddt2qSRI0eqrq5OM2bMiLx8yJAhysnJic8JAQB90lV9DygUCkmSMjMzo17+4osvKisrSxMnTlR5ebnOnbv8t6/a29sVDoejbgCAvq/Hz4Lr6urS6tWrNX36dE2cODHy8nvvvVejR49WXl6eDh06pO9973uqr6/XK6+80u2ciooKPfHEEz09BgAgSfW4gEpLS3X48GG9+270Lxd+4IEHIv89adIk5ebmavbs2WpoaNDYsZf+YuHy8nKVlZVF/hwOh5Wfn9/TYwEAkkSPCmjVqlV67bXXtHv3bo0aNeozs0VFF38g5ciRI90WUCAQUCAQ6MkxAABJzFRAzjk9+OCD2rZtm6qrq1VYWHjFtzl48KAkKTc3t0cHBAD0TaYCKi0t1ebNm7Vjxw6lp6erpaVFkhQMBjV48GA1NDRo8+bN+vrXv67hw4fr0KFDWrNmjWbMmKHJk40rCAAAfZqpgDZs2CDp4g+b/qmNGzdq2bJlSktL01tvvaWnn35abW1tys/P15IlS/SDH/wgbgcGAPQNKc455/sQfyocDisYDCZs/leM+V9b/nZ+aZv95f8Se/Y3ttEaZMhOvHIkivXqZBmyhpVakqT6K0ciLLvdJNu5LXuyJMn6U3JzDNlvXPqt1s/0XkPsWev9XPFk7Nm/Nu5SXGjITrvdNnvfO7Z8yJC17g3cZcha9sZJts+JDkO2U9JhXfxRnYyMjMvm2AUHAPCCAgIAeEEBAQC8oIAAAF5QQAAALyggAIAXFBAAwAsKCADgBQUEAPCCAgIAeNHvVvFYuU2G8FLb7KYbYs/ONO6osaxMmWkbrWnGvIVxi4yaDNkdxtmWtSZ3Gmd/Ysw/ZcgOMc6eaciWXTkSxbLR5nHj7PGGrHETj3lVkuX3OFvXTVnyHxhnpxuylmvfLmmdWMUDAOilKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlBAAAAvKCAAgBcUEADAC3bBxZG7YHwDwyKzu8clbLRusY0260hQVpLeNWRzjbMnGbLDjbN3GfPVhmyqcfafG7LWXX2Ws1j/To4ZssXG2daPFcv+PcvnpiT9ypC1nnu5IWvZd3hWF//O2QUHAOiVKCAAgBcUEADACwoIAOAFBQQA8IICAgB4QQEBALyggAAAXlBAAAAvKCAAgBfX+T5AX3LyO7Z8VlHs2ZdtozXIkL3JOPucMX/SkP3fxtmdhuwc4+wcQ/aUcbZ15ZBlpY119lFDNss427K6x/L3LUnNhqxlVY4kTTDmL79s5lLWz5+FhuzXrLMXxZ7NNeymCrdLWn/lHI+AAABeUEAAAC8oIACAFxQQAMALCggA4AUFBADwggICAHhBAQEAvKCAAABeUEAAAC8oIACAF+yCi6Pqvbb8UxsScw5JOm/IvmucbdnvJUnVhqxlt5skzTBk842zWwzZwcbZ1v17lv1uidwzZ90FZ9kDaFiNKEkKGrKHjbMt55akWwzZO42zpxuGF1j/Ei15y8VvE7vgAAC9l6mANmzYoMmTJysjI0MZGRkqLi7W66+/Hnn9+fPnVVpaquHDh2vo0KFasmSJWltb435oAEDyMxXQqFGjtG7dOtXV1Wn//v2aNWuWFi5cqN/+9reSpDVr1ujVV1/V1q1bVVNTo+PHj2vx4sUJOTgAILmZvge0YMGCqD//6Ec/0oYNG1RbW6tRo0bp+eef1+bNmzVr1ixJ0saNG3XDDTeotrZWX/3qV+N3agBA0uvx94A6Ozu1ZcsWtbW1qbi4WHV1dero6FBJSUkkM2HCBBUUFGjPnj2XndPe3q5wOBx1AwD0feYCev/99zV06FAFAgGtWLFC27Zt04033qiWlhalpaVp2LBhUfns7Gy1tFz++UQVFRUKBoORW36+9blKAIBkZC6g8ePH6+DBg9q7d69WrlyppUuX6oMPPujxAcrLyxUKhSK3o0ctvyAYAJCszD8HlJaWpnHjxkmSpk6dql//+td65plndNddd+nChQs6ffp01KOg1tZW5eRc/re9BwIBBQIB+8kBAEntqn8OqKurS+3t7Zo6dapSU1NVVVUVeV19fb2amppUXFx8te8GANDHmB4BlZeXa/78+SooKNCZM2e0efNmVVdX64033lAwGNTy5ctVVlamzMxMZWRk6MEHH1RxcTHPgAMAXMJUQCdOnNA3vvENNTc3KxgMavLkyXrjjTf0ta99TZL0k5/8RAMGDNCSJUvU3t6uuXPn6qc//WlCDt4b3fkXtvya9xJzDqvfJDifSCFD1rqOxTK72Tjbsv5GkgoMWcu5JdvqHuuKGgvr34llFc9E42zrupw758SezVpkHG5Zl9NknG35wH3HkL0QW8xUQM8///xnvn7QoEGqrKxUZWWlZSwAoB9iFxwAwAsKCADgBQUEAPCCAgIAeEEBAQC8oIAAAF5QQAAALyggAIAXFBAAwAvzNuxEc875PkKPhdtt+a7EHKNf6TRkY9wOEmFZUfOJcXaKMZ/Is1jylnNY9abr87Exf8bwDtKsw88asueMsy1nMVyg8P/PXunreYrrZV/xjx07xi+lA4A+4OjRoxo1atRlX9/rCqirq0vHjx9Xenq6UlL+8/8Tw+Gw8vPzdfToUWVkZHg8YWJxP/uO/nAfJe5nXxOP++mc05kzZ5SXl6cBAy7/nZ5e909wAwYM+MzGzMjI6NMX/4+4n31Hf7iPEvezr7na+xkMXnlfOU9CAAB4QQEBALxImgIKBAJau3atAoGA76MkFPez7+gP91HifvY11/J+9ronIQAA+oekeQQEAOhbKCAAgBcUEADACwoIAOBF0hRQZWWlvvCFL2jQoEEqKirSvn37fB8prh5//HGlpKRE3SZMmOD7WFdl9+7dWrBggfLy8pSSkqLt27dHvd45p8cee0y5ubkaPHiwSkpK9NFHH/k57FW40v1ctmzZJdd23rx5fg7bQxUVFbr55puVnp6ukSNHatGiRaqvr4/KnD9/XqWlpRo+fLiGDh2qJUuWqLW11dOJeyaW+zlz5sxLrueKFSs8nbhnNmzYoMmTJ0d+2LS4uFivv/565PXX6lomRQG9/PLLKisr09q1a/Xee+9pypQpmjt3rk6cOOH7aHH1pS99Sc3NzZHbu+++6/tIV6WtrU1TpkxRZWVlt69fv369nn32WT333HPau3evPve5z2nu3Lk6f/78NT7p1bnS/ZSkefPmRV3bl1566Rqe8OrV1NSotLRUtbW12rVrlzo6OjRnzhy1tbVFMmvWrNGrr76qrVu3qqamRsePH9fixYs9ntoulvspSffff3/U9Vy/fr2nE/fMqFGjtG7dOtXV1Wn//v2aNWuWFi5cqN/+9reSruG1dElg2rRprrS0NPLnzs5Ol5eX5yoqKjyeKr7Wrl3rpkyZ4vsYCSPJbdu2LfLnrq4ul5OT4/72b/828rLTp0+7QCDgXnrpJQ8njI9P30/nnFu6dKlbuHChl/MkyokTJ5wkV1NT45y7eO1SU1Pd1q1bI5nf/e53TpLbs2ePr2NetU/fT+ecu+2229x3vvMdf4dKkM9//vPuZz/72TW9lr3+EdCFCxdUV1enkpKSyMsGDBigkpIS7dmzx+PJ4u+jjz5SXl6exowZo/vuu09NTU2+j5QwjY2NamlpibquwWBQRUVFfe66SlJ1dbVGjhyp8ePHa+XKlTp16pTvI12VUCgkScrMzJQk1dXVqaOjI+p6TpgwQQUFBUl9PT99P//oxRdfVFZWliZOnKjy8nKdO2f9PQi9R2dnp7Zs2aK2tjYVFxdf02vZ65aRftrJkyfV2dmp7OzsqJdnZ2frww8/9HSq+CsqKtKmTZs0fvx4NTc364knntCtt96qw4cPKz093ffx4q6lpUWSur2uf3xdXzFv3jwtXrxYhYWFamho0Pe//33Nnz9fe/bs0cCBA30fz6yrq0urV6/W9OnTNXHiREkXr2daWpqGDRsWlU3m69nd/ZSke++9V6NHj1ZeXp4OHTqk733ve6qvr9crr7zi8bR277//voqLi3X+/HkNHTpU27Zt04033qiDBw9es2vZ6wuov5g/f37kvydPnqyioiKNHj1a//zP/6zly5d7PBmu1t133x3570mTJmny5MkaO3asqqurNXv2bI8n65nS0lIdPnw46b9HeSWXu58PPPBA5L8nTZqk3NxczZ49Ww0NDRo7duy1PmaPjR8/XgcPHlQoFNIvfvELLV26VDU1Ndf0DL3+n+CysrI0cODAS56B0draqpycHE+nSrxhw4bp+uuv15EjR3wfJSH+eO3623WVpDFjxigrKyspr+2qVav02muv6Z133on6tSk5OTm6cOGCTp8+HZVP1ut5ufvZnaKiIklKuuuZlpamcePGaerUqaqoqNCUKVP0zDPPXNNr2esLKC0tTVOnTlVVVVXkZV1dXaqqqlJxcbHHkyXW2bNn1dDQoNzcXN9HSYjCwkLl5OREXddwOKy9e/f26esqXfytv6dOnUqqa+uc06pVq7Rt2za9/fbbKiwsjHr91KlTlZqaGnU96+vr1dTUlFTX80r3szsHDx6UpKS6nt3p6upSe3v7tb2WcX1KQ4Js2bLFBQIBt2nTJvfBBx+4Bx54wA0bNsy1tLT4PlrcfPe733XV1dWusbHR/epXv3IlJSUuKyvLnThxwvfReuzMmTPuwIED7sCBA06Se+qpp9yBAwfcf/zHfzjnnFu3bp0bNmyY27Fjhzt06JBbuHChKywsdB9//LHnk9t81v08c+aMe+ihh9yePXtcY2Oje+utt9xNN93kvvjFL7rz58/7PnrMVq5c6YLBoKuurnbNzc2R27lz5yKZFStWuIKCAvf222+7/fv3u+LiYldcXOzx1HZXup9HjhxxTz75pNu/f79rbGx0O3bscGPGjHEzZszwfHKbRx55xNXU1LjGxkZ36NAh98gjj7iUlBT35ptvOueu3bVMigJyzrl/+Id/cAUFBS4tLc1NmzbN1dbW+j5SXN11110uNzfXpaWluT/7sz9zd911lzty5IjvY12Vd955x0m65LZ06VLn3MWnYj/66KMuOzvbBQIBN3v2bFdfX+/30D3wWffz3Llzbs6cOW7EiBEuNTXVjR492t1///1J9z9P3d0/SW7jxo2RzMcff+y+/e1vu89//vNuyJAh7o477nDNzc3+Dt0DV7qfTU1NbsaMGS4zM9MFAgE3btw491d/9VcuFAr5PbjRt771LTd69GiXlpbmRowY4WbPnh0pH+eu3bXk1zEAALzo9d8DAgD0TRQQAMALCggA4AUFBADwggICAHhBAQEAvKCAAABeUEAAAC8oIACAFxQQAMALCggA4AUFBADw4v8BRecSNCxPCt8AAAAASUVORK5CYII=\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Remarquez que, grâce à l'aléatoire, le même individu tiré une nouvelle fois peut subir d'autres transformations."
      ],
      "metadata": {
        "id": "er05v3k6PY9h"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "sample_in, sample_out = cifar10_train_trs[0]\n",
        "plt.imshow(sample_in.permute(1, 2, 0))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 465
        },
        "id": "kzn-lMnu4hTV",
        "outputId": "7856bee5-9636-4694-d155-aa9ce9c07551"
      },
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7aa9f01055a0>"
            ]
          },
          "metadata": {},
          "execution_count": 32
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 2.3 La classe `Dataloader`\n",
        "\n",
        "Généralement, on n'utilise pas l'objet `Dataset` directement, mais on va plutôt l'affecter à un objet de classe `Dataloader`, qui est un **conteneur itérable** sur nos données.\n",
        "\n",
        "Ce conteneur itérable permet de tirer les données par petit sous-échantillons, appelés `batch`, mais permet également de définir certaine règles sur le tirage des individus.\n",
        "\n",
        "On utilise l'instanciation suivante pour créer un conteneur itérable sur notre objet `cifar10_train_trs`, avec batch de 8 (batch_size=8) et un brassage des individus lorsque le jeu de données est entièrement tiré (`shuffle=True`)."
