ML-Intro-Course/9_MNIST_torch.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bRmpRHXBQI9u"
      },
      "source": [
        "# Introduction to PyTorch using the Digit-MNIST dataset\n",
        "\n",
        "<a href=\"https://colab.research.google.com/drive/1wR5WnJFE9mGov_-yPkLagGXTR41VAGu5\" target=\"_blank\">\n",
        "  <img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\">\n",
        "</a>\n",
        "\n",
        "Return to the [castle](https://github.com/Nkluge-correa/TeenyTinyCastle).\n",
        "\n",
        "[PyTorch](https://pytorch.org/) is a popular open-source machine learning framework developed by Facebook's AI Research team. It is built to provide flexibility and speed to researchers and developers working on deep learning projects.\n",
        "\n",
        "PyTorch offers an intuitive and Pythonic interface that enables easy construction and training of complex neural networks. PyTorch also offers a rich set of tools and utilities for data preprocessing, model evaluation, and deployment. It has gained significant traction in the deep learning community due to its ease of use, speed, and flexibility.\n",
        "\n",
        "\n",
        "![torch](https://user-images.githubusercontent.com/42931974/68615320-d6bf3380-04e8-11ea-84f8-dcef049f1ed3.gif)\n",
        "\n",
        "[Source](https://github.com/AnilNITT/PyTorch-for-beginner).\n",
        "\n",
        "In this tutorial, we will implement a neural network, not from scratch, but using PyTorch. We will be training our network via `supervised learning`.\n",
        "\n",
        "PyTorch offers domain-specific libraries such as [TorchText](https://pytorch.org/text/stable/index.html), [TorchVision](https://pytorch.org/vision/stable/index.html), and [TorchAudio](https://pytorch.org/audio/stable/index.html), all of which include datasets. For this tutorial, we will be using a TorchVision dataset.\n",
        "\n",
        "The `torchvision.datasets` module contains `Dataset` objects for many real-world vision data like CIFAR, COCO ([full list here](https://pytorch.org/vision/stable/datasets.html), or just use `dir(datasets)`).\n",
        "\n",
        "In this tutorial, we use the `Fashion_MNIST dataset`. Every TorchVision `Dataset` includes two arguments: `transform` and `target_transform` to modify the samples and labels, respectively.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WdV9gwEVQI9w",
        "outputId": "ec82a02f-73dc-4573-95c8-6180b4624542"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz\n",
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz to data/FashionMNIST/raw/train-images-idx3-ubyte.gz\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 26421880/26421880 [00:00<00:00, 105338326.41it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Extracting data/FashionMNIST/raw/train-images-idx3-ubyte.gz to data/FashionMNIST/raw\n",
            "\n",
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz\n",
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz to data/FashionMNIST/raw/train-labels-idx1-ubyte.gz\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 29515/29515 [00:00<00:00, 8981056.48it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Extracting data/FashionMNIST/raw/train-labels-idx1-ubyte.gz to data/FashionMNIST/raw\n",
            "\n",
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz\n",
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz to data/FashionMNIST/raw/t10k-images-idx3-ubyte.gz\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 4422102/4422102 [00:00<00:00, 65577598.62it/s]\n"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Extracting data/FashionMNIST/raw/t10k-images-idx3-ubyte.gz to data/FashionMNIST/raw\n",
            "\n",
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz\n",
            "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz to data/FashionMNIST/raw/t10k-labels-idx1-ubyte.gz\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "100%|██████████| 5148/5148 [00:00<00:00, 17427180.78it/s]"
