CIFAR10_danya_colab_2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CIFAR-10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from matrix_square_root_power import *\n",
    "from shampoo_optimizer import *\n",
    "from tensorflow.keras import datasets, layers, models\n",
    "import tensorflow as tf\n",
    "from __future__ import print_function\n",
    "import keras\n",
    "from keras.layers import Dense, Conv2D, BatchNormalization, Activation, Dropout\n",
    "from keras.layers import AveragePooling2D, Input, Flatten, GlobalAveragePooling2D\n",
    "from keras.optimizers import Adam\n",
    "from keras.callbacks import ModelCheckpoint, LearningRateScheduler\n",
    "from keras.callbacks import ReduceLROnPlateau\n",
    "from keras.preprocessing.image import ImageDataGenerator\n",
    "from keras.regularizers import l2\n",
    "from keras import backend as K\n",
    "from keras.models import Model, Sequential\n",
    "from keras.datasets import cifar10\n",
    "from keras import backend\n",
    "from keras.utils import np_utils\n",
    "import numpy\n",
    "from keras.datasets import cifar10\n",
    "from keras.layers import Flatten\n",
    "from keras.constraints import maxnorm\n",
    "from keras.optimizers import SGD\n",
    "from keras.layers.convolutional import Conv2D\n",
    "from keras.layers.convolutional import MaxPooling2D\n",
    "from keras.utils import np_utils\n",
    "from keras import backend as K"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "(train_images, train_labels), (test_images, test_labels) = datasets.cifar10.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50000, 32, 32, 3)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_images.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Vizualization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 25 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "class_names = ['airplane', 'automobile', 'bird', 'cat', 'deer',\n",
    "               'dog', 'frog', 'horse', 'ship', 'truck']\n",
    "\n",
    "plt.figure(figsize=(10,10))\n",
    "for i in range(25):\n",
    "    plt.subplot(5,5,i+1)\n",
    "    plt.xticks([])\n",
    "    plt.yticks([])\n",
    "    plt.grid(False)\n",
    "    plt.imshow(train_images[i], cmap=plt.cm.binary)\n",
    "    # The CIFAR labels happen to be arrays, \n",
    "    # which is why you need the extra index\n",
    "    plt.xlabel(class_names[train_labels[i][0]])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_images = train_images.astype('float32')\n",
    "test_images = test_images.astype('float32')\n",
    "train_images = train_images / 255.0\n",
    "test_images = test_images / 255.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_classes = test_labels.shape[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_7 (Conv2D)            (None, 30, 30, 32)        896       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_8 (Conv2D)            (None, 13, 13, 64)        18496     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2 (None, 6, 6, 64)          0         \n",
      "_________________________________________________________________\n",
      "conv2d_9 (Conv2D)            (None, 4, 4, 64)          36928     \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 1024)              0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 64)                65600     \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 10)                650       \n",
      "=================================================================\n",
      "Total params: 122,570\n",
      "Trainable params: 122,570\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model = models.Sequential()\n",
    "model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.Flatten())\n",
    "model.add(layers.Dense(64, activation='relu'))\n",
    "model.add(layers.Dense(10, activation='softmax'))\n",
    "# Compile model\n",
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def loss(model, x, y):\n",
    "    y_ = model(x)\n",
    "\n",
    "    return loss_object(y_true=y, y_pred=y_)\n",
    "\n",
    "def grad(model, inputs, targets):\n",
    "    with tf.GradientTape() as tape:\n",
    "        loss_value = loss(model, inputs, targets)\n",
    "        return loss_value, tape.gradient(loss_value, model.trainable_weights)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_images = tf.convert_to_tensor(train_images, dtype=tf.float32)\n",