      ],
      "metadata": {
        "id": "PkUQokMW6fna"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "cifar10_train_loader = DataLoader(cifar10_train_trs, batch_size=8, shuffle=True)"
      ],
      "metadata": {
        "id": "rNS_89Tg6qNP"
      },
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Comme tout conteneur itérable, on peut générer un iterateur avec la fonction `iter()`, puis piocher des individus dans celui-ci avec la fonction `next()`."
      ],
      "metadata": {
        "id": "PWWsOwrRQwE4"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "data_batch, labels_batch = next(iter(cifar10_train_loader))"
      ],
      "metadata": {
        "id": "f8_ZIbMYI1ia"
      },
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Cette itération nous a donné 2 séries d'entrées et sorties correspondantes"
      ],
      "metadata": {
        "id": "SvQe4DPtR_Zv"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(data_batch.shape)\n",
        "print(labels_batch.shape)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JFuLyE33JENM",
        "outputId": "55fd9b65-8a51-405c-c4d7-478f69af3a50"
      },
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([8, 3, 32, 32])\n",
            "torch.Size([8])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Pour visuliser nos 8 entrées et sorties."
      ],
      "metadata": {
        "id": "-X7T6blQSI5s"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "cols, rows = 3, 3\n",
        "figure = plt.figure(figsize=(8, 8))\n",
        "for i, data in enumerate(data_batch):\n",
        "    figure.add_subplot(rows, cols, i + 1)\n",
        "    plt.title(cifar10_train_trs.classes[labels_batch[i]])\n",
        "    plt.axis(\"off\")\n",
        "    plt.imshow(data.permute(1, 2, 0))\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 814
        },
        "id": "67chKMqEJMZU",
        "outputId": "0f2a26a8-902c-40a9-ad0b-48dfc4bc9b41"
      },
      "execution_count": 36,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n",
            "WARNING:matplotlib.image:Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 800x800 with 8 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "En réalité, l'utilisation des objets `Dataloader` se fait à l'aide d'une boucle `for`, où l'on va alors itérer sur tous les objets du jeu de données. (ici, on se limite à 5 itérations, pour que cela reste lisible)."
      ],
      "metadata": {
        "id": "hzXI4FrlSjy6"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Pour limiter la boucle\n",
        "counter = 0\n",
        "# On pourra entrainer dans des boucles de la sorte\n",
        "for items, labels in cifar10_train_loader:\n",
        "  print(items.shape)\n",
        "  batch_named_labels = [cifar10_train_trs.classes[label] for label in labels]\n",
        "  print(batch_named_labels)\n",
        "  counter += 1\n",
        "  if counter == 5:\n",
        "    break"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oFYE9TiEKl58",
        "outputId": "6ca765e4-a8e6-4269-b356-0e1e1bce83e4"
      },
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([8, 3, 32, 32])\n",
            "['horse', 'horse', 'horse', 'frog', 'horse', 'horse', 'cat', 'automobile']\n",
            "torch.Size([8, 3, 32, 32])\n",
            "['bird', 'dog', 'dog', 'truck', 'truck', 'ship', 'dog', 'dog']\n",
            "torch.Size([8, 3, 32, 32])\n",
            "['truck', 'bird', 'dog', 'ship', 'frog', 'cat', 'horse', 'cat']\n",
            "torch.Size([8, 3, 32, 32])\n",
            "['truck', 'frog', 'horse', 'deer', 'bird', 'airplane', 'horse', 'cat']\n",
            "torch.Size([8, 3, 32, 32])\n",
            "['horse', 'automobile', 'dog', 'frog', 'horse', 'frog', 'cat', 'cat']\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 2.4 Une classe `Dataset` personalisée\n",
        "\n",
        "Généralement, on va créer une classe personalisée héritée de la classe `Dataset`, afin de pouvoir utiliser plusieurs fonctions proposées par PyTorch (et `DataLoader`).\n",
        "\n",
        "Il y a trois méthodes à implémenter :\n",
        "\n",
        "* `__init__()` : Le constructeur, qui va mettre mémoire plusieurs éléments du jeu de données, comme les chemins d'accès, les transformations potentielles, et même parfois des éléments du jeu de données.\n",
        "* `__len__(self)` : La méthode qui retourne le nombre d'individus dans le jeu de données. Cette méthode est appelée lorsque l'on utilise `len()`.\n",
        "* `__getitem__(self, idx)` : Cette méthode est appelée lors de l'utilisation des crochets ou du `Dataloader()` et retourne l'élément à la position `idx`. C'est ici qu'on applique les transformations et que l'on précise où aller chercher les données si elles sont dans un endroit distant.\n",
        "\n",
        "Dans notre exemple (on crée une classe pour charger \"contraceptive.tsv\"), nous effectuons déjà plusieurs pré-traitements sur les données lors de la construction de l'objet, car les transformations désirées sont invariantes et particulièrement simples. De plus, tout le jeu de données est stocké directement dans les attributs de l'objet, car il est suffisamment petit.\n",
        "\n"
      ],
      "metadata": {
        "id": "V8L4kAXtpYha"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Classe qui hérite de Dataset\n",
        "class CustomDataset(Dataset):\n",
        "\n",
        "  # Notre constructeur\n",
        "  def __init__(self, file_path, transform=None, target_transform=None):\n",
        "    # On charge le jeu de données\n",
        "    all_data = pd.read_csv(file_path, sep=\"\\t\")\n",
        "    # On constitue les entrées\n",
        "    occ_dummies = pd.get_dummies(all_data[\"Husband_occupation\"], prefix=\"occ\")*1\n",
        "    data = pd.concat([all_data.drop([\"Husband_occupation\", \"target\"],\n",
        "                                    axis=1), occ_dummies], axis=1)\n",
        "    self.data = torch.tensor(data.to_numpy(), dtype=torch.float32)\n",
        "    # On constitue les sorties\n",
        "    self.targets = torch.tensor((all_data[\"target\"] - 1).to_numpy())\n",
        "    # On sauve le noms des variables\n",
        "    self.var_names = data.columns.tolist()\n",
        "    # On constitue les classes\n",
        "    self.classes = [\"none\", \"long-term\", \"short_term\"]\n",
        "    # On sauve les transformations\n",
        "    self.transform = transform\n",
        "    self.target_transform = target_transform\n",
        "\n",
        "  # Pour la longueur\n",
        "  def __len__(self):\n",
        "    return len(self.targets)\n",
        "\n",
        "  # Pour le tirage des individus\n",
        "  def __getitem__(self, idx):\n",
        "    # On prend l'entrée et la sortie correspondant à l'indice\n",
        "    item = self.data[idx, :]\n",
        "    target = self.targets[idx]\n",
        "    # On applique les éventuelles transformations\n",
        "    if self.transform:\n",
        "      item = self.transform(item)\n",
        "    if self.target_transform:\n",
        "      target = self.transform(target)\n",
        "    return item, target\n"
      ],
      "metadata": {
        "id": "nnk4DojupwTT"
      },
      "execution_count": 38,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "On peut maintenant créer une instance de cette classe avec le chemin d'accès à nos données. On peut voir que les crochets nous donnent bien un individu de notre jeu de données."
