3D_MNIST_Classifier_with_S4TF.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "3D MNIST Classifier with S4TF.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "machine_shape": "hm",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "swift",
      "display_name": "Swift"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/lukereichold/3D-MNIST-S4TF/blob/master/3D_MNIST_Classifier_with_S4TF.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sh3DhAz5BCuv",
        "colab_type": "text"
      },
      "source": [
        "## ⚙️ Install Dependencies"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qS0u3oWyrHBy",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "1e8f5b23-0fa5-4357-bf6b-84b816e44fa2"
      },
      "source": [
        "import TensorFlow\n",
        "import Foundation\n",
        "\n",
        "#if canImport(PythonKit)\n",
        "    import PythonKit\n",
        "#else\n",
        "    import Python\n",
        "#endif\n",
        "print(Python.version)"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "3.6.9 (default, Nov  7 2019, 10:44:02) \r\n",
            "[GCC 8.3.0]\r\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O-zbuX8hAr7W",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "let h5py = Python.import(\"h5py\")\n",
        "let np = Python.import(\"numpy\")\n",
        "let os = Python.import(\"os\")\n",
        "let plt = Python.import(\"matplotlib.pyplot\")\n",
        "let cm = Python.import(\"matplotlib.cm\")"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8WrzILOqrHGR",
        "colab_type": "code",
        "outputId": "05e74dbb-d1ce-4d28-d3f6-9f2dd2abc5cd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "%include \"EnableIPythonDisplay.swift\"\n",
        "IPythonDisplay.shell.enable_matplotlib(\"inline\")"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "('inline', 'module://ipykernel.pylab.backend_inline')\n"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mUOSyO8nk3WZ",
        "colab_type": "text"
      },
      "source": [
        "## ⬇ Fetch Dataset\n",
        "3D Point Vectors fetched initially from https://www.kaggle.com/daavoo/3d-mnist/data (29 Mb)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "d_FVkxV98SYd",
        "colab_type": "code",
        "outputId": "2c26950a-0ab1-415d-ba17-baa3b534c55b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 462
        }
      },
      "source": [
        "os.system(\"wget https://www.dropbox.com/s/mvre9rojnjp2a36/full_dataset_vectors.h5\")"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2020-03-27 19:40:55--  https://www.dropbox.com/s/mvre9rojnjp2a36/full_dataset_vectors.h5\r\n",
            "Resolving www.dropbox.com (www.dropbox.com)... 162.125.82.1, 2620:100:6032:1::a27d:5201\r\n",
            "Connecting to www.dropbox.com (www.dropbox.com)|162.125.82.1|:443... connected.\n",
            "HTTP request sent, awaiting response... 301 Moved Permanently\n",
            "Location: /s/raw/mvre9rojnjp2a36/full_dataset_vectors.h5 [following]\n",
            "--2020-03-27 19:40:56--  https://www.dropbox.com/s/raw/mvre9rojnjp2a36/full_dataset_vectors.h5\n",
            "Reusing existing connection to www.dropbox.com:443.\n",
            "HTTP request sent, awaiting response... 302 Found\n",
            "Location: https://uc7c21340e97f56af25c34e47592.dl.dropboxusercontent.com/cd/0/inline/A0slMNm65JRjGTBN0j5isS6CqZRUUTttHAOzsBf5cxrM1x22EX79WWbJ68PT1G29w6DgEUejbPgOKywNv5pB32xvlIqYm-9t4EJZjglFkfJoJxdmlpoBVENZxuuYOlfKF8w/file# [following]\n",
            "--2020-03-27 19:40:56--  https://uc7c21340e97f56af25c34e47592.dl.dropboxusercontent.com/cd/0/inline/A0slMNm65JRjGTBN0j5isS6CqZRUUTttHAOzsBf5cxrM1x22EX79WWbJ68PT1G29w6DgEUejbPgOKywNv5pB32xvlIqYm-9t4EJZjglFkfJoJxdmlpoBVENZxuuYOlfKF8w/file\n",
            "Resolving uc7c21340e97f56af25c34e47592.dl.dropboxusercontent.com (uc7c21340e97f56af25c34e47592.dl.dropboxusercontent.com)... 162.125.82.6, 2620:100:6032:6::a27d:5206\n",