          ]
        },
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Extracting data/FashionMNIST/raw/t10k-labels-idx1-ubyte.gz to data/FashionMNIST/raw\n",
            "\n"
          ]
        },
        {
          "name": "stderr",
          "output_type": "stream",
          "text": [
            "\n"
          ]
        }
      ],
      "source": [
        "from torchvision import datasets\n",
        "from torchvision.transforms import ToTensor\n",
        "\n",
        "training_data = datasets.FashionMNIST(\n",
        "    root=\"data\",\n",
        "    train=True,\n",
        "    download=True,\n",
        "    transform=ToTensor(),\n",
        ")\n",
        "\n",
        "test_data = datasets.FashionMNIST(\n",
        "    root=\"data\",\n",
        "    train=False,\n",
        "    download=True,\n",
        "    transform=ToTensor(),\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cuMaymrtQI9x"
      },
      "source": [
        "Loading the `datasets.FashionMNIST` gives us the same dataset we used in the previous notebook."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 807
        },
        "id": "OQVVeqe1QI9x",
        "outputId": "f503873a-9178-4980-ce00-743f9630139e"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 1000x1000 with 25 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "import torch\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "plt.figure(figsize=[10, 10])\n",
        "for i in range(25):\n",
        "    plt.subplot(5, 5, i+1)\n",
        "    plt.xticks([])\n",
        "    plt.yticks([])\n",
        "    plt.grid(False)\n",
        "    plt.imshow(torch.squeeze(training_data[i][0]), cmap=plt.cm.binary)\n",
        "\n",
        "plt.show()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "W8CG8XCnQI9x"
      },
      "source": [
        "PyTorch has its own methods. For example, here, we will pass the `Dataset` as an argument to a `DataLoader`. This wraps an iterable over our dataset and supports automatic batching, sampling, shuffling, and multiprocess data loading. Here we define a batch size of 64, i.e., each element in the `DataLoader` iterable will return a batch of 64 images and labels."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ay-mJBmDQI9x",
        "outputId": "93013ac6-078d-4d33-afee-62d4ca4b3485"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Sixe of train_data: 60000\n",
            "Sixe of test_data: 10000\n",
            "Shape of X [Batch, Channels, Height, Width]: torch.Size([64, 1, 28, 28])\n",
            "Shape of y: torch.Size([64]) torch.int64\n"
          ]
        }
      ],
      "source": [
        "from torch.utils.data import DataLoader\n",
        "\n",
        "batch_size = 64\n",
        "\n",
        "train_dataloader = DataLoader(training_data, batch_size=batch_size)\n",
        "test_dataloader = DataLoader(test_data, batch_size=batch_size)\n",
        "\n",
        "for X, y in train_dataloader:\n",
        "    print(f'Sixe of train_data: {len(train_dataloader.dataset)}')\n",
        "    print(f'Sixe of test_data: {len(test_dataloader.dataset)}')\n",
        "    print(f'Shape of X [Batch, Channels, Height, Width]: {X.shape}')\n",
        "    print(f'Shape of y: {y.shape} {y.dtype}')\n",
        "    break"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3B9fpnsAQI9x"
      },
      "source": [
        "Neural networks in PyTorch are usually built using classes, inducing the user to OOP (`object oriented programming`).\n",
        "\n",
        "To define a neural network in PyTorch, we create a class that inherits from `nn.Module`. We define the network layers in the `__init__` function and specify how data will pass through the network in the `forward` function. We move it to the GPU if available to accelerate operations in the neural network."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "blbZIoWhQI9x",