      ],
      "metadata": {
        "id": "VIDL1_N1WatN"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "file_path = \"/content/drive/MyDrive/Colab Notebooks/ml_data/TP8/contraceptive.tsv\"\n",
        "my_data = CustomDataset(file_path)\n",
        "my_data[0]"
      ],
      "metadata": {
        "id": "2oK4RJg7tFqg",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "dbe0bde8-a236-4b04-f9a4-fc922066499d"
      },
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(tensor([24.,  2.,  3.,  3.,  1.,  1.,  3.,  0.,  0.,  1.,  0.,  0.]),\n",
              " tensor(0))"
            ]
          },
          "metadata": {},
          "execution_count": 39
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "On peut maintenant poser un `DataLoader` sur ce jeu de données, et tirer des exemples de ce dernier."
      ],
      "metadata": {
        "id": "-8dwHU9IWzZM"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "my_dataloader = DataLoader(my_data, batch_size=4, shuffle=True)\n",
        "\n",
        "counter = 0\n",
        "for items, labels in my_dataloader:\n",
        "  print(items)\n",
        "  print(labels)\n",
        "  counter += 1\n",
        "  if counter == 5:\n",
        "    break"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uF_kXxl7y7pK",
        "outputId": "3d7557db-e081-4f0c-d92a-b248bde1b605"
      },
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[30.,  4.,  2.,  0.,  1.,  0.,  3.,  0.,  0.,  0.,  1.,  0.],\n",
            "        [36.,  4.,  4.,  2.,  1.,  1.,  3.,  0.,  0.,  0.,  1.,  0.],\n",
            "        [30.,  4.,  4.,  1.,  0.,  1.,  4.,  0.,  0.,  1.,  0.,  0.],\n",
            "        [22.,  4.,  4.,  1.,  1.,  1.,  3.,  0.,  0.,  1.,  0.,  0.]])\n",
            "tensor([0, 0, 1, 2])\n",
            "tensor([[44.,  4.,  4.,  4.,  1.,  0.,  3.,  0.,  0.,  1.,  0.,  0.],\n",
            "        [17.,  3.,  4.,  1.,  1.,  1.,  3.,  0.,  0.,  0.,  1.,  0.],\n",
            "        [44.,  4.,  4., 11.,  1.,  1.,  4.,  0.,  1.,  0.,  0.,  0.],\n",
            "        [38.,  3.,  3.,  4.,  0.,  0.,  2.,  1.,  0.,  0.,  1.,  0.]])\n",
            "tensor([1, 2, 2, 0])\n",
            "tensor([[29.,  4.,  2.,  0.,  1.,  0.,  2.,  0.,  0.,  1.,  0.,  0.],\n",
            "        [42.,  3.,  3.,  8.,  1.,  1.,  3.,  0.,  0.,  0.,  0.,  1.],\n",
            "        [45.,  4.,  4.,  1.,  1.,  1.,  4.,  0.,  1.,  0.,  0.,  0.],\n",
            "        [20.,  3.,  3.,  2.,  1.,  1.,  2.,  1.,  0.,  1.,  0.,  0.]])\n",
            "tensor([0, 0, 0, 2])\n",
            "tensor([[33.,  3.,  4.,  6.,  1.,  1.,  2.,  0.,  0.,  0.,  1.,  0.],\n",
            "        [32.,  3.,  3.,  4.,  1.,  0.,  3.,  0.,  0.,  0.,  1.,  0.],\n",
            "        [49.,  2.,  2.,  8.,  1.,  0.,  1.,  0.,  0.,  0.,  1.,  0.],\n",
            "        [28.,  1.,  4.,  0.,  1.,  0.,  1.,  0.,  0.,  0.,  0.,  1.]])\n",
            "tensor([2, 1, 0, 0])\n",
            "tensor([[21.,  3.,  3.,  1.,  1.,  1.,  1.,  0.,  0.,  0.,  1.,  0.],\n",
            "        [29.,  2.,  4.,  3.,  1.,  0.,  3.,  0.,  0.,  0.,  1.,  0.],\n",
            "        [30.,  3.,  4.,  2.,  1.,  1.,  4.,  0.,  1.,  0.,  0.,  0.],\n",
            "        [47.,  2.,  2.,  7.,  1.,  1.,  4.,  0.,  0.,  1.,  0.,  0.]])\n",
            "tensor([2, 2, 1, 1])\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Par la suite, on va utiliser ce jeu de données pour entrainer un modèle. On va donc séparer notre jeu de données en trois parties : entrainement (70%), validation (15%) et test (15%). On peut utiliser pour cela la fonction `random_split()` sur notre objet directement."
      ],
      "metadata": {
        "id": "IVwaBJ4eXMZr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "train_size = int(0.7 * len(my_data))\n",
        "valid_size = int(0.15 * len(my_data))\n",
        "test_size = len(my_data) - train_size - valid_size\n",
        "my_data_train, my_data_valid, my_data_test = \\\n",
        "  random_split(my_data, [train_size, valid_size, test_size])"
      ],
      "metadata": {
        "id": "9Dd4r3N42IWX"
      },
      "execution_count": 41,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Les objets résultants sont toujours des instances de la classes `Dataset`"
      ],
      "metadata": {
        "id": "WEPZnJOBXqqu"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(len(my_data_train), len(my_data_valid), len(my_data_test))\n",
        "my_data_train[0]"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-WHj0zyh4_Sx",
        "outputId": "0b641129-df57-4fa5-c575-903bca5fc410"
      },
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1031 220 222\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(tensor([26.,  4.,  4.,  3.,  1.,  1.,  4.,  0.,  0.,  0.,  1.,  0.]),\n",
              " tensor(2))"
            ]
          },
          "metadata": {},
          "execution_count": 42
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "On peut donc créer trois itérateurs sur ces jeux de données, que nous utiliserons plus tard."