    "test_images = tf.convert_to_tensor(test_images, dtype=tf.float32)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Shampoo, lr = 0.01, eps = 0.0001"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 000: Loss: 1.940, Accuracy: 53.990%\n",
      "Epoch 001: Loss: 1.857, Accuracy: 60.988%\n",
      "Epoch 002: Loss: 1.820, Accuracy: 64.503%\n",
      "Epoch 003: Loss: 1.801, Accuracy: 65.931%\n",
      "Epoch 004: Loss: 1.794, Accuracy: 66.404%\n",
      "Epoch 005: Loss: 1.787, Accuracy: 66.925%\n",
      "Epoch 006: Loss: 1.781, Accuracy: 67.460%\n",
      "Epoch 007: Loss: 1.778, Accuracy: 67.704%\n",
      "Epoch 008: Loss: 1.773, Accuracy: 68.113%\n",
      "Epoch 009: Loss: 1.771, Accuracy: 68.257%\n",
      "Epoch 010: Loss: 1.770, Accuracy: 68.281%\n",
      "Epoch 011: Loss: 1.770, Accuracy: 68.281%\n",
      "Epoch 012: Loss: 1.732, Accuracy: 72.694%\n",
      "Epoch 013: Loss: 1.700, Accuracy: 76.064%\n",
      "Epoch 014: Loss: 1.694, Accuracy: 76.500%\n",
      "Epoch 015: Loss: 1.691, Accuracy: 76.689%\n",
      "Epoch 016: Loss: 1.689, Accuracy: 76.765%\n",
      "Epoch 017: Loss: 1.689, Accuracy: 76.773%\n",
      "Epoch 018: Loss: 1.688, Accuracy: 76.773%\n",
      "Epoch 019: Loss: 1.688, Accuracy: 76.773%\n",
      "Epoch 020: Loss: 1.688, Accuracy: 76.777%\n",
      "Epoch 021: Loss: 1.687, Accuracy: 76.823%\n",
      "Epoch 022: Loss: 1.684, Accuracy: 77.133%\n",
      "Epoch 023: Loss: 1.683, Accuracy: 77.264%\n",
      "Epoch 024: Loss: 1.682, Accuracy: 77.266%\n",
      "Epoch 025: Loss: 1.682, Accuracy: 77.268%\n",
      "Epoch 026: Loss: 1.682, Accuracy: 77.268%\n",
      "Epoch 027: Loss: 1.682, Accuracy: 77.268%\n",
      "Epoch 028: Loss: 1.682, Accuracy: 77.268%\n",
      "Epoch 029: Loss: 1.682, Accuracy: 77.268%\n",
      "Epoch 030: Loss: 1.682, Accuracy: 77.268%\n",
      "Epoch 031: Loss: 1.665, Accuracy: 79.157%\n",
      "Epoch 032: Loss: 1.624, Accuracy: 83.644%\n",
      "Epoch 033: Loss: 1.618, Accuracy: 84.014%\n",
      "Epoch 034: Loss: 1.617, Accuracy: 84.119%\n",
      "Epoch 035: Loss: 1.614, Accuracy: 84.369%\n",
      "Epoch 036: Loss: 1.613, Accuracy: 84.465%\n",
      "Epoch 037: Loss: 1.612, Accuracy: 84.495%\n",
      "Epoch 038: Loss: 1.611, Accuracy: 84.625%\n",
      "Epoch 039: Loss: 1.609, Accuracy: 84.838%\n",
      "Epoch 040: Loss: 1.609, Accuracy: 84.838%\n",
      "Epoch 041: Loss: 1.609, Accuracy: 84.838%\n",
      "Epoch 042: Loss: 1.609, Accuracy: 84.838%\n",
      "Epoch 043: Loss: 1.609, Accuracy: 84.838%\n",
      "Epoch 044: Loss: 1.609, Accuracy: 84.838%\n",
      "Epoch 045: Loss: 1.608, Accuracy: 84.838%\n",
      "Epoch 046: Loss: 1.608, Accuracy: 84.838%\n",
      "Epoch 047: Loss: 1.608, Accuracy: 84.838%\n",
      "Epoch 048: Loss: 1.608, Accuracy: 84.838%\n",
      "Epoch 049: Loss: 1.608, Accuracy: 84.856%\n",
      "Epoch 050: Loss: 1.607, Accuracy: 84.946%\n",
      "Epoch 051: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 052: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 053: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 054: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 055: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 056: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 057: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 058: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 059: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 060: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 061: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 062: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 063: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 064: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 065: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 066: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 067: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 068: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 069: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 070: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 071: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 072: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 073: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 074: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 075: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 076: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 077: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 078: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 079: Loss: 1.607, Accuracy: 84.960%\n",
