task-2-nn/MT2022151.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "id": "a266UkggTfiC"
      },
      "outputs": [],
      "source": [
        "%matplotlib inline"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "em2_f3ynTfiK"
      },
      "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": 11,
      "metadata": {
        "id": "QTScnJoPTfiO"
      },
      "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": "u7JpE24eTfiQ"
      },
      "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": 12,
      "metadata": {
        "id": "Z9J1SGRpTfiR",
        "outputId": "5acb8323-3cf3-40cd-e52b-0a9bbe935bff",
        "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": "fR_2aHBbTfiS"
      },
      "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": 23,
      "metadata": {
        "id": "aPq30ChrTfiT"
      },
      "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, 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": "y9YjQ6LNTfiU"
      },
      "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": 24,
      "metadata": {
        "id": "MHN8LinnTfiW",
        "outputId": "025fabfc-c861-4a01-8f45-cde5c2958123",
        "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=10, bias=True)\n",
            "  )\n",
            ")\n"
          ]
        }
      ],
      "source": [
        "model = NeuralNetwork().to(device)\n",
        "print(model)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GSipU87XTfiX"
      },
      "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": 15,
      "metadata": {
        "id": "vlli6bT1TfiY",
        "outputId": "8b6a7dc0-026c-4e85-a6ed-9ebe1fb71580",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Predicted class: tensor([1], 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": "czQ2YVoXTfiY"
      },
      "source": [
        "--------------\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fvgUQkXiTfiZ"
      },
      "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": 16,
      "metadata": {
        "id": "oqOOjESCTfiZ",
        "outputId": "ef35eb79-60c1-418f-d667-ed1277d021ad",
        "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": "jaaEHrahTfia"
      },
      "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": 17,
      "metadata": {
        "id": "WLp9KQapTfib",
        "outputId": "900258e9-566b-4a08-a655-2ab0e3aa894f",
        "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": "yqYKVUSjTfib"
      },
      "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": "ao_j4lkSTfic",
        "outputId": "7c0ddb5e-354e-4866-f9a7-951085320f8c",
        "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": "HZUitJd4Tfic"
      },
      "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": "3DrDH-zfTfid",
        "outputId": "16f4fc5b-885a-47ff-9466-0dc5098b9fd2",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Before ReLU: tensor([[ 3.0979e-01,  1.6647e-01,  3.6700e-01, -6.0742e-02, -1.8407e-02,\n",
            "          3.7831e-01,  4.1790e-05,  7.2465e-01,  6.8680e-01,  9.4547e-01,\n",
            "          1.3231e-01,  1.0721e-01,  3.1017e-01, -3.3343e-01,  7.3524e-02,\n",
            "         -5.3752e-02,  5.5183e-02,  2.4888e-01, -7.4022e-01, -6.7520e-03],\n",
            "        [ 9.8758e-01,  8.6845e-02, -8.7820e-02, -4.2574e-01,  8.7517e-02,\n",
            "          4.3747e-01,  3.5882e-02,  9.6696e-01,  2.0986e-01,  5.5270e-01,\n",
            "          2.4116e-03,  2.9967e-01,  2.6855e-01, -2.6544e-01, -2.5122e-04,\n",
            "         -1.3215e-01, -6.6715e-02,  5.8060e-01, -6.0362e-01, -3.1581e-01],\n",
            "        [ 4.9443e-01,  3.4397e-02, -2.7965e-02, -3.5425e-01,  5.2899e-02,\n",
            "          5.8126e-01,  2.5706e-01,  7.7114e-01,  3.1612e-01,  7.9098e-01,\n",
            "          1.8244e-01,  5.9149e-02,  3.0813e-01, -4.0161e-01,  2.8383e-01,\n",
            "          1.3365e-01, -7.4862e-02,  5.6745e-02, -1.5265e-01,  1.2320e-01]],\n",
            "       grad_fn=<AddmmBackward0>)\n",
            "\n",
            "\n",
            "After ReLU: tensor([[3.0979e-01, 1.6647e-01, 3.6700e-01, 0.0000e+00, 0.0000e+00, 3.7831e-01,\n",
            "         4.1790e-05, 7.2465e-01, 6.8680e-01, 9.4547e-01, 1.3231e-01, 1.0721e-01,\n",
            "         3.1017e-01, 0.0000e+00, 7.3524e-02, 0.0000e+00, 5.5183e-02, 2.4888e-01,\n",
            "         0.0000e+00, 0.0000e+00],\n",
            "        [9.8758e-01, 8.6845e-02, 0.0000e+00, 0.0000e+00, 8.7517e-02, 4.3747e-01,\n",
            "         3.5882e-02, 9.6696e-01, 2.0986e-01, 5.5270e-01, 2.4116e-03, 2.9967e-01,\n",
            "         2.6855e-01, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 5.8060e-01,\n",
            "         0.0000e+00, 0.0000e+00],\n",
            "        [4.9443e-01, 3.4397e-02, 0.0000e+00, 0.0000e+00, 5.2899e-02, 5.8126e-01,\n",
            "         2.5706e-01, 7.7114e-01, 3.1612e-01, 7.9098e-01, 1.8244e-01, 5.9149e-02,\n",
            "         3.0813e-01, 0.0000e+00, 2.8383e-01, 1.3365e-01, 0.0000e+00, 5.6745e-02,\n",
            "         0.0000e+00, 1.2320e-01]], 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": "LSXeesLETfid"
      },
      "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": 20,
      "metadata": {
        "id": "tMgeFhwfTfid"
      },
      "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": "JImev-LoTfie"
      },
      "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": 21,
      "metadata": {
        "id": "1QDN06epTfie"
      },
      "outputs": [],
      "source": [
        "softmax = nn.Softmax(dim=1)\n",
        "pred_probab = softmax(logits)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MQ7WRCCaTfif"
      },
      "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": 22,
      "metadata": {
        "id": "pl1FpPItTfif",
        "outputId": "a6954da0-d45f-4dfc-c92b-389c9a6d8c99",
        "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([[-0.0031, -0.0039,  0.0338,  ..., -0.0132, -0.0356,  0.0012],\n",
            "        [ 0.0042,  0.0019, -0.0205,  ..., -0.0350, -0.0154, -0.0293]],\n",
            "       device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.0.bias | Size: torch.Size([512]) | Values : tensor([-0.0055, -0.0036], device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.2.weight | Size: torch.Size([512, 512]) | Values : tensor([[ 1.1152e-02, -2.2646e-02, -2.5694e-02,  ..., -4.2210e-02,\n",
            "          3.6723e-02, -1.9322e-02],\n",
            "        [ 1.6634e-02, -4.3681e-02, -3.2206e-02,  ..., -3.9255e-02,\n",
            "          9.5431e-05,  2.7839e-02]], device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.2.bias | Size: torch.Size([512]) | Values : tensor([ 0.0248, -0.0112], device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.4.weight | Size: torch.Size([10, 512]) | Values : tensor([[-0.0081,  0.0308,  0.0128,  ..., -0.0349, -0.0404,  0.0347],\n",
            "        [ 0.0295,  0.0230,  0.0263,  ..., -0.0025,  0.0293,  0.0380]],\n",
            "       device='cuda:0', grad_fn=<SliceBackward0>) \n",
            "\n",
            "Layer: linear_relu_stack.4.bias | Size: torch.Size([10]) | Values : tensor([0.0069, 0.0235], 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": "9yCplQIeTfif"
      },
      "source": [
        "--------------\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "j_i0WLVhTfig"
      },
      "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
}