        "outputId": "4e861066-6975-44f9-fa5f-66fa2374c743"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Using cpu device\n",
            "NeuralNetwork(\n",
            "  (flatten): Flatten(start_dim=1, end_dim=-1)\n",
            "  (linear_relu_stack): Sequential(\n",
            "    (0): Linear(in_features=784, out_features=512, bias=True)\n",
            "    (1): ReLU()\n",
            "    (2): Linear(in_features=512, out_features=256, bias=True)\n",
            "    (3): ReLU()\n",
            "    (4): Linear(in_features=256, out_features=128, bias=True)\n",
            "    (5): ReLU()\n",
            "    (6): Linear(in_features=128, out_features=10, bias=True)\n",
            "  )\n",
            ")\n"
          ]
        }
      ],
      "source": [
        "from torch import nn\n",
        "\n",
        "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
        "print(f\"Using {device} device\")\n",
        "\n",
        "class NeuralNetwork(nn.Module):\n",
        "    \"\"\"\n",
        "    A feedforward neural network that consists of\n",
        "    four fully connected layers with ReLU activation\n",
        "    functions, and a softmax activation function at\n",
        "    the output layer. The network takes input of\n",
        "    shape (batch_size, 1, 28, 28) and produces\n",
        "    output of shape (batch_size, 10), where each\n",
        "    element in the output represents the predicted\n",
        "    probability of the input image belonging to\n",
        "    one of the 10 classes (0-9).\n",
        "\n",
        "    Attributes:\n",
        "    -----------\n",
        "    flatten : nn.Flatten\n",
        "        A module that flattens the input to a 1D tensor.\n",
        "    linear_relu_stack : nn.Sequential\n",
        "        A sequence of fully connected layers with ReLU activation\n",
        "        functions.\n",
        "\n",
        "    Methods:\n",
        "    --------\n",
        "    forward(x)\n",
        "        Performs forward propagation on the input tensor x\n",
        "        and returns the output logits.\n",
        "    \"\"\"\n",
        "    def __init__(self):\n",
        "        super(NeuralNetwork, self).__init__()\n",
        "        self.flatten = nn.Flatten()\n",
        "        self.linear_relu_stack = nn.Sequential(\n",
        "            nn.Linear(28*28, 512),\n",
        "            nn.ReLU(),\n",
        "            nn.Linear(512, 256),\n",
        "            nn.ReLU(),\n",
        "            nn.Linear(256, 128),\n",
        "            nn.ReLU(),\n",
        "            nn.Linear(128, 10)\n",
        "        )\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.flatten(x)\n",
        "        logits = self.linear_relu_stack(x)\n",
        "        return logits\n",
        "\n",
        "\n",
        "model = NeuralNetwork().to(device)\n",
        "print(model)\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BzFuG5e8QI9y"
      },
      "source": [
        "As before, we need a `loss function` and an `optimizer` to train a model. In a single training loop, the model makes predictions on the training dataset (fed to it in batches) and backpropagates the prediction error to adjust the model’s parameters.\n",
        "\n",
        "We will also create a function to evaluate our model after training."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "aMUJfe8QQI9y"
      },
      "outputs": [],
      "source": [
        "# our loss function\n",
        "loss_fn = nn.CrossEntropyLoss()\n",
        "\n",
        "# the optimizer\n",
        "optimizer = torch.optim.SGD(model.parameters(), lr=1e-3)\n",
        "\n",
        "\n",
        "def train(dataloader, model, loss_fn, optimizer):\n",
        "    size = len(dataloader.dataset)\n",
        "    model.train()\n",
        "\n",
        "    # for every batch in the dataloader\n",
        "    for batch, (X, y) in enumerate(dataloader):\n",