      ],
      "metadata": {
        "id": "VxDCGzKlXzkr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "train_loader = DataLoader(my_data_train, batch_size=8, shuffle=True)\n",
        "valid_loader =  DataLoader(my_data_valid, batch_size=8, shuffle=True)\n",
        "test_loader =  DataLoader(my_data_test, batch_size=8, shuffle=True)"
      ],
      "metadata": {
        "id": "42sjhJfN-rRW"
      },
      "execution_count": 43,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "RdeOOu2_NRJS"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## 3. Les Modèles\n",
        "\n",
        "Dans cette partie, nous allons enfin commencer à utiliser des modèles. Nous allons voir comment utiliser des modèles pré-entrainés, comment construire notre propre modèle, comment l'entrainer, l'évaluer puis faire des prédictions avec ce dernier. Pour finir, nous allons voir comment sauvegarder un modèle."
      ],
      "metadata": {
        "id": "vSXzU6XsNSPg"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 3.1 Les Modèles Pré-entrainés\n",
        "\n",
        "Il existe de nombreux modèles pré-entrainés dans PyTorch (pour les images : https://pytorch.org/vision/stable/index.html; pour le texte : https://pytorch.org/text/stable/models.html; pour l'audio : https://pytorch.org/audio/stable/models.html). Ils peuvent se charger pré-entrainés, ou juste pour avoir la structure de ces derniers.\n",
        "\n",
        "Ces modèles pré-entrainés peuvent servir de base pour de nouveau modèles, ou des parties peuvent être ré-utilisées dans de nouveau modèle, ou faire du *transfer learning*.\n",
        "\n",
        "Ici, nous allons utiliser le modèle *Alexnet* (https://en.wikipedia.org/wiki/AlexNet et https://pytorch.org/hub/pytorch_vision_alexnet/) qui était le gagnant du concours ImageNet de 2012, avec une erreur moyenne top-5 de 15.3% (10.8 points meilleur que le modèle suivant).\n",
        "\n",
        "Le modèle (pré-entrainé) se charge de façon suivante."
      ],
      "metadata": {
        "id": "rPOeNauZNx10"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "alexnet = models.alexnet(pretrained=True)"
      ],
      "metadata": {
        "id": "2192YB0WMmSq",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9f4638a4-e02f-4156-d547-38f6d06a1b41"
      },
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.\n",
            "  warnings.warn(\n",
            "/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=AlexNet_Weights.IMAGENET1K_V1`. You can also use `weights=AlexNet_Weights.DEFAULT` to get the most up-to-date weights.\n",
            "  warnings.warn(msg)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "On peut afficher sa structure de la manière suivante."
      ],
      "metadata": {
        "id": "O9tetg9QaqGE"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "print(alexnet)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "dsQpud7KN81i",
        "outputId": "bd73251b-fee8-4f86-8daf-7b2c75a25db4"
      },
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "AlexNet(\n",
            "  (features): Sequential(\n",
            "    (0): Conv2d(3, 64, kernel_size=(11, 11), stride=(4, 4), padding=(2, 2))\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (3): Conv2d(64, 192, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))\n",
            "    (4): ReLU(inplace=True)\n",
            "    (5): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "    (6): Conv2d(192, 384, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (7): ReLU(inplace=True)\n",
            "    (8): Conv2d(384, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (9): ReLU(inplace=True)\n",
            "    (10): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
            "    (11): ReLU(inplace=True)\n",
            "    (12): MaxPool2d(kernel_size=3, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  )\n",
            "  (avgpool): AdaptiveAvgPool2d(output_size=(6, 6))\n",
            "  (classifier): Sequential(\n",
            "    (0): Dropout(p=0.5, inplace=False)\n",
            "    (1): Linear(in_features=9216, out_features=4096, bias=True)\n",
            "    (2): ReLU(inplace=True)\n",
            "    (3): Dropout(p=0.5, inplace=False)\n",
            "    (4): Linear(in_features=4096, out_features=4096, bias=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): Linear(in_features=4096, out_features=1000, bias=True)\n",
            "  )\n",
            ")\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Afin de comprendre les labels de sortie (1000 classes), on charge le document suivant, qui contient l'interprétation des labels."
      ],
      "metadata": {
        "id": "u-Djcgi0awNF"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "with open(\"/content/drive/MyDrive/Colab Notebooks/\"\n",
        "          \"ml_data/TP8/imagenet_class_labels.txt\") as f:\n",
        "  classes = [line.strip() for line in f.readlines()]"
      ],
      "metadata": {
        "id": "REFocSLQQaFG"
      },
      "execution_count": 46,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Les images entrées dans Alexnet, en format PIL, ont subis les transformations suivantes, que l'on redéfinit ici."
      ],
      "metadata": {
        "id": "taPW42noa-6z"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "transform = transforms.Compose([\n",
        "    transforms.Resize(256),\n",
        "    transforms.CenterCrop(224),\n",
        "    transforms.ToTensor(),\n",
        "    transforms.Normalize(\n",
        "        mean=[0.485, 0.456, 0.406],\n",
        "        std=[0.229, 0.224, 0.225])\n",
        "])"
      ],
      "metadata": {
        "id": "kya6IR-DPPwC"
      },
      "execution_count": 47,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Une image (de citron) a été téléchargée depuis internet (libre à vous de tester une autre image). On la charge, la transforme, la transforme en batch (avec une taille de 1 pour la première dimension), puis on la passe dans Alexnet. On voit qu'en traduisant la sortie, celui-ci prédit la bonne classe avec 45% de certitude."