      "Epoch 080: Loss: 1.607, Accuracy: 84.960%\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-63-a96d10399a11>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     17\u001b[0m         \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtrain_labels\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     18\u001b[0m         \u001b[0;31m# Optimize the model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 19\u001b[0;31m         \u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrads\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     20\u001b[0m         \u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mapply_gradients\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mzip\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgrads\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrainable_weights\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     21\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-59-fa66445214eb>\u001b[0m in \u001b[0;36mgrad\u001b[0;34m(model, inputs, targets)\u001b[0m\n\u001b[1;32m      7\u001b[0m     \u001b[0;32mwith\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mGradientTape\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mtape\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m         \u001b[0mloss_value\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mloss\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtargets\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 9\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtape\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgradient\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrainable_weights\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/eager/backprop.py\u001b[0m in \u001b[0;36mgradient\u001b[0;34m(self, target, sources, output_gradients, unconnected_gradients)\u001b[0m\n\u001b[1;32m   1021\u001b[0m                           for x in nest.flatten(output_gradients)]\n\u001b[1;32m   1022\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1023\u001b[0;31m     flat_grad = imperative_grad.imperative_grad(\n\u001b[0m\u001b[1;32m   1024\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_tape\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1025\u001b[0m         \u001b[0mflat_targets\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/eager/imperative_grad.py\u001b[0m in \u001b[0;36mimperative_grad\u001b[0;34m(tape, target, sources, output_gradients, sources_raw, unconnected_gradients)\u001b[0m\n\u001b[1;32m     69\u001b[0m         \"Unknown value for unconnected_gradients: %r\" % unconnected_gradients)\n\u001b[1;32m     70\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 71\u001b[0;31m   return pywrap_tensorflow.TFE_Py_TapeGradient(\n\u001b[0m\u001b[1;32m     72\u001b[0m       \u001b[0mtape\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_tape\u001b[0m\u001b[0;34m,\u001b[0m  \u001b[0;31m# pylint: disable=protected-access\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     73\u001b[0m       \u001b[0mtarget\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/eager/backprop.py\u001b[0m in \u001b[0;36m_gradient_function\u001b[0;34m(op_name, attr_tuple, num_inputs, inputs, outputs, out_grads, skip_input_indices)\u001b[0m\n\u001b[1;32m    139\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mnum_inputs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    140\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 141\u001b[0;31m   \u001b[0;32mreturn\u001b[0m \u001b[0mgrad_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmock_op\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0mout_grads\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    142\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    143\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/ops/nn_grad.py\u001b[0m in \u001b[0;36m_Conv2DGrad\u001b[0;34m(op, grad)\u001b[0m\n\u001b[1;32m    595\u001b[0m           \u001b[0muse_cudnn_on_gpu\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0muse_cudnn_on_gpu\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    596\u001b[0m           data_format=data_format),\n\u001b[0;32m--> 597\u001b[0;31m       gen_nn_ops.conv2d_backprop_filter(\n\u001b[0m\u001b[1;32m    598\u001b[0m           \u001b[0mop\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    599\u001b[0m           \u001b[0mshape_1\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/ops/gen_nn_ops.py\u001b[0m in \u001b[0;36mconv2d_backprop_filter\u001b[0;34m(input, filter_sizes, out_backprop, strides, padding, use_cudnn_on_gpu, explicit_paddings, data_format, dilations, name)\u001b[0m\n\u001b[1;32m   1074\u001b[0m   \u001b[0;32mif\u001b[0m \u001b[0mtld\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_eager\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1075\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1076\u001b[0;31m       _result = _pywrap_tensorflow.TFE_Py_FastPathExecute(\n\u001b[0m\u001b[1;32m   1077\u001b[0m         \u001b[0m_ctx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_context_handle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtld\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdevice_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"Conv2DBackpropFilter\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1078\u001b[0m         \u001b[0mtld\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mop_callbacks\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfilter_sizes\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout_backprop\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"strides\"\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "loss_object = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
    "optimizer = ShampooOptimizer(learning_rate=0.01)\n",
    "\n",
    "shampoo_train_loss_results = []\n",
    "shampoo_train_accuracy_results = []\n",
    "\n",
    "epochs = 250\n",
    "batch_size = 128\n",
    "\n",
    "for epoch in range(epochs):\n",
    "    epoch_loss_avg = tf.keras.metrics.Mean()\n",
    "    epoch_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()\n",
    "\n",
    "    # Training loop - using batches of 128\n",
    "    for i in range(train_images.shape[0] // batch_size):\n",
    "        x = train_images[i:i+batch_size]\n",
    "        y = train_labels[i:i+batch_size]\n",
    "        # Optimize the model\n",
    "        loss_value, grads = grad(model, x, y)\n",
    "        optimizer.apply_gradients(zip(grads, model.trainable_weights))\n",
    "\n",
    "        # Track progress\n",
    "        epoch_loss_avg(loss_value)  # Add current batch loss\n",
    "        # Compare predicted label to actual label\n",
    "        epoch_accuracy(y, model(x))\n",
    "\n",
    "        # End epoch\n",
    "    shampoo_train_loss_results.append(epoch_loss_avg.result())\n",
    "    shampoo_train_accuracy_results.append(epoch_accuracy.result())\n",
    "\n",
    "    print(\"Epoch {:03d}: Loss: {:.3f}, Accuracy: {:.3%}\".format(epoch,\n",
    "                                                                epoch_loss_avg.result(),\n",
    "                                                                epoch_accuracy.result()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axes = plt.subplots(2, sharex=True, figsize=(12, 8))\n",
    "fig.suptitle('Training Metrics')\n",
    "\n",
    "axes[0].set_ylabel(\"Loss\", fontsize=14)\n",
    "axes[0].plot(shampoo_train_loss_results, label = r\"Shampoo, $\\alpha = {0}$\".format(0.01))\n",
    "axes[0].grid()\n",
    "axes[0].legend()\n",
    "\n",
    "axes[1].set_ylabel(\"Accuracy\", fontsize=14)\n",
    "axes[1].set_xlabel(\"Epoch\", fontsize=14)\n",
    "axes[1].plot(shampoo_train_accuracy_results, label = r\"Shampoo, $\\alpha = {0}$\".format(0.01))\n",
    "axes[1].grid()\n",
    "axes[1].legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Adam, lr = 0.001"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 000: Loss: 2.047, Accuracy: 41.488%\n",
      "Epoch 001: Loss: 1.904, Accuracy: 55.633%\n",
      "Epoch 002: Loss: 1.792, Accuracy: 66.997%\n",
      "Epoch 003: Loss: 1.705, Accuracy: 75.563%\n",
      "Epoch 004: Loss: 1.631, Accuracy: 83.037%\n",
      "Epoch 005: Loss: 1.605, Accuracy: 85.533%\n",
      "Epoch 006: Loss: 1.596, Accuracy: 86.324%\n",
      "Epoch 007: Loss: 1.587, Accuracy: 87.190%\n",
      "Epoch 008: Loss: 1.577, Accuracy: 88.155%\n",
      "Epoch 009: Loss: 1.575, Accuracy: 88.321%\n",
      "Epoch 010: Loss: 1.572, Accuracy: 88.536%\n",
      "Epoch 011: Loss: 1.572, Accuracy: 88.536%\n",
      "Epoch 012: Loss: 1.572, Accuracy: 88.556%\n",
      "Epoch 013: Loss: 1.563, Accuracy: 89.595%\n",
      "Epoch 014: Loss: 1.556, Accuracy: 90.485%\n",
      "Epoch 015: Loss: 1.552, Accuracy: 90.911%\n",
