task-2-nn/IMT2020111.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 5,
      "metadata": {
        "id": "IMtHLx3lep6t"
      },
      "outputs": [],
      "source": [
        "%matplotlib inline"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kwrUnFtnep6v"
      },
      "source": [
        "\n",
        "[Learn the Basics](intro.html) ||\n",
        "[Quickstart](quickstart_tutorial.html) ||\n",
        "[Tensors](tensorqs_tutorial.html) ||\n",
        "[Datasets & DataLoaders](data_tutorial.html) ||\n",
        "[Transforms](transforms_tutorial.html) ||\n",
        "**Build Model** ||\n",
        "[Autograd](autogradqs_tutorial.html) ||\n",
        "[Optimization](optimization_tutorial.html) ||\n",
        "[Save & Load Model](saveloadrun_tutorial.html)\n",
        "\n",
        "# Build the Neural Network\n",
        "\n",
        "Neural networks comprise of layers/modules that perform operations on data.\n",
        "The [torch.nn](https://pytorch.org/docs/stable/nn.html) namespace provides all the building blocks you need to\n",
        "build your own neural network. Every module in PyTorch subclasses the [nn.Module](https://pytorch.org/docs/stable/generated/torch.nn.Module.html).\n",
        "A neural network is a module itself that consists of other modules (layers). This nested structure allows for\n",
        "building and managing complex architectures easily.\n",
        "\n",
        "In the following sections, we'll build a neural network to classify images in the FashionMNIST dataset.\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 6,
      "metadata": {
        "id": "YLHLSvU6ep6w"
      },
      "outputs": [],
      "source": [
        "import os\n",
        "import torch\n",
        "from torch import nn\n",
        "from torch.utils.data import DataLoader\n",
        "from torchvision import datasets, transforms"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xwosOYMQep6x"
      },
      "source": [
        "## Get Device for Training\n",
        "We want to be able to train our model on a hardware accelerator like the GPU,\n",
        "if it is available. Let's check to see if\n",
        "[torch.cuda](https://pytorch.org/docs/stable/notes/cuda.html) is available, else we\n",
        "continue to use the CPU.\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 7,
      "metadata": {
        "id": "jlw-_8tOep6x",
        "outputId": "a437810a-20c1-4e1b-b15e-31429dfab4fb",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Using cuda device\n"
          ]
        }
      ],
      "source": [
        "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
        "print(f\"Using {device} device\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bzLKd7gGep6x"
      },
      "source": [
        "## Define the Class\n",
        "We define our neural network by subclassing ``nn.Module``, and\n",
        "initialize the neural network layers in ``__init__``. Every ``nn.Module`` subclass implements\n",
        "the operations on input data in the ``forward`` method.\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 8,
      "metadata": {
        "id": "Mtj4y6AAep6y"
      },
      "outputs": [],
      "source": [
        "class NeuralNetwork(nn.Module):\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, 512),\n",
        "            nn.ReLU(),\n",
        "            nn.Linear(512, 10),\n",
        "        )\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.flatten(x)\n",
        "        logits = self.linear_relu_stack(x)\n",
        "        return logits"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sVbO2Tujep6y"
      },
      "source": [
        "We create an instance of ``NeuralNetwork``, and move it to the ``device``, and print\n",
        "its structure.\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "Plt0J2OIep6z",
        "outputId": "26ad7351-9b00-454e-9b86-54f490015a6f",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "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=512, bias=True)\n",
            "    (3): ReLU()\n",
            "    (4): Linear(in_features=512, out_features=10, bias=True)\n",
            "  )\n",
            ")\n"
          ]
        }
      ],
      "source": [
        "model = NeuralNetwork().to(device)\n",
        "print(model)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ShroZrumep6z"
      },
      "source": [
        "To use the model, we pass it the input data. This executes the model's ``forward``,\n",
        "along with some [background operations](https://github.com/pytorch/pytorch/blob/270111b7b611d174967ed204776985cefca9c144/torch/nn/modules/module.py#L866).\n",
        "Do not call ``model.forward()`` directly!\n",
        "\n",
        "Calling the model on the input returns a 2-dimensional tensor with dim=0 corresponding to each output of 10 raw predicted values for each class, and dim=1 corresponding to the individual values of each output.  .\n",
        "We get the prediction probabilities by passing it through an instance of the ``nn.Softmax`` module.\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "id": "hD48hBoIep6z",