            "Connecting to uc7c21340e97f56af25c34e47592.dl.dropboxusercontent.com (uc7c21340e97f56af25c34e47592.dl.dropboxusercontent.com)|162.125.82.6|:443... connected.\n",
            "HTTP request sent, awaiting response... 302 FOUND\n",
            "Location: /cd/0/inline2/A0sJCg-PdSnslbsr6GWHt0VvtZZvJRnLc4HgUaLQnZ5IQ0LHrhWwi-aQ0z25M96JP1bAK-R_qDl-E4m5jXW7P-E-V3VquPatq6QpXccpr-OuWe4ZZOwE9NMK__0S1mlx72kxP3tM3d4p0LPkwv8AupMgv583L0ds2voPAjWSXJ6Fy4xhu86DKMBOehb0rP8nGQDnAcw9jdTAI-OwqIMmLJaKJ7LHMkRvulPcWQT0-R4xRx_-C4qX8gadC6vahklLz5xWmL93fFkNjwMEDb5lI09JDQEaxjkLrJU7hmLpvXMJHPxp-Wvci6-w4lXUK5Q1HO_ncco4PhbTRauBoDvu5y01C7tBR-hzgeqDuOTke6YVLA/file [following]\n",
            "--2020-03-27 19:40:57--  https://uc7c21340e97f56af25c34e47592.dl.dropboxusercontent.com/cd/0/inline2/A0sJCg-PdSnslbsr6GWHt0VvtZZvJRnLc4HgUaLQnZ5IQ0LHrhWwi-aQ0z25M96JP1bAK-R_qDl-E4m5jXW7P-E-V3VquPatq6QpXccpr-OuWe4ZZOwE9NMK__0S1mlx72kxP3tM3d4p0LPkwv8AupMgv583L0ds2voPAjWSXJ6Fy4xhu86DKMBOehb0rP8nGQDnAcw9jdTAI-OwqIMmLJaKJ7LHMkRvulPcWQT0-R4xRx_-C4qX8gadC6vahklLz5xWmL93fFkNjwMEDb5lI09JDQEaxjkLrJU7hmLpvXMJHPxp-Wvci6-w4lXUK5Q1HO_ncco4PhbTRauBoDvu5y01C7tBR-hzgeqDuOTke6YVLA/file\n",
            "Reusing existing connection to uc7c21340e97f56af25c34e47592.dl.dropboxusercontent.com:443.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 30582936 (29M) [application/octet-stream]\n",
            "Saving to: ‘full_dataset_vectors.h5’\n",
            "\n",
            "full_dataset_vector 100%[===================>]  29.17M  4.10MB/s    in 8.0s    \n",
            "\n",
            "2020-03-27 19:41:06 (3.65 MB/s) - ‘full_dataset_vectors.h5’ saved [30582936/30582936]\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0\n"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VZeMCfywmBV4",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "64d194c5-ffe3-4259-977c-326c235faaf7"
      },
      "source": [
        "let dataset = h5py.File(\"full_dataset_vectors.h5\", \"r\")\n",
        "\n",
        "func floatTensor(from data: PythonObject) -> Tensor<Float> {\n",
        "  return Tensor<Float>(numpy: data.value.astype(np.float32))!\n",
        "}\n",
        "\n",
        "func intTensor(from data: PythonObject) -> Tensor<Int32> {\n",
        "  return Tensor<Int32>(numpy: data.value.astype(np.int32))!\n",
        "}\n",
        "\n",
        "var trainingFeatures = floatTensor(from: dataset[\"X_train\"])\n",
        "var trainingLabels = intTensor(from: dataset[\"y_train\"])\n",
        "\n",
        "var testFeatures = floatTensor(from: dataset[\"X_test\"])\n",
        "var testLabels = intTensor(from: dataset[\"y_test\"])"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "sys:1: H5pyDeprecationWarning: dataset.value has been deprecated. Use dataset[()] instead.\r\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ESxkVgimlQXr",
        "colab_type": "text"
      },
      "source": [
        "## 📐 Verify Initial Dataset Shapes"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pFKdWnHtBmNb",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "assert(trainingFeatures.shape == TensorShape([10_000, 4096]))\n",
        "assert(trainingLabels.shape == TensorShape([10_000]))\n",
        "\n",
        "assert(testFeatures.shape == TensorShape([2_000, 4096]))\n",
        "assert(testLabels.shape == TensorShape([2_000]))\n",
        "\n",
        "let trainingSize = trainingFeatures.shape[0]\n",
        "let testSize = testFeatures.shape[0]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0HAql2lRJgBP",
        "colab_type": "text"
      },
      "source": [
        "## 🌈 Give Each Point a Color\n",
        "\n",
        "Each sample represents an \"MNIST digit\" in three-dimensional space (x, y, z coordinates), where each  of the 4096 possible points has a color. "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xsuzDT6kxoPv",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "var coloredTrainingSamples = Tensor<Float>(ones: TensorShape([10_000, 4096, 3]))\n",
