a3_q1_mnist_baseline.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "a3-q1-mnist-baseline.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "mount_file_id": "1vDC-MHS06ZyIxPQDx2x1by4Rg72uHPe6",
      "authorship_tag": "ABX9TyPM6dRnu7lSWLyPE60gw561",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/mehrotrasan16/Keras-Deep-Learning/blob/03-tf-lite-quantizations-ASLMNIST/a3_q1_mnist_baseline.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "u9A1WE08_FLP",
        "outputId": "2835a932-6aa1-4eda-b253-1bd789e003dd",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "import tempfile\n",
        "import os\n",
        "\n",
        "import tensorflow as tf\n",
        "import numpy as np\n",
        "\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "from tensorflow import keras\n",
        "from keras import layers, models\n",
        "\n",
        "%load_ext tensorboard"
      ],
      "execution_count": 37,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "The tensorboard extension is already loaded. To reload it, use:\n",
            "  %reload_ext tensorboard\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-Qmn6m9TCfmB"
      },
      "source": [
        "# Input dataset"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RBiMnvxXsjZK"
      },
      "source": [
        "# Load ASL MNIST \n",
        "train_pd = pd.read_csv('/content/drive/My Drive/Colab Notebooks/CS581-TestResults/assignment3/q1/SignLanguageMNIST/sign_mnist_train/sign_mnist_train.csv')\n",
        "test_pd = pd.read_csv('/content/drive/My Drive/Colab Notebooks/CS581-TestResults/assignment3/q1/SignLanguageMNIST/sign_mnist_test/sign_mnist_test.csv')\n",
        "train_labels = train_pd['label'].values\n",
        "test_labels = test_pd['label'].values\n",
        "del(train_pd['label'])\n",
        "del(test_pd['label'])\n",
        "train_images = train_pd.values\n",
        "test_images = test_pd.values"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qtx_RFdSvZn-",
        "outputId": "e14b14b6-9022-4482-e3a8-ee9d341b69ab",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "test_images.shape, test_labels.shape,train_images.shape,train_labels.shape"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((7172, 784), (7172,), (27455, 784), (27455,))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qZl-zkaOvxMJ"
      },
      "source": [
        "train_images = train_images.reshape(train_images.shape[0],28,28,1)\n",
        "test_images = test_images.reshape(test_images.shape[0],28,28,1)"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3wfl2PvovLNU"
      },
      "source": [
        "# Normalize the input image so that each pixel value is between 0 to 1.\n",
        "train_images = train_images / 255.0\n",
        "test_images = test_images / 255.0"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hcipR4Y6zCOz",
        "outputId": "36a45468-4c17-4300-e06e-394d932e6241",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "test_images.shape, test_labels.shape,train_images.shape,train_labels.shape"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((7172, 28, 28, 1), (7172,), (27455, 28, 28, 1), (27455,))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ip3oYVQYEH3_"
      },
      "source": [
        "# Define the model architecture.\n",
        "model = models.Sequential([\n",
        "keras.layers.Conv2D(64, kernel_size=(3,3), padding='same', activation='relu'),\n",
        "layers.MaxPool2D(pool_size=(2,2)),\n",
        "keras.layers.Conv2D(96, kernel_size=(3,3), padding='same', activation='relu',use_bias=False),\n",
        "layers.MaxPool2D(pool_size=(2,2)),\n",
        "keras.layers.Conv2D(128, kernel_size=(3,3), padding='same', activation='relu',use_bias=False),\n",
        "layers.MaxPool2D(pool_size=(2,2)),\n",
        "layers.Flatten(),\n",
        "keras.layers.Dense(512, activation='relu',use_bias=False),\n",
        "layers.Dropout(0.60),\n",
        "layers.Dense(25,activation='softmax')\n",
        "])"
      ],
      "execution_count": 46,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pbY-KGMPvbW9",
        "outputId": "7889b8ae-2ac6-475c-dcac-dc4aa369413d",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "\n",
        "# Train the digit classification model\n",
        "model.compile(optimizer='adam',\n",
        "              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
        "              metrics=['accuracy'])\n",
        "\n",
