task-2-nn/IMT2020102.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 30,
      "metadata": {
        "id": "nWyY3XBaevZU"
      },
      "outputs": [],
      "source": [
        "%matplotlib inline"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-m0M9s0ZevZY"
      },
      "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": 31,
      "metadata": {
        "id": "XzbQ5Y2bevZa"
      },
      "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": "AvVQsuQUevZb"
      },
      "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": 32,
      "metadata": {
        "id": "KaPjSCaYevZb",
        "outputId": "6436310b-dcd8-4131-e8fc-dbfef63406cb",
        "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": "i4ISj00FevZb"
      },
      "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": 33,
      "metadata": {
        "id": "eD_IcIKPevZc"
      },
      "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, 1024),\n",
        "            nn.ReLU(),\n",
        "            nn.Linear(1024, 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": "zAwM4yo_evZc"
      },
      "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": 34,
      "metadata": {
        "id": "nM_fcjQ4evZd",
        "outputId": "ead5bbd7-0b84-401b-c52d-d167f724895e",
        "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=1024, bias=True)\n",
            "    (1): ReLU()\n",
            "    (2): Linear(in_features=1024, out_features=512, bias=True)\n",
            "    (3): ReLU()\n",
            "    (4): Linear(in_features=512, out_features=512, bias=True)\n",
            "    (5): ReLU()\n",
            "    (6): 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": "b-qr4bD7evZe"
      },
      "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": 35,
      "metadata": {
        "id": "ALMK_UPcevZe",
        "outputId": "1f6c6124-bfc3-422c-9469-d755452342e3",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Predicted class: tensor([2], 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": "L322gkxcevZf"
      },
      "source": [
        "--------------\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "er4MSramevZf"
      },
      "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": 36,
      "metadata": {
        "id": "d8rl6CY4evZf",
        "outputId": "407f08a1-8669-4283-8e72-d450fd5543a6",
        "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": "XaO2yHRvevZf"
      },
      "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": 37,
      "metadata": {
        "id": "tuE_quTeevZg",
        "outputId": "9238586a-6c2e-4624-dc9c-5ab455627868",
        "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": "8bW8j008evZg"
      },
      "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": 38,
      "metadata": {
        "id": "dF3QywBlevZg",
        "outputId": "9f759a8e-756e-45e9-a64c-669bb5931a6e",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "torch.Size([3, 20])\n"
          ]
        }
      ],
      "source": [
        "layer1 = nn.Linear(in_features=28*28, out_features=20)\n",
        "hidden1 = layer1(flat_image)\n",
        "print(hidden1.size())"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "o8Qhka-4evZg"
      },
      "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": 39,
      "metadata": {
        "id": "-GgcC-wUevZh",
        "outputId": "de4f647c-a39e-428e-a556-8741b4cbcfbd",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Before ReLU: tensor([[-0.3244,  0.3188, -0.4965,  0.1404,  0.1048,  0.0751, -0.5065,  0.4562,\n",
            "          0.0717, -0.0532,  0.0795, -0.0190,  0.3622, -0.2354,  0.6490, -0.0780,\n",
            "          0.5821,  0.6012,  0.2950, -0.3647],\n",
            "        [-0.4132,  0.2370, -0.4907, -0.2600, -0.1971, -0.3062, -0.4167,  0.4202,\n",
            "         -0.2582,  0.1843,  0.1660, -0.1644, -0.1411,  0.0249,  0.4024, -0.1912,\n",
            "          0.6233,  0.3782,  0.4203, -0.0548],\n",
            "        [-0.3270,  0.1316, -0.3147,  0.4877, -0.1004,  0.0665, -0.3365,  0.0637,\n",