        "var coloredTestSamples = Tensor<Float>(ones: TensorShape([2_000, 4096, 3]))\n",
        "\n",
        "let trainingNp = trainingFeatures.makeNumpyArray()\n",
        "let testNp = testFeatures.makeNumpyArray()\n",
        "let scalarMap = cm.ScalarMappable(cmap: \"Oranges\")\n",
        "\n",
        "func addColorDimension(scalars: PythonObject) -> Tensor<Float> {\n",
        "  let rgba = scalarMap.to_rgba(scalars).astype(np.float32)\n",
        "  let colorValues = Tensor<Float>(numpy: rgba)!\n",
        "  return colorValues[TensorRange.ellipsis, ..<3]\n",
        "}\n",
        "\n",
        "for i in 0 ..< trainingSize {\n",
        "  coloredTrainingSamples[i] = addColorDimension(scalars: trainingNp[i])\n",
        "}\n",
        "\n",
        "for i in 0 ..< testSize {\n",
        "  coloredTestSamples[i] = addColorDimension(scalars: testNp[i])\n",
        "}"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qnMhtvftNTxB",
        "colab_type": "text"
      },
      "source": [
        "## 🔲 Expand Points to Three Dimensions\n",
        "\n",
        "Reshape each sample to be represented as a rank-4 tensor of shape `(16, 16, 16, 3)`."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "F0aL-YkN2Ztb",
        "colab_type": "code",
        "outputId": "90e00b6d-81a5-4b74-af21-37b5f0514917",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "coloredTrainingSamples = coloredTrainingSamples.reshaped(to: TensorShape(trainingSize, 16, 16, 16, 3))\n",
        "coloredTestSamples = coloredTestSamples.reshaped(to: TensorShape(testSize, 16, 16, 16, 3))\n",
        "\n",
        "let coloredTrainingSamplesNp = coloredTrainingSamples.makeNumpyArray()\n",
        "let trainingLabelsNp = trainingLabels.makeNumpyArray()\n",
        "\n",
        "print(coloredTrainingSamples.shape)\n",
        "print(coloredTestSamples.shape)"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[10000, 16, 16, 16, 3]\r\n",
            "[2000, 16, 16, 16, 3]\r\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HOcXqn4LrREg",
        "colab_type": "text"
      },
      "source": [
        "## 📚 Defining Our Network"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oJeeJY9TrQ3L",
        "colab_type": "code",
        "cellView": "both",
        "colab": {}
      },
      "source": [
        "let frame_height = 16, frame_width = 16, frame_depth = 16\n",
        "let RGB_channels = 3\n",
        "let classCount = 10\n",
        "\n",
        "let input_shape = (frame_depth, frame_height, frame_width, RGB_channels)\n",
        "\n",
        "struct Classifier: Layer {\n",
        "    \n",
        "    typealias Input = Tensor<Float>\n",
        "    typealias Output = Tensor<Float>\n",
        "    \n",
        "    var conv1 = Conv3D<Float>(filterShape: (3, 3, 3, RGB_channels, 8), padding: .same, activation: relu)\n",
        "    var conv2 = Conv3D<Float>(filterShape: (3, 3, 3, 8, 16), padding: .same, activation: relu)\n",
        "    var pool = MaxPool3D<Float>(poolSize: (2, 2, 2), strides: (2, 2, 2))\n",
        "\n",
        "    var conv3 = Conv3D<Float>(filterShape: (3, 3, 3, 16, 32), padding: .same, activation: relu)\n",
        "    var conv4 = Conv3D<Float>(filterShape: (3, 3, 3, 32, 64), padding: .same, activation: relu)\n",
        "\n",
        "    var batchNorm = BatchNorm<Float>(featureCount: 64)\n",
        "    var flatten = Flatten<Float>()\n",
        "    \n",
        "    var dense1 = Dense<Float>(inputSize: 4096, outputSize: 4096, activation: relu)\n",
        "    var dense2 = Dense<Float>(inputSize: 4096, outputSize: 1024, activation: relu)\n",
        "\n",
        "    var dropout25 = Dropout<Float>(probability: 0.25)\n",
        "    var dropout50 = Dropout<Float>(probability: 0.5)\n",
        "  \n",
        "    var output = Dense<Float>(inputSize: 1024, outputSize: classCount, activation: softmax)\n",
        "\n",
        "    @differentiable\n",
        "    func callAsFunction(_ input: Input) -> Output {\n",
        "      return input\n",