        "history = model.fit(\n",
        "  train_images,\n",
        "  train_labels,\n",
        "  epochs=20,\n",
        "  batch_size=128,\n",
        "  validation_data=(test_images,test_labels)\n",
        ")\n"
      ],
      "execution_count": 47,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "215/215 [==============================] - 2s 7ms/step - loss: 3.1043 - accuracy: 0.1755 - val_loss: 2.9757 - val_accuracy: 0.3088\n",
            "Epoch 2/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.7231 - accuracy: 0.5731 - val_loss: 2.6533 - val_accuracy: 0.6454\n",
            "Epoch 3/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.5451 - accuracy: 0.7468 - val_loss: 2.6100 - val_accuracy: 0.6793\n",
            "Epoch 4/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.4614 - accuracy: 0.8291 - val_loss: 2.5039 - val_accuracy: 0.7864\n",
            "Epoch 5/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.4063 - accuracy: 0.8834 - val_loss: 2.4396 - val_accuracy: 0.8491\n",
            "Epoch 6/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3835 - accuracy: 0.9042 - val_loss: 2.4202 - val_accuracy: 0.8677\n",
            "Epoch 7/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3759 - accuracy: 0.9112 - val_loss: 2.4176 - val_accuracy: 0.8723\n",
            "Epoch 8/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3734 - accuracy: 0.9133 - val_loss: 2.4208 - val_accuracy: 0.8656\n",
            "Epoch 9/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3725 - accuracy: 0.9137 - val_loss: 2.4068 - val_accuracy: 0.8812\n",
            "Epoch 10/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3711 - accuracy: 0.9151 - val_loss: 2.4208 - val_accuracy: 0.8635\n",
            "Epoch 11/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3598 - accuracy: 0.9268 - val_loss: 2.3667 - val_accuracy: 0.9198\n",
            "Epoch 12/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3373 - accuracy: 0.9492 - val_loss: 2.3574 - val_accuracy: 0.9321\n",
            "Epoch 13/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3351 - accuracy: 0.9511 - val_loss: 2.3504 - val_accuracy: 0.9377\n",
            "Epoch 14/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3331 - accuracy: 0.9528 - val_loss: 2.3692 - val_accuracy: 0.9173\n",
            "Epoch 15/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3325 - accuracy: 0.9535 - val_loss: 2.3521 - val_accuracy: 0.9340\n",
            "Epoch 16/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3319 - accuracy: 0.9541 - val_loss: 2.3539 - val_accuracy: 0.9318\n",
            "Epoch 17/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3317 - accuracy: 0.9542 - val_loss: 2.3500 - val_accuracy: 0.9349\n",
            "Epoch 18/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.3062 - accuracy: 0.9805 - val_loss: 2.3234 - val_accuracy: 0.9646\n",
            "Epoch 19/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.2926 - accuracy: 0.9934 - val_loss: 2.3299 - val_accuracy: 0.9573\n",
            "Epoch 20/20\n",
            "215/215 [==============================] - 1s 6ms/step - loss: 2.2894 - accuracy: 0.9967 - val_loss: 2.3239 - val_accuracy: 0.9639\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sE8LlDX_EIvw",
        "outputId": "127a9628-4621-4c6c-c83b-f61b8006f51d",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "model.summary()"
      ],
      "execution_count": 48,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_3 (Conv2D)            (None, 28, 28, 64)        640       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_3 (MaxPooling2 (None, 14, 14, 64)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_4 (Conv2D)            (None, 14, 14, 96)        55296     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_4 (MaxPooling2 (None, 7, 7, 96)          0         \n",
            "_________________________________________________________________\n",
            "conv2d_5 (Conv2D)            (None, 7, 7, 128)         110592    \n",
            "_________________________________________________________________\n",
            "max_pooling2d_5 (MaxPooling2 (None, 3, 3, 128)         0         \n",
            "_________________________________________________________________\n",
            "flatten_1 (Flatten)          (None, 1152)              0         \n",
            "_________________________________________________________________\n",