            "          0.1413, -0.1947,  0.2984, -0.0828, -0.1510, -0.3157,  0.4676, -0.0472,\n",
            "          0.8394,  0.8564,  0.1126, -0.1385]], grad_fn=<AddmmBackward0>)\n",
            "\n",
            "\n",
            "After ReLU: tensor([[0.0000, 0.3188, 0.0000, 0.1404, 0.1048, 0.0751, 0.0000, 0.4562, 0.0717,\n",
            "         0.0000, 0.0795, 0.0000, 0.3622, 0.0000, 0.6490, 0.0000, 0.5821, 0.6012,\n",
            "         0.2950, 0.0000],\n",
            "        [0.0000, 0.2370, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.4202, 0.0000,\n",
            "         0.1843, 0.1660, 0.0000, 0.0000, 0.0249, 0.4024, 0.0000, 0.6233, 0.3782,\n",
            "         0.4203, 0.0000],\n",
            "        [0.0000, 0.1316, 0.0000, 0.4877, 0.0000, 0.0665, 0.0000, 0.0637, 0.1413,\n",
            "         0.0000, 0.2984, 0.0000, 0.0000, 0.0000, 0.4676, 0.0000, 0.8394, 0.8564,\n",
            "         0.1126, 0.0000]], 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": "7LPaL3VJevZh"
      },
      "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": 40,
      "metadata": {
        "id": "kXKBq6WrevZh"
      },
      "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": "r3ciGmFXevZh"
      },
      "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": 41,
      "metadata": {
        "id": "s3zcoQXqevZh"
      },
      "outputs": [],
      "source": [
        "softmax = nn.Softmax(dim=1)\n",
        "pred_probab = softmax(logits)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VD8PYixCevZi"
      },
      "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": 42,
      "metadata": {
        "id": "RevE41myevZi",
        "outputId": "c37d6c31-ef32-4c43-c9b5-a6ae18cdc051",
        "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=1024, bias=True)\n",
            "    (1): ReLU()\n",
            "    (2): Linear(in_features=1024, out_features=512, bias=True)\n",
            "    (3): ReLU()\n",
            "    (4): Linear(in_features=512, out_features=512, bias=True)\n",
            "    (5): ReLU()\n",
            "    (6): Linear(in_features=512, out_features=10, bias=True)\n",
            "  )\n",
            ")\n",
            "\n",
            "\n",
            "Layer: linear_relu_stack.0.weight | Size: torch.Size([1024, 784]) | Values : tensor([[ 0.0215,  0.0109,  0.0141,  ...,  0.0353,  0.0015, -0.0259],\n",
            "        [ 0.0082, -0.0227,  0.0263,  ..., -0.0009, -0.0273,  0.0295]],\n",
            "       device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.0.bias | Size: torch.Size([1024]) | Values : tensor([0.0337, 0.0246], device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.2.weight | Size: torch.Size([512, 1024]) | Values : tensor([[-0.0245,  0.0014,  0.0239,  ...,  0.0293, -0.0280,  0.0104],\n",
            "        [ 0.0019, -0.0089, -0.0158,  ...,  0.0156,  0.0139, -0.0231]],\n",
            "       device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.2.bias | Size: torch.Size([512]) | Values : tensor([ 0.0128, -0.0233], device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.4.weight | Size: torch.Size([512, 512]) | Values : tensor([[-0.0267,  0.0094, -0.0279,  ..., -0.0084,  0.0153,  0.0191],\n",
            "        [-0.0051, -0.0378,  0.0154,  ...,  0.0187, -0.0098,  0.0170]],\n",
            "       device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.4.bias | Size: torch.Size([512]) | Values : tensor([-0.0139,  0.0160], device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.6.weight | Size: torch.Size([10, 512]) | Values : tensor([[ 0.0042, -0.0002, -0.0324,  ...,  0.0059, -0.0337, -0.0097],\n",
            "        [ 0.0181,  0.0210, -0.0339,  ..., -0.0057,  0.0181,  0.0378]],\n",
            "       device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.6.bias | Size: torch.Size([10]) | Values : tensor([-0.0252, -0.0284], 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": "RtwrFCTzevZi"
      },
      "source": [
        "--------------\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pUj0nRJ9evZi"
      },
      "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
}