        "\n",
        "        # move the batch to the device (CPU or GPU)\n",
        "        X, y = X.to(device), y.to(device)\n",
        "\n",
        "        # make a prediction\n",
        "        pred = model(X)\n",
        "\n",
        "        # use the loss function to calculate the loss\n",
        "        loss = loss_fn(pred, y)\n",
        "\n",
        "        # reset the gradients of the optimizer to zero\n",
        "        optimizer.zero_grad()\n",
        "\n",
        "        # backpropagate the loss\n",
        "        loss.backward()\n",
        "\n",
        "        # perform an update to the parameters\n",
        "        optimizer.step()\n",
        "\n",
        "        if batch % 100 == 0:\n",
        "            loss, current = loss.item(), batch * len(X)\n",
        "            print(f'>> Loss: {loss:>7f}  [{current:>5d}/{size:>5d}]')\n",
        "\n",
        "\n",
        "def test(dataloader, model, loss_fn):\n",
        "    size = len(dataloader.dataset)\n",
        "    num_batches = len(dataloader)\n",
        "    model.eval()\n",
        "    test_loss, correct = 0, 0\n",
        "    with torch.no_grad():\n",
        "        for X, y in dataloader:\n",
        "            X, y = X.to(device), y.to(device)\n",
        "            pred = model(X)\n",
        "            test_loss += loss_fn(pred, y).item()\n",
        "\n",
        "            # calculate the accuracy of the predictions in a batch\n",
        "            correct += (pred.argmax(1) == y).type(torch.float).sum().item()\n",
        "    test_loss /= num_batches\n",
        "    correct /= size\n",
        "    print(\n",
        "        f\"Test Error: \\n>> Accuracy: {(100*correct):>0.1f}%\\n>> Avg loss: {test_loss:>8f} \\n\")\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EYrXYliLQI9y"
      },
      "source": [
        "The training process is conducted over several iterations (_epochs_). As usual, the model learns new parameters during each epoch to make better predictions.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eHn4eXmCQI9y",
        "outputId": "9fbabab0-6b89-41fe-b607-5ae4e52dc85b"
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1\n",
            "--------------------------------\n",
            ">> Loss: 2.300707  [    0/60000]\n",
            ">> Loss: 2.302816  [ 6400/60000]\n",
            ">> Loss: 2.301031  [12800/60000]\n",
            ">> Loss: 2.301285  [19200/60000]\n",
            ">> Loss: 2.294868  [25600/60000]\n",
            ">> Loss: 2.283386  [32000/60000]\n",
            ">> Loss: 2.287358  [38400/60000]\n",
            ">> Loss: 2.279450  [44800/60000]\n",
            ">> Loss: 2.281208  [51200/60000]\n",
            ">> Loss: 2.267759  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 20.8%\n",
            ">> Avg loss: 2.277023 \n",
            "\n",
            "Epoch 2\n",
            "--------------------------------\n",
            ">> Loss: 2.275969  [    0/60000]\n",
            ">> Loss: 2.279758  [ 6400/60000]\n",
            ">> Loss: 2.271825  [12800/60000]\n",
            ">> Loss: 2.275166  [19200/60000]\n",
            ">> Loss: 2.268454  [25600/60000]\n",
            ">> Loss: 2.251772  [32000/60000]\n",
            ">> Loss: 2.260715  [38400/60000]\n",
            ">> Loss: 2.246759  [44800/60000]\n",
            ">> Loss: 2.249928  [51200/60000]\n",
            ">> Loss: 2.232908  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 28.7%\n",
            ">> Avg loss: 2.241486 \n",
            "\n",
            "Epoch 3\n",
            "--------------------------------\n",
            ">> Loss: 2.243896  [    0/60000]\n",
            ">> Loss: 2.246746  [ 6400/60000]\n",
            ">> Loss: 2.228395  [12800/60000]\n",
            ">> Loss: 2.234327  [19200/60000]\n",
            ">> Loss: 2.224177  [25600/60000]\n",
            ">> Loss: 2.197957  [32000/60000]\n",