            "dense_2 (Dense)              (None, 512)               589824    \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 512)               0         \n",
            "_________________________________________________________________\n",
            "dense_3 (Dense)              (None, 25)                12825     \n",
            "=================================================================\n",
            "Total params: 769,177\n",
            "Trainable params: 769,177\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OZ5jVUfxAUZA",
        "outputId": "dc467627-ab1c-44aa-f19a-bd1a367c09ac",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 562
        }
      },
      "source": [
        "plt.plot(history.history['accuracy'], label='accuracy')\n",
        "plt.plot(history.history['val_accuracy'], label = 'val_accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.ylim([0.5, 1])\n",
        "plt.legend(loc='lower right')\n",
        "plt.show()\n",
        "\n",
        "plt.plot(history.history['loss'], label='loss')\n",
        "plt.plot(history.history['val_loss'], label = 'val_loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Loss')\n",
        "#plt.ylim([0.5, 1])\n",
        "plt.legend(loc='lower right')\n",
        "plt.show()\n",
        "\n",
        "test_loss, test_acc = model.evaluate(test_images,  test_labels, verbose=2)"
      ],
      "execution_count": 49,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "225/225 - 0s - loss: 2.3239 - accuracy: 0.9639\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZRSX1HbVBOZc"
      },
      "source": [
        "tflite_models_dir = '/content/drive/My Drive/Colab Notebooks/CS581-TestResults/assignment3/q1/tflite_models'"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "A9tdCNrCIfg0"
      },
      "source": [
        "# Baseline to TFLite"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wLUkkua9IfFV",
        "outputId": "8666a98b-ba77-4c7c-856a-84a3f49d7a20",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "converter = tf.lite.TFLiteConverter.from_keras_model(model)\n",
        "tflite_model = converter.convert()\n",
        "tf_lite_model_file = '/content/drive/My Drive/Colab Notebooks/CS581-TestResults/assignment3/q1/tflite_models/sign_mnist_model.tflite'\n",
        "\n",
        "with open(tf_lite_model_file,'wb') as f:\n",
        "    f.write(tflite_model)"
      ],
      "execution_count": 12,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/training/tracking/tracking.py:111: Model.state_updates (from tensorflow.python.keras.engine.training) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "This property should not be used in TensorFlow 2.0, as updates are applied automatically.\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/training/tracking/tracking.py:111: Layer.updates (from tensorflow.python.keras.engine.base_layer) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "This property should not be used in TensorFlow 2.0, as updates are applied automatically.\n",
            "INFO:tensorflow:Assets written to: /tmp/tmp2n64pzqn/assets\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KvnF0eiPIgFK"
      },
      "source": [
        "### Baseline TFLite Size(MB)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NK6vwIQwNb5i",
        "outputId": "32b01fe0-2103-4860-bc87-e958599c9221",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "(os.path.getsize(tf_lite_model_file) / 1024) / 1024"
      ],
      "execution_count": 13,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "1.4630508422851562"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "24vL-4wzIj4c"
      },
      "source": [
        "### Baseline TFLite Accuracy"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eEAy2K1bRSkl"
      },
      "source": [
        "def evaluate_model(interpreter):\n",
        "    input_index = interpreter.get_input_details()[0][\"index\"]\n",
        "    output_index = interpreter.get_output_details()[0]['index']\n",
        "    predictions = []\n",
        "\n",
        "    for img in test_images:\n",
        "        timg = np.expand_dims(img,axis = 0).astype(np.float32)\n",
        "        interpreter.set_tensor(input_index,timg)\n",
        "        \n",
        "        interpreter.invoke()\n",
        "\n",
        "        output = interpreter.tensor(output_index)\n",
        "        pred = np.argmax(output()[0])\n",
        "        predictions.append(pred)\n",