      "Epoch 016: Loss: 1.539, Accuracy: 92.095%\n",
      "Epoch 017: Loss: 1.538, Accuracy: 92.181%\n",
      "Epoch 018: Loss: 1.535, Accuracy: 92.422%\n",
      "Epoch 019: Loss: 1.535, Accuracy: 92.440%\n",
      "Epoch 020: Loss: 1.532, Accuracy: 92.660%\n",
      "Epoch 021: Loss: 1.531, Accuracy: 92.843%\n",
      "Epoch 022: Loss: 1.529, Accuracy: 93.127%\n",
      "Epoch 023: Loss: 1.528, Accuracy: 93.163%\n",
      "Epoch 024: Loss: 1.527, Accuracy: 93.297%\n",
      "Epoch 025: Loss: 1.524, Accuracy: 93.652%\n",
      "Epoch 026: Loss: 1.524, Accuracy: 93.646%\n",
      "Epoch 027: Loss: 1.518, Accuracy: 94.183%\n",
      "Epoch 028: Loss: 1.513, Accuracy: 94.673%\n",
      "Epoch 029: Loss: 1.515, Accuracy: 94.515%\n",
      "Epoch 030: Loss: 1.513, Accuracy: 94.700%\n",
      "Epoch 031: Loss: 1.515, Accuracy: 94.567%\n",
      "Epoch 032: Loss: 1.511, Accuracy: 94.904%\n",
      "Epoch 033: Loss: 1.508, Accuracy: 95.206%\n",
      "Epoch 034: Loss: 1.507, Accuracy: 95.270%\n",
      "Epoch 035: Loss: 1.507, Accuracy: 95.258%\n",
      "Epoch 036: Loss: 1.507, Accuracy: 95.262%\n",
      "Epoch 037: Loss: 1.507, Accuracy: 95.284%\n",
      "Epoch 038: Loss: 1.507, Accuracy: 95.284%\n",
      "Epoch 039: Loss: 1.507, Accuracy: 95.284%\n",
      "Epoch 040: Loss: 1.507, Accuracy: 95.284%\n",
      "Epoch 041: Loss: 1.506, Accuracy: 95.365%\n",
      "Epoch 042: Loss: 1.520, Accuracy: 94.157%\n",
      "Epoch 043: Loss: 1.511, Accuracy: 94.988%\n",
      "Epoch 044: Loss: 1.507, Accuracy: 95.323%\n",
      "Epoch 045: Loss: 1.506, Accuracy: 95.457%\n",
      "Epoch 046: Loss: 1.509, Accuracy: 95.188%\n",
      "Epoch 047: Loss: 1.505, Accuracy: 95.531%\n",
      "Epoch 048: Loss: 1.505, Accuracy: 95.531%\n",
      "Epoch 049: Loss: 1.505, Accuracy: 95.531%\n",
      "Epoch 050: Loss: 1.505, Accuracy: 95.531%\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-68-ecff981d4b7f>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     27\u001b[0m         \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtrain_labels\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     28\u001b[0m         \u001b[0;31m# Optimize the model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 29\u001b[0;31m         \u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrads\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     30\u001b[0m         \u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mapply_gradients\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mzip\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgrads\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrainable_weights\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     31\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-59-fa66445214eb>\u001b[0m in \u001b[0;36mgrad\u001b[0;34m(model, inputs, targets)\u001b[0m\n\u001b[1;32m      7\u001b[0m     \u001b[0;32mwith\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mGradientTape\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mtape\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m         \u001b[0mloss_value\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mloss\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtargets\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 9\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtape\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgradient\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrainable_weights\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/eager/backprop.py\u001b[0m in \u001b[0;36mgradient\u001b[0;34m(self, target, sources, output_gradients, unconnected_gradients)\u001b[0m\n\u001b[1;32m   1021\u001b[0m                           for x in nest.flatten(output_gradients)]\n\u001b[1;32m   1022\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1023\u001b[0;31m     flat_grad = imperative_grad.imperative_grad(\n\u001b[0m\u001b[1;32m   1024\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_tape\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1025\u001b[0m         \u001b[0mflat_targets\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/eager/imperative_grad.py\u001b[0m in \u001b[0;36mimperative_grad\u001b[0;34m(tape, target, sources, output_gradients, sources_raw, unconnected_gradients)\u001b[0m\n\u001b[1;32m     69\u001b[0m         \"Unknown value for unconnected_gradients: %r\" % unconnected_gradients)\n\u001b[1;32m     