        "      .sequenced(through: conv1, conv2, pool)\n",
        "      .sequenced(through: conv3, conv4, batchNorm, pool)\n",
        "      .sequenced(through: dropout25, flatten, dense1, dropout50)\n",
        "      .sequenced(through: dense2, dropout50, output)\n",
        "    }\n",
        "}"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UmLYuKqzrTpF",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "var model = Classifier()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RbSJPaWAqjIc",
        "colab_type": "code",
        "outputId": "093d1b32-ae10-45c9-d1d5-b9c8e3caf6fe",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "let dummy = Tensor<Float>(randomNormal: TensorShape(1, 16, 16, 16, 3))\n",
        "let eval = model(dummy)\n",
        "\n",
        "assert(eval.shape == TensorShape([1, 10]))\n",
        "print(eval.shape)"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[1, 10]\r\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VOeJ8yHOrYae",
        "colab_type": "text"
      },
      "source": [
        "## 📈 Training"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t0skMdiQqre6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "let batchSize = 100\n",
        "let epochs = 50"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AjJvdVFLrmhk",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "let optimizer = Adam(for: model, learningRate: 1e-5, decay: 1e-6)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N0DvT7Flrduz",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "var trainAccHistory = np.zeros(epochs)\n",
        "var valAccHistory = np.zeros(epochs)\n",
        "\n",
        "var trainLossHistory = np.zeros(epochs)\n",
        "var valLossHistory = np.zeros(epochs)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-oyj71strm-i",
        "colab_type": "code",
        "outputId": "64d50318-7fb7-4848-e35b-8e9f9581db44",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 867
        }
      },
      "source": [
        "for epoch in 0 ..< epochs {\n",
        "\n",
        "    Context.local.learningPhase = .training\n",
        "\n",
        "    // Shuffle samples\n",
        "    let shuffledIndices = np.random.permutation(trainingSize)\n",
        "    let shuffledSamples = coloredTrainingSamplesNp[shuffledIndices]\n",
        "    let shuffledLabels = trainingLabelsNp[shuffledIndices]\n",
        "\n",
        "    // Loop over each batch of samples\n",
        "    for batchStart in stride(from: 0, to: trainingSize, by: batchSize) {\n",
        "      let batchRange = batchStart ..< batchStart + batchSize\n",
        "      let labels = Tensor<Int32>(numpy: shuffledLabels[batchRange])!\n",
        "      let samples = Tensor<Float>(numpy: shuffledSamples[batchRange])!\n",
        "\n",
        "      let (_, gradients) = valueWithGradient(at: model) { model -> Tensor<Float> in\n",
        "          let logits = model(samples)\n",
        "          return softmaxCrossEntropy(logits: logits, labels: labels)\n",
        "      }\n",
        "      optimizer.update(&model, along: gradients)\n",
        "    }\n",
        "\n",
        "    // Evaluate model\n",
        "    Context.local.learningPhase = .inference\n",
        "\n",
        "    var correctTrainGuessCount = 0\n",
        "    var totalTrainGuessCount = 0\n",
        "\n",
        "    for batchStart in stride(from: 0, to: trainingSize, by: batchSize) {\n",
        "      let batchRange = batchStart ..< batchStart + batchSize\n",
        "      let labels = Tensor<Int32>(numpy: shuffledLabels[batchRange])!\n",
        "      let samples = Tensor<Float>(numpy: shuffledSamples[batchRange])!\n",
        "\n",
        "      let logits = model(samples)\n",
        "\n",
        "      // accuracy\n",
        "      let correctPredictions = logits.argmax(squeezingAxis: 1) .== labels\n",
        "      correctTrainGuessCount += Int(Tensor<Int32>(correctPredictions).sum().scalarized())\n",