        "\n",
        "    predictions = np.array(predictions)\n",
        "    accuracy= (predictions == test_labels).mean()\n",
        "    return accuracy    "
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-Qv6W2ctIBeY"
      },
      "source": [
        "interpreter= tf.lite.Interpreter(model_path='/content/drive/My Drive/Colab Notebooks/CS581-TestResults/assignment3/q1/tflite_models/sign_mnist_model.tflite')\n",
        "interpreter.allocate_tensors()"
      ],
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zNvmHVFYISsq"
      },
      "source": [
        "tflite_test_acc = evaluate_model(interpreter)"
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C_qiPT1BIcTN",
        "outputId": "5228a9d5-3104-4b1b-d098-32c3a2a34c70",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "tflite_test_acc"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8063301728945901"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 17
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MfKHxy5DB0fe"
      },
      "source": [
        "# Baseline to float16 optimization"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xieSDY55I6LU"
      },
      "source": [
        "converter = tf.lite.TFLiteConverter.from_keras_model(model)\n",
        "converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
        "converter.target_spec.supported_types = [tf.float16]"
      ],
      "execution_count": 18,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kum56nYqJXWz",
        "outputId": "5d99bd88-fd8a-44ca-99e5-f60a25d29008",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "tflite_fp16_model = converter.convert()\n",
        "tflite_fp16_file = '/content/drive/My Drive/Colab Notebooks/CS581-TestResults/assignment3/q1/tflite_models/sign_mnist_model_quant_f16.tflite'\n",
        "with open(tflite_fp16_file,'wb') as f:\n",
        "    f.write(tflite_fp16_model)"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmpvigo2_uy/assets\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmpvigo2_uy/assets\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-blQxSqmIsJF"
      },
      "source": [
        "### TFLite float16 optimization size(MB)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yhk995FIM1P0",
        "outputId": "ddebbbd4-a2c7-44d3-a879-512433d0f5a4",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "(os.path.getsize(tflite_fp16_file) / 1024) / 1024"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.7343902587890625"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "B6LmsHcnIwx7"
      },
      "source": [
        "### TFLite float16 optimization accuracy"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xzPpUDTnNZ_F"
      },
      "source": [
        "interpreter_f16= tf.lite.Interpreter(model_path='/content/drive/My Drive/Colab Notebooks/CS581-TestResults/assignment3/q1/tflite_models/sign_mnist_model_quant_f16.tflite')\n",
        "interpreter_f16.allocate_tensors()"
      ],
      "execution_count": 21,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fIWxe_SiI6M-",
        "outputId": "47b08101-871a-4596-d306-72038c2c3e1d",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "tflite_f16_test_acc = evaluate_model(interpreter_f16)\n",
        "tflite_f16_test_acc"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8063301728945901"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oB01-No2D6s5"
      },
      "source": [
        "# Dynamic Range Optimization"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vaN2xVxoD9QR",
        "outputId": "1bf8d799-774b-4b67-a642-2a221dace905",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "converter = tf.lite.TFLiteConverter.from_keras_model(model)\n",
        "converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
        "\n",
        "tflite_model_quant_dyn = converter.convert()"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmptv6b11ym/assets\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmptv6b11ym/assets\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "KtWPd3ZZoXMF",
        "outputId": "756ca802-951f-41b6-cd6b-2e3978573215",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "tflite_dyn_model = converter.convert()\n",
        "tflite_dyn_file = '/content/drive/My Drive/Colab Notebooks/CS581-TestResults/assignment3/q1/tflite_models/sign_mnist_model_quant_dyn.tflite'\n",
        "with open(tflite_dyn_file,'wb') as f:\n",