        "outputId": "a2546ec0-f94b-4c31-f803-78939d137c57",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Predicted class: tensor([9], device='cuda:0')\n"
          ]
        }
      ],
      "source": [
        "X = torch.rand(1, 28, 28, device=device)\n",
        "logits = model(X)\n",
        "pred_probab = nn.Softmax(dim=1)(logits)\n",
        "y_pred = pred_probab.argmax(1)\n",
        "print(f\"Predicted class: {y_pred}\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0GKO4-OIep60"
      },
      "source": [
        "--------------\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Uix73flhep60"
      },
      "source": [
        "## Model Layers\n",
        "\n",
        "Let's break down the layers in the FashionMNIST model. To illustrate it, we\n",
        "will take a sample minibatch of 3 images of size 28x28 and see what happens to it as\n",
        "we pass it through the network.\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "id": "HgMbeunCep60",
        "outputId": "c9203691-1fc7-4324-c94a-dd5987aaa536",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([3, 28, 28])\n"
          ]
        }
      ],
      "source": [
        "input_image = torch.rand(3,28,28)\n",
        "print(input_image.size())"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TuN8IK-Eep60"
      },
      "source": [
        "### nn.Flatten\n",
        "We initialize the [nn.Flatten](https://pytorch.org/docs/stable/generated/torch.nn.Flatten.html)\n",
        "layer to convert each 2D 28x28 image into a contiguous array of 784 pixel values (\n",
        "the minibatch dimension (at dim=0) is maintained).\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "raD_B5fPep60",
        "outputId": "dac4e664-14e7-459d-9b40-1d0359acd992",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([3, 784])\n"
          ]
        }
      ],
      "source": [
        "flatten = nn.Flatten()\n",
        "flat_image = flatten(input_image)\n",
        "print(flat_image.size())"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OR7M-0Oeep60"
      },
      "source": [
        "### nn.Linear\n",
        "The [linear layer](https://pytorch.org/docs/stable/generated/torch.nn.Linear.html)\n",
        "is a module that applies a linear transformation on the input using its stored weights and biases.\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {
        "id": "Vd0WJyFKep60",
        "outputId": "b1d46151-b143-48af-e26a-5885ab311c92",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([3, 21])\n"
          ]
        }
      ],
      "source": [
        "layer1 = nn.Linear(in_features=28*28, out_features=21)\n",
        "hidden1 = layer1(flat_image)\n",
        "print(hidden1.size())"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lj7n4Iymep61"
      },
      "source": [
        "### nn.ReLU\n",
        "Non-linear activations are what create the complex mappings between the model's inputs and outputs.\n",
        "They are applied after linear transformations to introduce *nonlinearity*, helping neural networks\n",
        "learn a wide variety of phenomena.\n",
        "\n",
        "In this model, we use [nn.ReLU](https://pytorch.org/docs/stable/generated/torch.nn.ReLU.html) between our\n",
        "linear layers, but there's other activations to introduce non-linearity in your model.\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 19,
      "metadata": {
        "id": "0TizmZQ6ep61",
        "outputId": "a69c18d3-8feb-4abe-e461-ccc1cf3e9d9f",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Before ReLU: tensor([[-0.2459,  0.1897, -0.1310, -0.3354, -0.1577, -0.4556,  0.6840,  0.3185,\n",
            "          0.3411,  0.2010, -0.1325, -0.1212,  0.2680, -0.2419,  0.0667,  0.1347,\n",
            "          0.3281, -0.8500,  0.1278, -0.1704,  0.5306],\n",
            "        [-0.3858,  0.2442,  0.1080, -0.2229, -0.1401, -0.1724,  0.5045,  0.2530,\n",
            "          0.5246,  0.0579,  0.1191, -0.2515, -0.0708,  0.0901,  0.5621,  0.0665,\n",
            "          0.2788, -0.3786, -0.0512, -0.3569,  0.6405],\n",
            "        [-0.2852, -0.0832, -0.1369, -0.1512, -0.1110, -0.3379,  0.6369,  0.4842,\n",
            "          0.1853,  0.1549, -0.2191,  0.0245,  0.0142,  0.0652,  0.3645, -0.0369,\n",
            "          0.1389, -0.2177, -0.0811, -0.2676,  0.3432]],\n",
            "       grad_fn=<AddmmBackward0>)\n",
            "\n",
            "\n",
            "After ReLU: tensor([[0.0000, 0.1897, 0.0000, 0.0000, 0.0000, 0.0000, 0.6840, 0.3185, 0.3411,\n",
            "         0.2010, 0.0000, 0.0000, 0.2680, 0.0000, 0.0667, 0.1347, 0.3281, 0.0000,\n",
            "         0.1278, 0.0000, 0.5306],\n",
            "        [0.0000, 0.2442, 0.1080, 0.0000, 0.0000, 0.0000, 0.5045, 0.2530, 0.5246,\n",
            "         0.0579, 0.1191, 0.0000, 0.0000, 0.0901, 0.5621, 0.0665, 0.2788, 0.0000,\n",
            "         0.0000, 0.0000, 0.6405],\n",
            "        [0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.6369, 0.4842, 0.1853,\n",