      ],
      "metadata": {
        "id": "DAROGQambQrQ"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# On charge l'image et on la montre\n",
        "img = Image.open(\"/content/drive/MyDrive/Colab Notebooks/ml_data/TP8/lemon.png\")\n",
        "plt.imshow(img)\n",
        "# On transforme l'image en batch\n",
        "img_tensor = transform(img)\n",
        "img_batch = img_tensor.unsqueeze(0)\n",
        "# On met le modèle en mode évaluation et on l'affecte au dispositif\n",
        "alexnet.eval()\n",
        "alexnet.to(device)\n",
        "# On passe notre image en batch dans le modèle\n",
        "y = alexnet(img_batch.to(device))\n",
        "# On affiche les réponses les plus probables avec un softmax\n",
        "prob = torch.nn.functional.softmax(y, dim=1)[0] * 100\n",
        "_, indices = torch.sort(y, descending=True)\n",
        "for i in indices[0][:5]:\n",
        "  print(classes[i], prob[i].item())"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 522
        },
        "id": "WVuj6loePuTs",
        "outputId": "b91eacad-fdfb-43e7-8b40-8dd7a08809fa"
      },
      "execution_count": 48,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "951: 'lemon', 45.035125732421875\n",
            "950: 'orange', 36.466033935546875\n",
            "945: 'bell pepper', 15.7482328414917\n",
            "940: 'spaghetti squash', 0.6504407525062561\n",
            "948: 'Granny Smith', 0.5691735148429871\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 3.2 Créer un modèle"
      ],
      "metadata": {
        "id": "SawhyMa5zRhl"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "On va maintenant créer un modèle très simple qui traitera notre jeu de données \"contraception.tsv\". Ce dernier se composera de :\n",
        "\n",
        "* 12 neurones d'entrées\n",
        "* Une couche de 12 neurones entièrement connectées, puis activation Relu\n",
        "* Une couche de 6 neurones entièrement connectées, puis activation Relu\n",
        "* Une couche de 3 neurones de sortie qui donnerons les log-odds.\n",
        "\n",
        "Pour créer ce modèle, on va faire une classe héritée de `nn.Module`. Il faudra implémenter deux méthodes :\n",
        "\n",
        "* `__init__(self)` : Le constructeur, qui définit les différentes couches entre les neurones. On utilise ici `nn.Linear(nb_inputs, nb_outputs)` qui créer des poids entre `nb_inputs` et `nb_outputs` neurones. Avant les déclarations, il ne faut pas oublier d'appeler le constructeur de la classe mère avec `super(MiniNet, self).__init__()`.\n",
        "\n",
        "* `forward(self, x)` : La méthode qui effectue le passage en avant des données `x`. Ici, on définit les fonctions d'activation à ce niveau, mais il est aussi possible de les définir au niveau du constructeur (nous le verrons plus tard).\n",
        "\n",
        "Notez que le résultat de notre modèle est ici en log-odd, car nous n'avons pas utilisé une activation de sortie *softmax*. La raison est que nous allons utiliser plus tard la fonctionnelle `nn.CrossEntropyLoss()` pour entrainer ce modèle, et que celle-ci prends en entrée (pour les prédictions) des log-odd (par contre, les vraies classes peuvent être donneés via leur label numérique dans cette fonctionnelle)."
      ],
      "metadata": {
        "id": "LEggQcx4bzPX"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "class MiniNet(nn.Module):\n",
        "\n",
        "  def __init__(self):\n",
        "    super(MiniNet, self).__init__() # On doit appeler le constructeur mère.\n",
        "    self.full_1 = nn.Linear(12, 12) # entrées (12) -> première couche (12)\n",
        "    self.full_2 = nn.Linear(12, 6) # première couche (12) -> deuxième couche (6)\n",
        "    self.full_3 = nn.Linear(6, 3) # deuxième couche (6) -> sorties logit (3)\n",
        "\n",
        "  def forward(self, x):\n",
        "      z_1 = F.relu(self.full_1(x)) # z_1 est le relu de la 1ère couche\n",
        "      z_2 = F.relu(self.full_2(z_1)) # z_2 est le relu de la 2ème couche\n",
        "      z_3 = self.full_3(z_2) # z_3 est le résultat de la 3ème couche\n",
        "      return z_3 # on retourne z_3"
      ],
      "metadata": {
        "id": "OOwKQzb8zWzz"
      },
      "execution_count": 49,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "On affecte notre modèle au dispositif de calcul utilisé, puis on l'affiche."
      ],
      "metadata": {
        "id": "0XSJy7zzee7_"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "mini_model = MiniNet().to(device)\n",
        "print(mini_model)"
      ],
      "metadata": {
        "id": "9SFi6By41uN6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "023ed742-705c-4a91-aa2e-382ee7359961"
      },
      "execution_count": 50,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "MiniNet(\n",
            "  (full_1): Linear(in_features=12, out_features=12, bias=True)\n",
            "  (full_2): Linear(in_features=12, out_features=6, bias=True)\n",
            "  (full_3): Linear(in_features=6, out_features=3, bias=True)\n",
            ")\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "On voit qu'on peut déjà faire passer un individu de notre jeu de données dans le modèle. Seulement, comme les poids sont initilisés aléatoirement et que le modèle n'a pas été entrainé, le résultat sera aléatoire."
      ],
      "metadata": {
        "id": "9vwXYziLfVZz"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "sample = my_data_test[0][0]\n",
        "sample_batch = sample.unsqueeze(0)\n",
        "logit_outputs = mini_model(sample_batch.to(device))\n",
        "prob_outputs = nn.Softmax(dim=1)(logit_outputs)\n",
        "print(my_data_test[0][1].item())\n",
        "print(prob_outputs)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "pLEVgazz6GO-",
        "outputId": "aa9a12d5-e000-4f4b-d6a3-2add66444256"
      },
      "execution_count": 51,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "0\n",
            "tensor([[0.7716, 0.1794, 0.0490]], device='cuda:0', grad_fn=<SoftmaxBackward0>)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 3.3 Entrainer un modèle\n",
        "\n",
        "Pour entrainer notre modèle, on doit maintenant définir la fonctionnelle de perte, qui sera `nn.CrossEntropyLoss()`. On doit également donner un optimisateur (parmi https://pytorch.org/docs/stable/optim.html), qui va dicter comment les poids sont mis à jour par rapport à leur gradient. On utilise ici la *descente du gradient stochastique* avec `torch.optim.SGD(mini_model.parameters(), lr=0.01)`, où les paramètres du modèle sont donnés, et le taux d'apprentissage est fixé à `lr=0.01`."