        "    f.write(tflite_dyn_model)"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmph8_fuiq8/assets\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmph8_fuiq8/assets\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BBt2-4ttqzdO"
      },
      "source": [
        "## Dynamic Optimization Size"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PpTAyOjGod9j",
        "outputId": "326a9ff1-2333-4e57-e499-21d6eb930992",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "(os.path.getsize(tflite_dyn_file) / 1024) / 1024"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.372100830078125"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 25
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "K6je1Jk0q3OP"
      },
      "source": [
        "## Dynamic Optimization Accuracy"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T_chrMWDp7P3"
      },
      "source": [
        "interpreter_dyn = tf.lite.Interpreter(model_path='/content/drive/My Drive/Colab Notebooks/CS581-TestResults/assignment3/q1/tflite_models/sign_mnist_model_quant_dyn.tflite')\n",
        "interpreter_dyn.allocate_tensors()"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "FguZw0nurzJB",
        "outputId": "3f3b4d19-f7d5-4530-c016-162809cf2880",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "tflite_dyn_test_acc = evaluate_model(interpreter_dyn)\n",
        "tflite_dyn_test_acc"
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.8063301728945901"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 27
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "peSU5JjtCBAF"
      },
      "source": [
        "# Baseline to int8 quantization (weights and activations)\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yJ426WP6KJvh",
        "outputId": "dc3fcc3d-aae6-4980-8dbd-cc9240451308",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "type(train_images),train_images.shape"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(numpy.ndarray, (27455, 28, 28, 1))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 28
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "P-xhCPECKdv7"
      },
      "source": [
        "train_images = train_images.astype(np.float32)"
      ],
      "execution_count": 29,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VS5Bd-1GCGbV",
        "outputId": "4e11ba4e-f0f3-4c5e-97d3-35f1fd010f2e",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "def representative_data_gen():\n",
        "  for input_value in tf.data.Dataset.from_tensor_slices(train_images).batch(1).take(100):\n",
        "    yield [input_value]\n",
        "\n",
        "converter = tf.lite.TFLiteConverter.from_keras_model(model)\n",
        "converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
        "converter.representative_dataset = representative_data_gen\n",
        "# Ensure that if any ops can't be quantized, the converter throws an error\n",
        "converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]\n",
        "# Set the input and output tensors to uint8 (APIs added in r2.3)\n",
        "converter.inference_input_type = tf.uint8\n",
        "converter.inference_output_type = tf.uint8\n",
        "\n",
        "tflite_int8_model = converter.convert()"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmpsvzvz8s2/assets\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmpsvzvz8s2/assets\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5rVtGtYvLx6n"
      },
      "source": [
        "tflite_int8_file = '/content/drive/My Drive/Colab Notebooks/CS581-TestResults/assignment3/q1/tflite_models/sign_mnist_model_quant_int8.tflite'\n",
        "with open(tflite_int8_file,'wb') as f:\n",
        "    f.write(tflite_int8_model)"
      ],
      "execution_count": 31,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_zW2sM5oL0Yu"
      },
      "source": [
        "### TFLite to int8 optimization size"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PDL4qWRADNTX",
        "outputId": "92465834-f43b-48c6-f2de-2b54937021b0",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "(os.path.getsize(tflite_int8_file) / 1024) / 1024"
      ],
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "0.376708984375"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 32
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "h-qa3chvL-T6"
      },
      "source": [
        "### TLFITE to int8 optimization accuracy"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "U599cI9Nm996"
      },
      "source": [