        "\n",
        "      // loss\n",
        "      trainLossHistory[epoch] += PythonObject(softmaxCrossEntropy(logits: logits, labels: labels).scalarized())\n",
        "    }\n",
        "\n",
        "    let trainAcc = Float(correctTrainGuessCount) / Float(trainingSize)\n",
        "    trainAccHistory[epoch] = PythonObject(trainAcc)\n",
        "\n",
        "    var correctValGuessCount = 0\n",
        "    var totalValGuessCount = 0\n",
        "\n",
        "    for batchStart in stride(from: 0, to: testSize, by: batchSize) {\n",
        "        let batchRange = batchStart ..< batchStart + batchSize\n",
        "        let labels = testLabels[batchRange]\n",
        "        let samples = coloredTestSamples[batchRange]\n",
        "\n",
        "        let logits = model(samples)\n",
        "\n",
        "        // accuracy\n",
        "        let correctPredictions = logits.argmax(squeezingAxis: 1) .== labels\n",
        "        correctValGuessCount += Int(Tensor<Int32>(correctPredictions).sum().scalarized())\n",
        "\n",
        "        // loss\n",
        "        valLossHistory[epoch] += PythonObject(softmaxCrossEntropy(logits: logits, labels: labels).scalarized())\n",
        "    }\n",
        "\n",
        "    let valAcc = Float(correctValGuessCount) / Float(testSize)\n",
        "    valAccHistory[epoch] = PythonObject(valAcc)\n",
        "    \n",
        "    print(\"\\(epoch + 1) | Training accuracy: \\(trainAcc) | Validation accuracy: \\(valAcc)\")\n",
        "}"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1 | Training accuracy: 0.1573 | Validation accuracy: 0.161\n",
            "2 | Training accuracy: 0.2549 | Validation accuracy: 0.265\n",
            "3 | Training accuracy: 0.3134 | Validation accuracy: 0.317\n",
            "4 | Training accuracy: 0.3721 | Validation accuracy: 0.369\n",
            "5 | Training accuracy: 0.4083 | Validation accuracy: 0.398\n",
            "6 | Training accuracy: 0.4471 | Validation accuracy: 0.4485\n",
            "7 | Training accuracy: 0.4813 | Validation accuracy: 0.4725\n",
            "8 | Training accuracy: 0.4936 | Validation accuracy: 0.49\n",
            "9 | Training accuracy: 0.5097 | Validation accuracy: 0.504\n",
            "10 | Training accuracy: 0.5237 | Validation accuracy: 0.518\n",
            "11 | Training accuracy: 0.5399 | Validation accuracy: 0.5285\n",
            "12 | Training accuracy: 0.5605 | Validation accuracy: 0.551\n",
            "13 | Training accuracy: 0.575 | Validation accuracy: 0.5615\n",
            "14 | Training accuracy: 0.5822 | Validation accuracy: 0.568\n",
            "15 | Training accuracy: 0.5856 | Validation accuracy: 0.571\n",
            "16 | Training accuracy: 0.5918 | Validation accuracy: 0.5765\n",
            "17 | Training accuracy: 0.6033 | Validation accuracy: 0.589\n",
            "18 | Training accuracy: 0.6064 | Validation accuracy: 0.5865\n",
            "19 | Training accuracy: 0.6073 | Validation accuracy: 0.585\n",
            "20 | Training accuracy: 0.6186 | Validation accuracy: 0.6025\n",
            "21 | Training accuracy: 0.6216 | Validation accuracy: 0.608\n",
            "22 | Training accuracy: 0.6234 | Validation accuracy: 0.609\n",
            "23 | Training accuracy: 0.6276 | Validation accuracy: 0.612\n",
            "24 | Training accuracy: 0.6322 | Validation accuracy: 0.61\n",
            "25 | Training accuracy: 0.6377 | Validation accuracy: 0.615\n",
            "26 | Training accuracy: 0.6372 | Validation accuracy: 0.6165\n",
            "27 | Training accuracy: 0.6497 | Validation accuracy: 0.625\n",
            "28 | Training accuracy: 0.6526 | Validation accuracy: 0.635\n",
            "29 | Training accuracy: 0.6488 | Validation accuracy: 0.628\n",
            "30 | Training accuracy: 0.6515 | Validation accuracy: 0.631\n",
            "31 | Training accuracy: 0.66 | Validation accuracy: 0.633\n",
            "32 | Training accuracy: 0.6611 | Validation accuracy: 0.634\n",