            ">> Loss: 2.212344  [38400/60000]\n",
            ">> Loss: 2.185745  [44800/60000]\n",
            ">> Loss: 2.189231  [51200/60000]\n",
            ">> Loss: 2.162194  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 29.5%\n",
            ">> Avg loss: 2.171911 \n",
            "\n",
            "Epoch 4\n",
            "--------------------------------\n",
            ">> Loss: 2.179961  [    0/60000]\n",
            ">> Loss: 2.180126  [ 6400/60000]\n",
            ">> Loss: 2.140685  [12800/60000]\n",
            ">> Loss: 2.150341  [19200/60000]\n",
            ">> Loss: 2.128035  [25600/60000]\n",
            ">> Loss: 2.081604  [32000/60000]\n",
            ">> Loss: 2.103177  [38400/60000]\n",
            ">> Loss: 2.048687  [44800/60000]\n",
            ">> Loss: 2.051389  [51200/60000]\n",
            ">> Loss: 1.998331  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 37.1%\n",
            ">> Avg loss: 2.013372 \n",
            "\n",
            "Epoch 5\n",
            "--------------------------------\n",
            ">> Loss: 2.036612  [    0/60000]\n",
            ">> Loss: 2.021858  [ 6400/60000]\n",
            ">> Loss: 1.942581  [12800/60000]\n",
            ">> Loss: 1.956688  [19200/60000]\n",
            ">> Loss: 1.897852  [25600/60000]\n",
            ">> Loss: 1.842977  [32000/60000]\n",
            ">> Loss: 1.860637  [38400/60000]\n",
            ">> Loss: 1.775344  [44800/60000]\n",
            ">> Loss: 1.794867  [51200/60000]\n",
            ">> Loss: 1.701439  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 43.6%\n",
            ">> Avg loss: 1.726013 \n",
            "\n",
            "Epoch 6\n",
            "--------------------------------\n",
            ">> Loss: 1.784411  [    0/60000]\n",
            ">> Loss: 1.744549  [ 6400/60000]\n",
            ">> Loss: 1.633147  [12800/60000]\n",
            ">> Loss: 1.657062  [19200/60000]\n",
            ">> Loss: 1.580856  [25600/60000]\n",
            ">> Loss: 1.563868  [32000/60000]\n",
            ">> Loss: 1.563860  [38400/60000]\n",
            ">> Loss: 1.489828  [44800/60000]\n",
            ">> Loss: 1.524417  [51200/60000]\n",
            ">> Loss: 1.427534  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 53.2%\n",
            ">> Avg loss: 1.449983 \n",
            "\n",
            "Epoch 7\n",
            "--------------------------------\n",
            ">> Loss: 1.535621  [    0/60000]\n",
            ">> Loss: 1.496325  [ 6400/60000]\n",
            ">> Loss: 1.364728  [12800/60000]\n",
            ">> Loss: 1.406826  [19200/60000]\n",
            ">> Loss: 1.332364  [25600/60000]\n",
            ">> Loss: 1.348769  [32000/60000]\n",
            ">> Loss: 1.347738  [38400/60000]\n",
            ">> Loss: 1.282905  [44800/60000]\n",
            ">> Loss: 1.315721  [51200/60000]\n",
            ">> Loss: 1.244001  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 56.5%\n",
            ">> Avg loss: 1.254507 \n",
            "\n",
            "Epoch 8\n",
            "--------------------------------\n",
            ">> Loss: 1.344694  [    0/60000]\n",
            ">> Loss: 1.318388  [ 6400/60000]\n",
            ">> Loss: 1.167307  [12800/60000]\n",
            ">> Loss: 1.245180  [19200/60000]\n",
            ">> Loss: 1.156284  [25600/60000]\n",
            ">> Loss: 1.192663  [32000/60000]\n",
            ">> Loss: 1.207504  [38400/60000]\n",
            ">> Loss: 1.140078  [44800/60000]\n",
            ">> Loss: 1.174623  [51200/60000]\n",
            ">> Loss: 1.123350  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 59.0%\n",
            ">> Avg loss: 1.124394 \n",
            "\n",
            "Epoch 9\n",
            "--------------------------------\n",
            ">> Loss: 1.212616  [    0/60000]\n",
            ">> Loss: 1.200843  [ 6400/60000]\n",
            ">> Loss: 1.028962  [12800/60000]\n",
            ">> Loss: 1.138399  [19200/60000]\n",