        "# Helper function to evaluate a TFLite model on all images\n",
        "def evaluate_model_int8(tflite_file, model_type):\n",
        "  global test_images\n",
        "  global test_labels\n",
        "\n",
        "  test_image_indices = range(test_images.shape[0])\n",
        "  predictions = run_tflite_model(tflite_file, test_image_indices)\n",
        "\n",
        "  accuracy = (np.sum(test_labels== predictions) * 100) / len(test_images)\n",
        "\n",
        "  print('%s model accuracy is %.4f%% (Number of test samples=%d)' % (\n",
        "      model_type, accuracy, len(test_images)))"
      ],
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b5R05MBcm6Jg"
      },
      "source": [
        "# Helper function to run inference on a TFLite model\n",
        "def run_tflite_model(tflite_file, test_image_indices):\n",
        "  global test_images\n",
        "\n",
        "  # Initialize the interpreter\n",
        "  interpreter = tf.lite.Interpreter(model_path=str(tflite_file))\n",
        "  interpreter.allocate_tensors()\n",
        "\n",
        "  input_details = interpreter.get_input_details()[0]\n",
        "  output_details = interpreter.get_output_details()[0]\n",
        "\n",
        "  predictions = np.zeros((len(test_image_indices),), dtype=int)\n",
        "  for i, test_image_index in enumerate(test_image_indices):\n",
        "    test_image = test_images[test_image_index]\n",
        "    test_label = test_labels[test_image_index]\n",
        "\n",
        "    # Check if the input type is quantized, then rescale input data to uint8\n",
        "    if input_details['dtype'] == np.uint8:\n",
        "      input_scale, input_zero_point = input_details[\"quantization\"]\n",
        "      test_image = test_image / input_scale + input_zero_point\n",
        "\n",
        "    test_image = np.expand_dims(test_image, axis=0).astype(input_details[\"dtype\"])\n",
        "    interpreter.set_tensor(input_details[\"index\"], test_image)\n",
        "    interpreter.invoke()\n",
        "    output = interpreter.get_tensor(output_details[\"index\"])[0]\n",
        "\n",
        "    predictions[i] = output.argmax()\n",
        "\n",
        "  return predictions\n"
      ],
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5WZlv1lTvQJb",
        "outputId": "a75bebd8-6922-4739-e3fc-07a09227ec9b",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "evaluate_model_int8(tflite_int8_file,model_type='Quantized')"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Quantized model accuracy is 80.4657% (Number of test samples=7172)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Fc3j7eOzx2Rh"
      },
      "source": [
        ""
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zLbwfLBx31rv"
      },
      "source": [
        "# Quantization Aware Optimization - on the Whole model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PkX0aBogGGsT",
        "outputId": "4f25dd80-3926-4121-df1a-2ed6ba34e4f0",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "! pip install -q tensorflow-model-optimization\n"
      ],
      "execution_count": 38,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\u001b[?25l\r\u001b[K     |██                              | 10kB 24.7MB/s eta 0:00:01\r\u001b[K     |███▉                            | 20kB 3.5MB/s eta 0:00:01\r\u001b[K     |█████▊                          | 30kB 4.6MB/s eta 0:00:01\r\u001b[K     |███████▋                        | 40kB 4.8MB/s eta 0:00:01\r\u001b[K     |█████████▌                      | 51kB 4.0MB/s eta 0:00:01\r\u001b[K     |███████████▍                    | 61kB 4.5MB/s eta 0:00:01\r\u001b[K     |█████████████▎                  | 71kB 4.7MB/s eta 0:00:01\r\u001b[K     |███████████████▏                | 81kB 5.1MB/s eta 0:00:01\r\u001b[K     |█████████████████               | 92kB 5.4MB/s eta 0:00:01\r\u001b[K     |███████████████████             | 102kB 5.4MB/s eta 0:00:01\r\u001b[K     |████████████████████▉           | 112kB 5.4MB/s eta 0:00:01\r\u001b[K     |██████████████████████▊         | 122kB 5.4MB/s eta 0:00:01\r\u001b[K     |████████████████████████▊       | 133kB 5.4MB/s eta 0:00:01\r\u001b[K     |██████████████████████████▋     | 143kB 5.4MB/s eta 0:00:01\r\u001b[K     |████████████████████████████▌   | 153kB 5.4MB/s eta 0:00:01\r\u001b[K     |██████████████████████████████▍ | 163kB 5.4MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 174kB 5.4MB/s \n",
