alexnet/AlexNet_for_CIFAR-10_with_Pylearn2_Keras_LRN.ipynb

{
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": null
      },
      "outputs": [],
      "source": [
        "# Copyright 2022 Google LLC\n",
        "#\n",
        "# Licensed under the Apache License, Version 2.0 (the \"License\");\n",
        "# you may not use this file except in compliance with the License.\n",
        "# You may obtain a copy of the License at\n",
        "#\n",
        "#      http://www.apache.org/licenses/LICENSE-2.0\n",
        "#\n",
        "# Unless required by applicable law or agreed to in writing, software\n",
        "# distributed under the License is distributed on an \"AS IS\" BASIS,\n",
        "# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
        "# See the License for the specific language governing permissions and\n",
        "# limitations under the License."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": null
      },
      "source": [
        "[![View on GitHub][github-badge]][github-cifar10-pylearn2-lrn] [![Open In Colab][colab-badge]][colab-cifar10-pylearn2-lrn] [![Open in Binder][binder-badge]][binder-cifar10-pylearn2-lrn]\n",
        "\n",
        "[github-badge]: https://img.shields.io/badge/View-on%20GitHub-blue?logo=GitHub\n",
        "[colab-badge]: https://colab.research.google.com/assets/colab-badge.svg\n",
        "[binder-badge]: https://static.mybinder.org/badge_logo.svg\n",
        "\n",
        "[github-cifar10-pylearn2-lrn]: https://github.com/mbrukman/reimplementing-ml-papers/blob/main/alexnet/AlexNet_for_CIFAR-10_with_Pylearn2_Keras_LRN.ipynb\n",
        "[colab-cifar10-pylearn2-lrn]: https://colab.research.google.com/github/mbrukman/reimplementing-ml-papers/blob/main/alexnet/AlexNet_for_CIFAR-10_with_Pylearn2_Keras_LRN.ipynb\n",
        "[binder-cifar10-pylearn2-lrn]: https://mybinder.org/v2/gh/mbrukman/reimplementing-ml-papers/main?filepath=alexnet/AlexNet_for_CIFAR-10_with_Pylearn2_Keras_LRN.ipynb"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": null
      },
      "outputs": [],
      "source": [
        "%%bash\n",
        "\n",
        "readonly GH_USER=\"mbrukman\"\n",
        "readonly GH_REPO=\"reimplementing-ml-papers\"\n",
        "readonly GH_BRANCH=\"main\"\n",
        "\n",
        "# Download the Pylearn2/Keras implementation of LocalResponseNormalization. We\n",
        "# will use this layer below, even though the other implementation of LRN is\n",
        "# needed because it is used by default in AlexNet model definition, so it als\n",
        "# has to be downloaded.\n",
        "readonly PYLEARN2_LRN=\"pylearn2_local_response_normalization.py\"\n",
        "if ! [ -f \"${PYLEARN2_LRN}\" ]; then\n",
        "  # We save the target file under a different name, as we already have a file\n",
        "  # by the name of `local_response_normalization.py` in the current directory.\n",
        "  curl -s -o \"${PYLEARN2_LRN}\" \"https://raw.githubusercontent.com/${GH_USER}/${GH_REPO}/${GH_BRANCH}/third_party/pylearn2/local_response_normalization.py\"\n",
        "fi\n",
        "\n",
        "# Download the AlexNet CIFAR-10 model definition and LocalResponseNormalization\n",
        "# layer needed to construct it and our library to process the CIFAR-10 dataset.\n",
        "for path in alexnet/{alexnet_cifar10,local_response_normalization}.py \\\n",
        "            datasets/cifar-10/cifar10_keras.py ; do\n",
        "  module=\"$(basename \"${path}\")\"\n",
        "  if ! [ -f \"${module}\" ]; then\n",
        "    curl -s -o \"${module}\" \"https://raw.githubusercontent.com/${GH_USER}/${GH_REPO}/${GH_BRANCH}/${path}\"\n",
        "  fi\n",
        "done"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": null
      },
      "outputs": [],
      "source": [
        "from tensorflow import keras\n",
        "\n",
        "# Local imports downloaded above.\n",
        "from pylearn2_local_response_normalization import LRN2D\n",
        "from alexnet_cifar10 import AlexNet\n",
        "from cifar10_keras import CIFAR10"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": null
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Model: \"CIFAR-10-Pylearn2-Keras-LRN2D\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " Conv1 (Conv2D)              (None, 32, 32, 64)        4864      \n",
            "                                                                 \n",
            " MaxPool1 (MaxPooling2D)     (None, 15, 15, 64)        0         \n",
            "                                                                 \n",
            " LRN1 (LRN2D)                (None, 15, 15, 64)        0         \n",
            "                                                                 \n",
            " Conv2 (Conv2D)              (None, 15, 15, 64)        102464    \n",
            "                                                                 \n",
            " LRN2 (LRN2D)                (None, 15, 15, 64)        0         \n",
            "                                                                 \n",
            " MaxPool2 (MaxPooling2D)     (None, 7, 7, 64)          0         \n",
            "                                                                 \n",
            " Local3 (Conv2D)             (None, 7, 7, 64)          36928     \n",
            "                                                                 \n",
            " Local4 (Conv2D)             (None, 7, 7, 32)          18464     \n",
            "                                                                 \n",