      ],
      "metadata": {
        "id": "J6uqUxBmfwRg"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "loss_fn = nn.CrossEntropyLoss()\n",
        "optimizer = torch.optim.SGD(mini_model.parameters(), lr=0.01)"
      ],
      "metadata": {
        "id": "r3DZUm3i9494"
      },
      "execution_count": 52,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "La boucle d'entrainement et validation s'effectue de la façon suivante :"
      ],
      "metadata": {
        "id": "2_ryD4eqhWBr"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# On définit le nombre d'epochs\n",
        "n_epochs = 200\n",
        "# Pour stocker les pertes d'entrainement et de validations (graphique à suivre)\n",
        "train_losses = []\n",
        "valid_losses = []\n",
        "# La boucle sur les epochs\n",
        "for epoch in range(n_epochs):\n",
        "\n",
        "    # Entrainement\n",
        "\n",
        "    # On met le modèle en mode entrainement\n",
        "    mini_model.train()\n",
        "\n",
        "    # Pour cumuler la perte\n",
        "    train_loss = 0\n",
        "    # Pour cumuler le nombre d'inidividus bien classés\n",
        "    train_correct = 0\n",
        "    # Boucle sur notre DataLoader d'entrainement\n",
        "    for inputs, labels in train_loader:\n",
        "\n",
        "        # On met nos tenseurs sur notre dispositif de calcul\n",
        "        inputs = inputs.to(device)\n",
        "        labels = labels.to(device)\n",
        "\n",
        "        # On met les gradients à zéro (sinon ils s'accumulent)\n",
        "        optimizer.zero_grad()\n",
        "\n",
        "        # On calcule les log-odds de nos inputs\n",
        "        logit_outputs = mini_model(inputs)\n",
        "        # On calcule la perte\n",
        "        loss = loss_fn(logit_outputs, labels)\n",
        "\n",
        "        # On va maintenant propager le gradient\n",
        "        loss.backward()\n",
        "        # Et ajuster les poids selon l'optimisateur\n",
        "        optimizer.step()\n",
        "\n",
        "        # On ajoute la perte au cumul\n",
        "        train_loss += loss.item()\n",
        "        # On ajoute le nombre d'individus bien classés\n",
        "        train_correct += (logit_outputs.argmax(1) == labels).sum().item()\n",
        "\n",
        "    # Validation\n",
        "\n",
        "    # On passe le modèle en mode évaluation (les poids ne bouge plus)\n",
        "    mini_model.eval()\n",
        "\n",
        "    # Pour cumuler la perte\n",
        "    valid_loss = 0\n",
        "    # Pour cumuler le nombre d'inidividus bien classés\n",
        "    valid_correct = 0\n",
        "    # Boucle sur notre DataLoader de validation\n",
        "    for inputs, labels in valid_loader:\n",
        "\n",
        "        # On met nos tenseurs sur notre dispositif de calcul\n",
        "        inputs = inputs.to(device)\n",
        "        labels = labels.to(device)\n",
        "\n",
        "        # On calcule les log-odds de nos inputs\n",
        "        logit_outputs = mini_model(inputs)\n",
        "        # On calcule la perte\n",
        "        loss = loss_fn(logit_outputs, labels)\n",
        "\n",
        "        # On ajoute la perte au cumul\n",
        "        valid_loss += loss.item()\n",
        "        # On ajoute le nombre d'individus bien classés\n",
        "        valid_correct += (logit_outputs.argmax(1) == labels).sum().item()\n",
        "\n",
        "\n",
        "    # On stocke les pertes d'entrainement et de validations\n",
        "    train_losses.append(train_loss / len(train_loader.dataset))\n",
        "    valid_losses.append(valid_loss / len(valid_loader.dataset))\n",
        "\n",
        "    # Si l'epoch est un multiple de 10, on affiche un message\n",
        "    if epoch % 10 == 0:\n",
        "      print(f\"Epoch {epoch} : \"\n",
        "            f\"train loss = {train_loss / len(train_loader.dataset):.4f} \"\n",
        "            f\"train accuracy = {train_correct / len(train_loader.dataset):.3%} \"\n",
        "            f\"valid loss = {valid_loss / len(valid_loader.dataset):.4f} \"\n",
        "            f\"valid accuracy = {valid_correct / len(valid_loader.dataset):.3%} \")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9-1X8mPS_FJp",
        "outputId": "c5b63509-185c-4a43-9fe0-53f1cb8cf877"
      },
      "execution_count": 53,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 0 : train loss = 0.1345 train accuracy = 42.386% valid loss = 0.1352 valid accuracy = 45.455% \n",
            "Epoch 10 : train loss = 0.1234 train accuracy = 51.018% valid loss = 0.1270 valid accuracy = 51.818% \n",
            "Epoch 20 : train loss = 0.1187 train accuracy = 53.831% valid loss = 0.1203 valid accuracy = 49.545% \n",
            "Epoch 30 : train loss = 0.1164 train accuracy = 55.189% valid loss = 0.1216 valid accuracy = 49.091% \n",
            "Epoch 40 : train loss = 0.1145 train accuracy = 56.644% valid loss = 0.1204 valid accuracy = 47.727% \n",
            "Epoch 50 : train loss = 0.1128 train accuracy = 55.965% valid loss = 0.1230 valid accuracy = 49.091% \n",
            "Epoch 60 : train loss = 0.1126 train accuracy = 57.323% valid loss = 0.1206 valid accuracy = 50.909% \n",
            "Epoch 70 : train loss = 0.1123 train accuracy = 56.450% valid loss = 0.1220 valid accuracy = 51.364% \n",
            "Epoch 80 : train loss = 0.1129 train accuracy = 57.129% valid loss = 0.1202 valid accuracy = 49.545% \n",
            "Epoch 90 : train loss = 0.1106 train accuracy = 59.263% valid loss = 0.1212 valid accuracy = 50.909% \n",
            "Epoch 100 : train loss = 0.1115 train accuracy = 57.808% valid loss = 0.1281 valid accuracy = 45.455% \n",
            "Epoch 110 : train loss = 0.1096 train accuracy = 57.711% valid loss = 0.1259 valid accuracy = 48.636% \n",
            "Epoch 120 : train loss = 0.1103 train accuracy = 58.584% valid loss = 0.1175 valid accuracy = 48.636% \n",
            "Epoch 130 : train loss = 0.1103 train accuracy = 58.972% valid loss = 0.1193 valid accuracy = 48.182% \n",
            "Epoch 140 : train loss = 0.1091 train accuracy = 58.390% valid loss = 0.1251 valid accuracy = 48.636% \n",
            "Epoch 150 : train loss = 0.1094 train accuracy = 58.681% valid loss = 0.1213 valid accuracy = 48.636% \n",
            "Epoch 160 : train loss = 0.1092 train accuracy = 58.972% valid loss = 0.1202 valid accuracy = 52.273% \n",
            "Epoch 170 : train loss = 0.1086 train accuracy = 58.875% valid loss = 0.1252 valid accuracy = 48.182% \n",
            "Epoch 180 : train loss = 0.1092 train accuracy = 58.390% valid loss = 0.1244 valid accuracy = 49.545% \n",
            "Epoch 190 : train loss = 0.1087 train accuracy = 60.524% valid loss = 0.1299 valid accuracy = 44.545% \n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "On va maintenant afficher les courbes, le modèle ne semble pas être surajusté (mais on aurait pu s'arrêter avant)."