            "         0.1549, 0.0000, 0.0245, 0.0142, 0.0652, 0.3645, 0.0000, 0.1389, 0.0000,\n",
            "         0.0000, 0.0000, 0.3432]], grad_fn=<ReluBackward0>)\n"
          ]
        }
      ],
      "source": [
        "print(f\"Before ReLU: {hidden1}\\n\\n\")\n",
        "hidden1 = nn.ReLU()(hidden1)\n",
        "print(f\"After ReLU: {hidden1}\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HtqH0fmJep61"
      },
      "source": [
        "### nn.Sequential\n",
        "[nn.Sequential](https://pytorch.org/docs/stable/generated/torch.nn.Sequential.html) is an ordered\n",
        "container of modules. The data is passed through all the modules in the same order as defined. You can use\n",
        "sequential containers to put together a quick network like ``seq_modules``.\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "id": "ZNN1PyJ4ep61"
      },
      "outputs": [],
      "source": [
        "seq_modules = nn.Sequential(\n",
        "    flatten,\n",
        "    layer1,\n",
        "    nn.ReLU(),\n",
        "    nn.Linear(20, 10)\n",
        ")\n",
        "input_image = torch.rand(3,28,28)\n",
        "logits = seq_modules(input_image)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mlwzDtWOep61"
      },
      "source": [
        "### nn.Softmax\n",
        "The last linear layer of the neural network returns `logits` - raw values in [-\\infty, \\infty] - which are passed to the\n",
        "[nn.Softmax](https://pytorch.org/docs/stable/generated/torch.nn.Softmax.html) module. The logits are scaled to values\n",
        "[0, 1] representing the model's predicted probabilities for each class. ``dim`` parameter indicates the dimension along\n",
        "which the values must sum to 1.\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "id": "HntZgH2oep61"
      },
      "outputs": [],
      "source": [
        "softmax = nn.Softmax(dim=1)\n",
        "pred_probab = softmax(logits)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "l-N65pa2ep61"
      },
      "source": [
        "## Model Parameters\n",
        "Many layers inside a neural network are *parameterized*, i.e. have associated weights\n",
        "and biases that are optimized during training. Subclassing ``nn.Module`` automatically\n",
        "tracks all fields defined inside your model object, and makes all parameters\n",
        "accessible using your model's ``parameters()`` or ``named_parameters()`` methods.\n",
        "\n",
        "In this example, we iterate over each parameter, and print its size and a preview of its values.\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "id": "GldTr7ULep61",
        "outputId": "3c94bb58-b8b7-4b8d-d324-aad299e6efc9",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model structure: 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=512, bias=True)\n",
            "    (3): ReLU()\n",
            "    (4): Linear(in_features=512, out_features=10, bias=True)\n",
            "  )\n",
            ")\n",
            "\n",
            "\n",
            "Layer: linear_relu_stack.0.weight | Size: torch.Size([512, 784]) | Values : tensor([[-0.0229, -0.0032,  0.0272,  ...,  0.0158, -0.0332, -0.0328],\n",
            "        [-0.0314, -0.0271, -0.0288,  ..., -0.0045, -0.0184, -0.0043]],\n",
            "       device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.0.bias | Size: torch.Size([512]) | Values : tensor([0.0092, 0.0064], device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.2.weight | Size: torch.Size([512, 512]) | Values : tensor([[-0.0014, -0.0249, -0.0044,  ..., -0.0309, -0.0201, -0.0006],\n",
            "        [-0.0419,  0.0149, -0.0374,  ..., -0.0026, -0.0220,  0.0220]],\n",
            "       device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.2.bias | Size: torch.Size([512]) | Values : tensor([-0.0219,  0.0148], device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.4.weight | Size: torch.Size([10, 512]) | Values : tensor([[ 0.0011,  0.0397, -0.0162,  ...,  0.0068,  0.0077,  0.0382],\n",
            "        [-0.0331, -0.0009,  0.0025,  ..., -0.0309, -0.0247, -0.0028]],\n",
            "       device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.4.bias | Size: torch.Size([10]) | Values : tensor([ 0.0094, -0.0225], device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n"
          ]
        }
      ],
      "source": [
        "print(f\"Model structure: {model}\\n\\n\")\n",
        "\n",
        "for name, param in model.named_parameters():\n",
        "    print(f\"Layer: {name} | Size: {param.size()} | Values : {param[:2]} \\n\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xbKhPJ65ep61"
      },
      "source": [
        "--------------\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yApGiospep62"
      },
      "source": [
        "## Further Reading\n",
        "- [torch.nn API](https://pytorch.org/docs/stable/nn.html)\n",
        "\n"
      ]
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.10.4"
    },
    "colab": {
      "provenance": []
    },
    "accelerator": "GPU",
    "gpuClass": "standard"
  },
  "nbformat": 4,
  "nbformat_minor": 0
}