            ">> Loss: 1.034074  [25600/60000]\n",
            ">> Loss: 1.078335  [32000/60000]\n",
            ">> Loss: 1.113373  [38400/60000]\n",
            ">> Loss: 1.041191  [44800/60000]\n",
            ">> Loss: 1.070676  [51200/60000]\n",
            ">> Loss: 1.039989  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 61.4%\n",
            ">> Avg loss: 1.032725 \n",
            "\n",
            "Epoch 10\n",
            "--------------------------------\n",
            ">> Loss: 1.113246  [    0/60000]\n",
            ">> Loss: 1.120748  [ 6400/60000]\n",
            ">> Loss: 0.927113  [12800/60000]\n",
            ">> Loss: 1.061001  [19200/60000]\n",
            ">> Loss: 0.947829  [25600/60000]\n",
            ">> Loss: 0.990108  [32000/60000]\n",
            ">> Loss: 1.047203  [38400/60000]\n",
            ">> Loss: 0.972910  [44800/60000]\n",
            ">> Loss: 0.989396  [51200/60000]\n",
            ">> Loss: 0.979301  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 63.7%\n",
            ">> Avg loss: 0.965095 \n",
            "\n",
            "Epoch 11\n",
            "--------------------------------\n",
            ">> Loss: 1.034356  [    0/60000]\n",
            ">> Loss: 1.062503  [ 6400/60000]\n",
            ">> Loss: 0.849190  [12800/60000]\n",
            ">> Loss: 1.000777  [19200/60000]\n",
            ">> Loss: 0.886895  [25600/60000]\n",
            ">> Loss: 0.919303  [32000/60000]\n",
            ">> Loss: 0.998082  [38400/60000]\n",
            ">> Loss: 0.924915  [44800/60000]\n",
            ">> Loss: 0.925186  [51200/60000]\n",
            ">> Loss: 0.933067  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 65.6%\n",
            ">> Avg loss: 0.913602 \n",
            "\n",
            "Epoch 12\n",
            "--------------------------------\n",
            ">> Loss: 0.970244  [    0/60000]\n",
            ">> Loss: 1.016292  [ 6400/60000]\n",
            ">> Loss: 0.787810  [12800/60000]\n",
            ">> Loss: 0.951466  [19200/60000]\n",
            ">> Loss: 0.842993  [25600/60000]\n",
            ">> Loss: 0.861495  [32000/60000]\n",
            ">> Loss: 0.959498  [38400/60000]\n",
            ">> Loss: 0.890179  [44800/60000]\n",
            ">> Loss: 0.874978  [51200/60000]\n",
            ">> Loss: 0.896137  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 67.2%\n",
            ">> Avg loss: 0.873316 \n",
            "\n",
            "Epoch 13\n",
            "--------------------------------\n",
            ">> Loss: 0.917380  [    0/60000]\n",
            ">> Loss: 0.976684  [ 6400/60000]\n",
            ">> Loss: 0.738077  [12800/60000]\n",
            ">> Loss: 0.910363  [19200/60000]\n",
            ">> Loss: 0.810300  [25600/60000]\n",
            ">> Loss: 0.814237  [32000/60000]\n",
            ">> Loss: 0.927966  [38400/60000]\n",
            ">> Loss: 0.863704  [44800/60000]\n",
            ">> Loss: 0.835693  [51200/60000]\n",
            ">> Loss: 0.865179  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 68.5%\n",
            ">> Avg loss: 0.840866 \n",
            "\n",
            "Epoch 14\n",
            "--------------------------------\n",
            ">> Loss: 0.872970  [    0/60000]\n",
            ">> Loss: 0.941023  [ 6400/60000]\n",
            ">> Loss: 0.696965  [12800/60000]\n",
            ">> Loss: 0.875664  [19200/60000]\n",
            ">> Loss: 0.784204  [25600/60000]\n",
            ">> Loss: 0.775430  [32000/60000]\n",
            ">> Loss: 0.901064  [38400/60000]\n",
            ">> Loss: 0.841903  [44800/60000]\n",
            ">> Loss: 0.804989  [51200/60000]\n",
            ">> Loss: 0.838034  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 69.2%\n",
            ">> Avg loss: 0.813881 \n",
            "\n",
            "Epoch 15\n",
            "--------------------------------\n",
            ">> Loss: 0.834721  [    0/60000]\n",
            ">> Loss: 0.908676  [ 6400/60000]\n",