      ],
      "metadata": {
        "id": "St_chLb7jNuT"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "fig, ax = plt.subplots()\n",
        "ax.plot(train_losses, c=\"red\", label=\"Train\")\n",
        "ax.plot(valid_losses, c=\"blue\", label=\"Valid\")\n",
        "ax.set_xlabel(\"Epoch\")\n",
        "ax.set_ylabel(\"Loss\")\n",
        "ax.legend()\n",
        "plt.plot()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 468
        },
        "id": "wholHVdiDNs0",
        "outputId": "4920c07a-79d2-4616-8770-6a6620d63dc4"
      },
      "execution_count": 54,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[]"
            ]
          },
          "metadata": {},
          "execution_count": 54
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "On va maintenant tester notre modèle entrainé sur le jeu de test."
      ],
      "metadata": {
        "id": "oBLk6PAyjWFO"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# On met le modèle en mode évaluation\n",
        "mini_model.eval()\n",
        "\n",
        "# Pour cumuler le nombre d'inidividus bien classés\n",
        "test_correct = 0\n",
        "# Boucle sur notre DataLoader de test\n",
        "for inputs, labels in test_loader:\n",
        "\n",
        "  # On met nos tenseurs sur notre dispositif de calcul\n",
        "  inputs = inputs.to(device)\n",
        "  labels = labels.to(device)\n",
        "\n",
        "  # On calcule les log-odds de nos inputs\n",
        "  logit_outputs = mini_model(inputs)\n",
        "\n",
        "  # On ajoute le nombre d'individus bien classés\n",
        "  test_correct += (logit_outputs.argmax(1) == labels).sum().item()\n",
        "\n",
        "# On calcule l'exactitude\n",
        "test_accuracy = test_correct / len(test_loader.dataset)\n",
        "\n",
        "# On l'affiche\n",
        "print(f\"Test Accuracy: {test_accuracy}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "G1HU-pNWG4pI",
        "outputId": "8fe71ed2-d066-422e-c1d5-a359670a39a6"
      },
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Test Accuracy: 0.5045045045045045\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "### 3.4 Sauver et charger un modèle\n",
        "\n",
        "Une fois le modèle entrainer, on peut avoir envie de la sauvegarder afin de le réutiliser pour faire des prédictions (ou pour le partager). Il existe deux manières de faire des sauvegardes."
      ],
      "metadata": {
        "id": "uJKbPhH8OdTO"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "La première consiste à sauvegarder, via `torch.save()`, uniquement les **paramètres** du modèle, qui sont obtenus avec la méthode `state_dict()`. L'extension privilégiée pour sauvegarder un modèle PyTorch est `.pt` ou `.pth`."
      ],
      "metadata": {
        "id": "5tYM-SrNlFHb"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "weights_path = \\\n",
        "  \"/content/drive/MyDrive/Colab Notebooks/ml_data/TP8/mini_model_weights.pth\"\n",
        "torch.save(mini_model.state_dict(), weights_path)"
      ],
      "metadata": {
        "id": "1f2Y8oxROccC"
      },
      "execution_count": 56,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "Avec cette méthode de sauvegarde, on doit d'abord instancier un modèle qui a la même structure, puis faire appel à la méthode `load_state_dict()` sur un `torch.load()` pour mettre à jour ses poids."
      ],
      "metadata": {
        "id": "RJ6r4E3slbtM"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "new_model = MiniNet()\n",
        "new_model.load_state_dict(torch.load(weights_path, weights_only=True))\n",
        "new_model.to(device)\n",
        "print(new_model(sample_batch.to(device)))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DNtVNNgHhnLX",
        "outputId": "db042e1a-3b60-4cc3-8ff7-6cf1255f0542"
      },
      "execution_count": 57,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[ 0.2077, -2.6023, -1.3191]], device='cuda:0',\n",
            "       grad_fn=<AddmmBackward0>)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "L'autre manière de procéder est de sauvegarder entièrement le modèle, en appliquant `torch.save()` directement sur le modèle."
      ],
      "metadata": {
        "id": "dPayLAcMl6Tk"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "model_path = \"/content/drive/MyDrive/Colab Notebooks/ml_data/TP8/mini_model.pth\"\n",
        "torch.save(mini_model, model_path)"
      ],
      "metadata": {
        "id": "mO3ghuuSiN82"
      },
      "execution_count": 58,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "On peut alors le charger directement avec `torch.load()`."
      ],
      "metadata": {
        "id": "pOmiYTmwmFX8"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "new_model2 = torch.load(model_path)\n",
        "print(new_model2(sample_batch.to(device)))"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OD7GOWMcjgOW",
        "outputId": "8e459a85-b09b-4b39-b9e4-acd6d0d47aa4"
      },
      "execution_count": 59,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "tensor([[ 0.2077, -2.6023, -1.3191]], device='cuda:0',\n",
            "       grad_fn=<AddmmBackward0>)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "<ipython-input-59-00a7894d86e6>:1: FutureWarning: You are using `torch.load` with `weights_only=False` (the current default value), which uses the default pickle module implicitly. It is possible to construct malicious pickle data which will execute arbitrary code during unpickling (See https://github.com/pytorch/pytorch/blob/main/SECURITY.md#untrusted-models for more details). In a future release, the default value for `weights_only` will be flipped to `True`. This limits the functions that could be executed during unpickling. Arbitrary objects will no longer be allowed to be loaded via this mode unless they are explicitly allowlisted by the user via `torch.serialization.add_safe_globals`. We recommend you start setting `weights_only=True` for any use case where you don't have full control of the loaded file. Please open an issue on GitHub for any issues related to this experimental feature.\n",
            "  new_model2 = torch.load(model_path)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Bien que moins intuitive, la permière méthode de sauvegarde est privilégiée. En effet, la seconde méthode est très sensible à la moindre restructuration du code et présente des risques de sécurité."
      ],
      "metadata": {
        "id": "lzNo2S6-mePQ"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "---"
      ],
      "metadata": {
        "id": "ASn0-kswkdmC"
      }
    }
  ]
}