            "\u001b[?25h"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bYVjVQdZ35Fy",
        "outputId": "31d28ce9-5ee9-4887-b2f4-766ba31b8f80",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "import tensorflow_model_optimization as tfmot\n",
        "\n",
        "quantize_model = tfmot.quantization.keras.quantize_model\n",
        "\n",
        "# q_aware stands for for quantization aware.\n",
        "q_aware_model = quantize_model(model)\n",
        "\n",
        "# `quantize_model` requires a recompile.\n",
        "q_aware_model.compile(optimizer='adam',\n",
        "              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),\n",
        "              metrics=['accuracy'])\n",
        "\n",
        "q_aware_model.summary()"
      ],
      "execution_count": 50,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "quantize_layer_2 (QuantizeLa (None, 28, 28, 1)         3         \n",
            "_________________________________________________________________\n",
            "quant_conv2d_3 (QuantizeWrap (None, 28, 28, 64)        771       \n",
            "_________________________________________________________________\n",
            "quant_max_pooling2d_3 (Quant (None, 14, 14, 64)        1         \n",
            "_________________________________________________________________\n",
            "quant_conv2d_4 (QuantizeWrap (None, 14, 14, 96)        55491     \n",
            "_________________________________________________________________\n",
            "quant_max_pooling2d_4 (Quant (None, 7, 7, 96)          1         \n",
            "_________________________________________________________________\n",
            "quant_conv2d_5 (QuantizeWrap (None, 7, 7, 128)         110851    \n",
            "_________________________________________________________________\n",
            "quant_max_pooling2d_5 (Quant (None, 3, 3, 128)         1         \n",
            "_________________________________________________________________\n",
            "quant_flatten_1 (QuantizeWra (None, 1152)              1         \n",
            "_________________________________________________________________\n",
            "quant_dense_2 (QuantizeWrapp (None, 512)               589829    \n",
            "_________________________________________________________________\n",
            "quant_dropout_1 (QuantizeWra (None, 512)               1         \n",
            "_________________________________________________________________\n",
            "quant_dense_3 (QuantizeWrapp (None, 25)                12830     \n",
            "=================================================================\n",
            "Total params: 769,780\n",
            "Trainable params: 769,177\n",
            "Non-trainable params: 603\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TOTh5yVdF_vi",
        "outputId": "6f7f0402-120d-4012-8b9a-55200f237ed6",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "train_images_subset = train_images[0:1000] # out of 60000\n",
        "train_labels_subset = train_labels[0:1000]\n",
        "\n",
        "histroy = q_aware_model.fit(train_images_subset, train_labels_subset,\n",
        "                  batch_size=500, epochs=10, validation_data=(test_images,test_labels))"
      ],
      "execution_count": 51,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "2/2 [==============================] - 0s 189ms/step - loss: 2.9365 - accuracy: 0.4580 - val_loss: 2.8208 - val_accuracy: 0.5763\n",
            "Epoch 2/10\n",
            "2/2 [==============================] - 0s 94ms/step - loss: 2.8520 - accuracy: 0.5140 - val_loss: 2.7133 - val_accuracy: 0.6670\n",
            "Epoch 3/10\n",
            "2/2 [==============================] - 0s 94ms/step - loss: 2.7572 - accuracy: 0.6150 - val_loss: 2.6098 - val_accuracy: 0.7250\n",
            "Epoch 4/10\n",
            "2/2 [==============================] - 0s 95ms/step - loss: 2.6496 - accuracy: 0.7300 - val_loss: 2.5291 - val_accuracy: 0.7875\n",
            "Epoch 5/10\n",
            "2/2 [==============================] - 0s 93ms/step - loss: 2.5743 - accuracy: 0.7910 - val_loss: 2.4686 - val_accuracy: 0.8503\n",
            "Epoch 6/10\n",
            "2/2 [==============================] - 0s 93ms/step - loss: 2.5133 - accuracy: 0.8410 - val_loss: 2.4391 - val_accuracy: 0.8772\n",
            "Epoch 7/10\n",
            "2/2 [==============================] - 0s 92ms/step - loss: 2.4579 - accuracy: 0.8980 - val_loss: 2.4195 - val_accuracy: 0.9037\n",