            ">> Loss: 0.662182  [12800/60000]\n",
            ">> Loss: 0.846129  [19200/60000]\n",
            ">> Loss: 0.762123  [25600/60000]\n",
            ">> Loss: 0.743424  [32000/60000]\n",
            ">> Loss: 0.877070  [38400/60000]\n",
            ">> Loss: 0.823204  [44800/60000]\n",
            ">> Loss: 0.781011  [51200/60000]\n",
            ">> Loss: 0.814059  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 70.2%\n",
            ">> Avg loss: 0.790730 \n",
            "\n",
            "Epoch 16\n",
            "--------------------------------\n",
            ">> Loss: 0.801312  [    0/60000]\n",
            ">> Loss: 0.878314  [ 6400/60000]\n",
            ">> Loss: 0.632577  [12800/60000]\n",
            ">> Loss: 0.821186  [19200/60000]\n",
            ">> Loss: 0.743208  [25600/60000]\n",
            ">> Loss: 0.716651  [32000/60000]\n",
            ">> Loss: 0.855126  [38400/60000]\n",
            ">> Loss: 0.806264  [44800/60000]\n",
            ">> Loss: 0.761709  [51200/60000]\n",
            ">> Loss: 0.792518  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 71.0%\n",
            ">> Avg loss: 0.770374 \n",
            "\n",
            "Epoch 17\n",
            "--------------------------------\n",
            ">> Loss: 0.771427  [    0/60000]\n",
            ">> Loss: 0.849991  [ 6400/60000]\n",
            ">> Loss: 0.606937  [12800/60000]\n",
            ">> Loss: 0.799830  [19200/60000]\n",
            ">> Loss: 0.726487  [25600/60000]\n",
            ">> Loss: 0.693962  [32000/60000]\n",
            ">> Loss: 0.834276  [38400/60000]\n",
            ">> Loss: 0.790568  [44800/60000]\n",
            ">> Loss: 0.745580  [51200/60000]\n",
            ">> Loss: 0.773139  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 71.9%\n",
            ">> Avg loss: 0.752094 \n",
            "\n",
            "Epoch 18\n",
            "--------------------------------\n",
            ">> Loss: 0.744174  [    0/60000]\n",
            ">> Loss: 0.823787  [ 6400/60000]\n",
            ">> Loss: 0.584241  [12800/60000]\n",
            ">> Loss: 0.781318  [19200/60000]\n",
            ">> Loss: 0.711610  [25600/60000]\n",
            ">> Loss: 0.674501  [32000/60000]\n",
            ">> Loss: 0.814207  [38400/60000]\n",
            ">> Loss: 0.775721  [44800/60000]\n",
            ">> Loss: 0.731552  [51200/60000]\n",
            ">> Loss: 0.755625  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 72.6%\n",
            ">> Avg loss: 0.735432 \n",
            "\n",
            "Epoch 19\n",
            "--------------------------------\n",
            ">> Loss: 0.719287  [    0/60000]\n",
            ">> Loss: 0.799903  [ 6400/60000]\n",
            ">> Loss: 0.564088  [12800/60000]\n",
            ">> Loss: 0.764797  [19200/60000]\n",
            ">> Loss: 0.698410  [25600/60000]\n",
            ">> Loss: 0.657768  [32000/60000]\n",
            ">> Loss: 0.794931  [38400/60000]\n",
            ">> Loss: 0.762093  [44800/60000]\n",
            ">> Loss: 0.718880  [51200/60000]\n",
            ">> Loss: 0.739815  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 73.4%\n",
            ">> Avg loss: 0.720104 \n",
            "\n",
            "Epoch 20\n",
            "--------------------------------\n",
            ">> Loss: 0.696302  [    0/60000]\n",
            ">> Loss: 0.778544  [ 6400/60000]\n",
            ">> Loss: 0.545843  [12800/60000]\n",
            ">> Loss: 0.749926  [19200/60000]\n",
            ">> Loss: 0.687097  [25600/60000]\n",
            ">> Loss: 0.643290  [32000/60000]\n",
            ">> Loss: 0.776363  [38400/60000]\n",
            ">> Loss: 0.749181  [44800/60000]\n",
            ">> Loss: 0.707411  [51200/60000]\n",
            ">> Loss: 0.725430  [57600/60000]\n",
            "Test Error: \n",
            ">> Accuracy: 74.2%\n",
            ">> Avg loss: 0.705862 \n",
            "\n"
          ]
        }
      ],
      "source": [
        "epochs = 20\n",