            " Flatten (Flatten)           (None, 1568)              0         \n",
            "                                                                 \n",
            " FC10 (Dense)                (None, 10)                15690     \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 178410 (696.91 KB)\n",
            "Trainable params: 178410 (696.91 KB)\n",
            "Non-trainable params: 0 (0.00 Byte)\n",
            "_________________________________________________________________\n"
          ]
        }
      ],
      "source": [
        "model = AlexNet(lrn=LRN2D, lrn_name='Pylearn2-Keras-LRN2D')\n",
        "model.summary()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": null
      },
      "outputs": [],
      "source": [
        "%%capture --no-stderr\n",
        "\n",
        "# This will download the CIFAR-10 dataset via the Keras library, which writes to\n",
        "# stdout, so we silence it above to avoid extraneous output.\n",
        "cifar10_data = CIFAR10()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": null
      },
      "outputs": [],
      "source": [
        "# Compile the model with optimizer and loss function.\n",
        "model.compile(optimizer=keras.optimizers.Adam(learning_rate=0.001),\n",
        "              loss=keras.losses.CategoricalCrossentropy(),\n",
        "              metrics=[keras.metrics.CategoricalAccuracy()])"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": null
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "1563/1563 [==============================] - 334s 213ms/step - loss: 1.5085 - accuracy: 0.4538\n",
            "Epoch 2/20\n",
            "1563/1563 [==============================] - 333s 213ms/step - loss: 1.1214 - accuracy: 0.6022\n",
            "Epoch 3/20\n",
            "1563/1563 [==============================] - 335s 214ms/step - loss: 0.9469 - accuracy: 0.6651\n",
            "Epoch 4/20\n",
            "1563/1563 [==============================] - 345s 221ms/step - loss: 0.8448 - accuracy: 0.7037\n",
            "Epoch 5/20\n",
            "1563/1563 [==============================] - 333s 213ms/step - loss: 0.7661 - accuracy: 0.7308\n",
            "Epoch 6/20\n",
            "1563/1563 [==============================] - 333s 213ms/step - loss: 0.7095 - accuracy: 0.7527\n",
            "Epoch 7/20\n",
            "1563/1563 [==============================] - 333s 213ms/step - loss: 0.6607 - accuracy: 0.7681\n",
            "Epoch 8/20\n",
            "1563/1563 [==============================] - 332s 212ms/step - loss: 0.6190 - accuracy: 0.7817\n",
            "Epoch 9/20\n",
            "1563/1563 [==============================] - 333s 213ms/step - loss: 0.5824 - accuracy: 0.7961\n",
            "Epoch 10/20\n",
            "1563/1563 [==============================] - 335s 214ms/step - loss: 0.5482 - accuracy: 0.8063\n",
            "Epoch 11/20\n",
            "1563/1563 [==============================] - 334s 213ms/step - loss: 0.5117 - accuracy: 0.8196\n",
            "Epoch 12/20\n",
            "1563/1563 [==============================] - 334s 213ms/step - loss: 0.4790 - accuracy: 0.8289\n",
            "Epoch 13/20\n",
            "1563/1563 [==============================] - 335s 214ms/step - loss: 0.4556 - accuracy: 0.8390\n",
            "Epoch 14/20\n",
            "1563/1563 [==============================] - 333s 213ms/step - loss: 0.4274 - accuracy: 0.8467\n",
            "Epoch 15/20\n",
            "1563/1563 [==============================] - 343s 220ms/step - loss: 0.4089 - accuracy: 0.8546\n",
            "Epoch 16/20\n",
            "1563/1563 [==============================] - 336s 215ms/step - loss: 0.3913 - accuracy: 0.8611\n",
            "Epoch 17/20\n",
            "1563/1563 [==============================] - 333s 213ms/step - loss: 0.3654 - accuracy: 0.8680\n",
            "Epoch 18/20\n",
            "1563/1563 [==============================] - 331s 212ms/step - loss: 0.3563 - accuracy: 0.8726\n",
            "Epoch 19/20\n",
            "1563/1563 [==============================] - 331s 211ms/step - loss: 0.3334 - accuracy: 0.8805\n",
            "Epoch 20/20\n",
            "1563/1563 [==============================] - 338s 216ms/step - loss: 0.3192 - accuracy: 0.8867\n"
          ]
        },
        {
          "data": {
            "text/plain": [
              "<keras.callbacks.History at 0x7f4a7e7a5850>"
            ]
          },
          "execution_count": null,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# Train the model\n",
        "model.fit(cifar10_data.x_train_scale_0_1(),\n",
        "          cifar10_data.y_train_categorical(),\n",
        "          epochs=20)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": null
      },
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "313/313 [==============================] - 16s 51ms/step - loss: 1.3028 - accuracy: 0.6922\n"
          ]
        },
        {
          "data": {
            "text/plain": [
              "[1.3027948141098022, 0.6922000050544739]"
            ]
          },
          "execution_count": null,
          "metadata": {},
          "output_type": "execute_result"
        }
      ],
      "source": [
        "# Evaluate the model\n",
        "model.evaluate(cifar10_data.x_test_scale_0_1(),\n",
        "               cifar10_data.y_test_categorical())"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "gpuClass": "standard",
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}