            "33 | Training accuracy: 0.6649 | Validation accuracy: 0.637\n",
            "34 | Training accuracy: 0.6676 | Validation accuracy: 0.639\n",
            "35 | Training accuracy: 0.674 | Validation accuracy: 0.642\n",
            "36 | Training accuracy: 0.6748 | Validation accuracy: 0.64\n",
            "37 | Training accuracy: 0.6785 | Validation accuracy: 0.6395\n",
            "38 | Training accuracy: 0.6786 | Validation accuracy: 0.6425\n",
            "39 | Training accuracy: 0.6793 | Validation accuracy: 0.6455\n",
            "40 | Training accuracy: 0.6852 | Validation accuracy: 0.643\n",
            "41 | Training accuracy: 0.6838 | Validation accuracy: 0.6475\n",
            "42 | Training accuracy: 0.6866 | Validation accuracy: 0.6425\n",
            "43 | Training accuracy: 0.6931 | Validation accuracy: 0.651\n",
            "44 | Training accuracy: 0.6961 | Validation accuracy: 0.654\n",
            "45 | Training accuracy: 0.6998 | Validation accuracy: 0.6565\n",
            "46 | Training accuracy: 0.7012 | Validation accuracy: 0.6585\n",
            "47 | Training accuracy: 0.7038 | Validation accuracy: 0.656\n",
            "48 | Training accuracy: 0.707 | Validation accuracy: 0.665\n",
            "49 | Training accuracy: 0.7079 | Validation accuracy: 0.6685\n",
            "50 | Training accuracy: 0.7123 | Validation accuracy: 0.6635\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7yhLGuCxre3N",
        "colab_type": "text"
      },
      "source": [
        "## 🕝 Inspecting training history"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EagDUPyrrftQ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 312
        },
        "outputId": "ee3d10a7-1b10-48ba-d90c-d98e971d31c1"
      },
      "source": [
        "plt.plot(trainAccHistory, label: \"Training Accuracy\")\n",
        "plt.plot(valAccHistory, label: \"Validation Accuracy\")\n",
        "plt.xlabel(\"Number of epochs\")\n",
        "plt.legend()\n",
        "plt.title(\"Accuracy\")\n",
        "plt.show()"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAEWCAYAAAB2X2wCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nUkp5CQ10pZTyEhroSinlJf4/paZjNYFmeCQAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "None\n"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "03rkXnMLrhQE",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 312
        },
        "outputId": "90c46f8b-d77e-41a7-dc1e-1116f7bb260b"
      },
      "source": [
        "plt.plot(trainLossHistory, label: \"Training Set\")\n",
        "plt.plot(valLossHistory, label: \"Validation Set\")\n",
        "plt.xlabel(\"Number of epochs\")\n",
        "plt.legend()\n",
        "plt.title(\"Loss\")\n",
        "plt.show()"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEWCAYAAACdaNcBAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nyV1EJIKU3EVEIkjJXUQkgpTcRUQi6P8DapbKcIR9110AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "None\n"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 17
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PoYNXefQrn0G",
        "colab_type": "text"
      },
      "source": [
        "## 🌐 Making Inferences\n",
        "\n",
        "With our model now trained, we can make inferences to predict the class labels of samples outside the training set."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fheOkQ5-Vv6n",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "1350b385-6749-44a3-b303-39f6e072067e"
      },
      "source": [
        "let randomIndex = Int.random(in: 0 ..< testSize)\n",
        "var randomSample = coloredTestSamples[randomIndex].reshaped(to: TensorShape(1, 16, 16, 16, 3))\n",
        "\n",
        "let label = testLabels[randomIndex]\n",
        "\n",
        "print(\"Predicted: \\(model(randomSample).argmax()), Actual: \\(label)\")"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Predicted: 3, Actual: 3\r\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}