        "for t in range(epochs):\n",
        "    print(f\"Epoch {t+1}\\n--------------------------------\")\n",
        "    train(train_dataloader, model, loss_fn, optimizer)\n",
        "    test(test_dataloader, model, loss_fn)\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2t0QIrb0QI9y"
      },
      "source": [
        "Like in Keras, we can easily save and load our model.\n",
        "\n",
        "A common way to save a model is to serialize the internal state dictionary (containing the model parameters). The process for loading a model includes re-creating the model structure and loading the state dictionary into it.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "P3ZUDE9aQI9y",
        "outputId": "7db629f2-24cb-473c-d4df-5c5ccc7ef2de"
      },
      "outputs": [
        {
          "data": {
            "text/plain": [
              "<All keys matched successfully>"
            ]
          },
          "execution_count": 7,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "torch.save(model.state_dict(), 'fashion_model.pt')\n",
        "model = NeuralNetwork()\n",
        "model.load_state_dict(torch.load('fashion_model.pt'))\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zu9Lhe_MQI9y"
      },
      "source": [
        "This model can now be used to make predictions. The trained model outputs raw prediction `logits`. To convert them to a probability distribution, we can use (just like before) a `softmax function`.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 556
        },
        "id": "dBOb5-rlQI9y",
        "outputId": "ae73a814-d799-44be-d775-348ec6ffceb1"
      },
      "outputs": [
        {
          "data": {
            "image/png": "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",
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ],
      "source": [
        "import torch.nn.functional as nnf\n",
        "import numpy as np\n",
        "\n",
        "model.eval()\n",
        "\n",
        "classes = [\n",
        "    \"T-shirt/top\",\n",
        "    \"Trouser\",\n",
        "    \"Pullover\",\n",
        "    \"Dress\",\n",
        "    \"Coat\",\n",
        "    \"Sandal\",\n",
        "    \"Shirt\",\n",
        "    \"Sneaker\",\n",
        "    \"Bag\",\n",
        "    \"Ankle boot\",\n",
        "]\n",
        "\n",
        "sample = np.random.randint(0, 9999)\n",
        "x, y = test_data[sample][0], test_data[sample][1]\n",
        "\n",
        "with torch.no_grad():\n",
        "    pred = model(x)\n",
        "\n",
        "prob = nnf.softmax(pred, dim=1)\n",
        "\n",
        "top_p, top_class = prob.topk(3, dim=1)\n",
        "\n",
        "predicted, actual = classes[top_class[0][0]], classes[y]\n",
        "\n",
        "plt.imshow(torch.squeeze(test_data[sample][0]), cmap=plt.cm.binary)\n",
        "plt.title(f'''\n",
        "Predicted: \"{predicted}\", Actual: \"{actual}\n",
        "1º -> {classes[top_class[0][0]]}: {(top_p[0][0]*100):>0.2f}%\n",
        "2º -> {classes[top_class[0][1]]}: {(top_p[0][1]*100):>0.2f}%\n",
        "2º -> {classes[top_class[0][2]]}: {(top_p[0][2]*100):>0.2f}%\n",
        "''')\n",
        "plt.show()\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xwwi4B0gQI9z"
      },
      "source": [
        "Pytorch gives the user greater visibility to much of what the Keras/Tensorflow library hides. However, in the end, both frameworks make sure the job gets done, and which one you choose to use (_you will probably use both if you want to practice ML_) is a matter of taste and preference! 🙃\n",
        "\n",
        "---\n",
        "\n",
        "Return to the [castle](https://github.com/Nkluge-correa/TeenyTinyCastle).\n"
      ]
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.9.13"
    },
    "orig_nbformat": 4,
    "vscode": {
      "interpreter": {
        "hash": "aca09746cf57686f00a55ae76e987247ecfb5dd0b3b2e2474d8dbbf0c5e3377e"
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}