            "Epoch 8/10\n",
            "2/2 [==============================] - 0s 95ms/step - loss: 2.4349 - accuracy: 0.9150 - val_loss: 2.4023 - val_accuracy: 0.9195\n",
            "Epoch 9/10\n",
            "2/2 [==============================] - 0s 93ms/step - loss: 2.4027 - accuracy: 0.9380 - val_loss: 2.3840 - val_accuracy: 0.9334\n",
            "Epoch 10/10\n",
            "2/2 [==============================] - 0s 96ms/step - loss: 2.4001 - accuracy: 0.9400 - val_loss: 2.3669 - val_accuracy: 0.9483\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B2W6PG4uGV2R",
        "outputId": "9a6742e7-52e2-40b3-a320-6a35508ee77b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 562
        }
      },
      "source": [
        "plt.plot(history.history['accuracy'], label='accuracy')\n",
        "plt.plot(history.history['val_accuracy'], label = 'val_accuracy')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.ylim([0.5, 1])\n",
        "plt.legend(loc='lower right')\n",
        "plt.show()\n",
        "\n",
        "plt.plot(history.history['loss'], label='loss')\n",
        "plt.plot(history.history['val_loss'], label = 'val_loss')\n",
        "plt.xlabel('Epoch')\n",
        "plt.ylabel('Loss')\n",
        "#plt.ylim([0.5, 1])\n",
        "plt.legend(loc='lower right')\n",
        "plt.show()\n",
        "\n",
        "test_loss, test_acc = q_aware_model.evaluate(test_images,  test_labels, verbose=2)"
      ],
      "execution_count": 52,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "225/225 - 1s - loss: 2.3669 - accuracy: 0.9483\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iqA-M96HGqWq",
        "outputId": "896074d8-4612-4360-ce94-b5542a9d0430",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "_, model = model.evaluate(\n",
        "    test_images, test_labels, verbose=0)\n",
        "\n",
        "_, q_aware_model_accuracy = q_aware_model.evaluate(\n",
        "   test_images, test_labels, verbose=0)\n",
        "\n",
        "print('Baseline test accuracy:', model)\n",
        "print('Quant test accuracy:', q_aware_model_accuracy)"
      ],
      "execution_count": 53,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Baseline test accuracy: 0.9638873338699341\n",
            "Quant test accuracy: 0.9482710361480713\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eVMMiQ83HzdG"
      },
      "source": [
        "## Qunatization Aware Model - to TFLite"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aLZFaqSbHu4M",
        "outputId": "d6371ee4-b043-48a9-e9c4-b5f3090743c8",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "converter = tf.lite.TFLiteConverter.from_keras_model(q_aware_model)\n",
        "converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
        "\n",
        "quantized_tflite_model = converter.convert()"
      ],
      "execution_count": 54,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmp1_o94711/assets\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmp1_o94711/assets\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UrrlzwAaH7az",
        "outputId": "32d4053a-1963-4f71-9899-2ec6821e1dee",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "converter = tf.lite.TFLiteConverter.from_keras_model(q_aware_model)\n",
        "converter.optimizations = [tf.lite.Optimize.DEFAULT]\n",
        "\n",
        "quantized_tflite_model = converter.convert()"
      ],
      "execution_count": 55,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmpnvfqbuoi/assets\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: /tmp/tmpnvfqbuoi/assets\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y0HseVimJl4h",
        "outputId": "4c6e05e7-fa1f-46da-8ea9-46a380683753",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "interpreter = tf.lite.Interpreter(model_content=quantized_tflite_model)\n",
        "interpreter.allocate_tensors()\n",
        "\n",
        "test_accuracy = evaluate_model(interpreter)\n",
        "\n",
        "print('Quant aware TFLite test_accuracy:', test_accuracy)\n",
        "print('Quant aware TF test accuracy:', q_aware_model_accuracy)"
      ],
      "execution_count": 56,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Quant TFLite test_accuracy: 0.9484104852203011\n",
            "Quant TF test accuracy: 0.9482710361480713\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YEL6QGhsKvDx"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}