70\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 71\u001b[0;31m   return pywrap_tensorflow.TFE_Py_TapeGradient(\n\u001b[0m\u001b[1;32m     72\u001b[0m       \u001b[0mtape\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_tape\u001b[0m\u001b[0;34m,\u001b[0m  \u001b[0;31m# pylint: disable=protected-access\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     73\u001b[0m       \u001b[0mtarget\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/eager/backprop.py\u001b[0m in \u001b[0;36m_gradient_function\u001b[0;34m(op_name, attr_tuple, num_inputs, inputs, outputs, out_grads, skip_input_indices)\u001b[0m\n\u001b[1;32m    139\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mnum_inputs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    140\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 141\u001b[0;31m   \u001b[0;32mreturn\u001b[0m \u001b[0mgrad_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmock_op\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0mout_grads\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    142\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    143\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/ops/nn_grad.py\u001b[0m in \u001b[0;36m_Conv2DGrad\u001b[0;34m(op, grad)\u001b[0m\n\u001b[1;32m    585\u001b[0m   \u001b[0;31m# in Eager mode.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    586\u001b[0m   return [\n\u001b[0;32m--> 587\u001b[0;31m       gen_nn_ops.conv2d_backprop_input(\n\u001b[0m\u001b[1;32m    588\u001b[0m           \u001b[0mshape_0\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    589\u001b[0m           \u001b[0mop\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/ops/gen_nn_ops.py\u001b[0m in \u001b[0;36mconv2d_backprop_input\u001b[0;34m(input_sizes, filter, out_backprop, strides, padding, use_cudnn_on_gpu, explicit_paddings, data_format, dilations, name)\u001b[0m\n\u001b[1;32m   1234\u001b[0m   \u001b[0;32mif\u001b[0m \u001b[0mtld\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_eager\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1235\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1236\u001b[0;31m       _result = _pywrap_tensorflow.TFE_Py_FastPathExecute(\n\u001b[0m\u001b[1;32m   1237\u001b[0m         \u001b[0m_ctx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_context_handle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtld\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdevice_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"Conv2DBackpropInput\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1238\u001b[0m         \u001b[0mtld\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mop_callbacks\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput_sizes\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfilter\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout_backprop\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"strides\"\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "model = models.Sequential()\n",
    "model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.Flatten())\n",
    "model.add(layers.Dense(64, activation='relu'))\n",
    "model.add(layers.Dense(10, activation='softmax'))\n",
    "\n",
    "loss_object = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
    "optimizer = tf.keras.optimizers.Adam(learning_rate=0.001)\n",
    "\n",
    "adam_train_loss_results = []\n",
    "adam_train_accuracy_results = []\n",
    "\n",
    "epochs = 250\n",
    "batch_size = 128\n",
    "\n",
    "for epoch in range(epochs):\n",
    "    epoch_loss_avg = tf.keras.metrics.Mean()\n",
    "    epoch_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()\n",
    "\n",
    "    # Training loop - using batches of 128\n",
    "    for i in range(train_images.shape[0] // batch_size):\n",
    "        x = train_images[i:i+batch_size]\n",
    "        y = train_labels[i:i+batch_size]\n",
    "        # Optimize the model\n",
    "        loss_value, grads = grad(model, x, y)\n",
    "        optimizer.apply_gradients(zip(grads, model.trainable_weights))\n",
    "\n",
    "        # Track progress\n",
    "        epoch_loss_avg(loss_value)  # Add current batch loss\n",
    "        # Compare predicted label to actual label\n",
    "        epoch_accuracy(y, model(x))\n",
    "\n",
    "        # End epoch\n",
    "    adam_train_loss_results.append(epoch_loss_avg.result())\n",
    "    adam_train_accuracy_results.append(epoch_accuracy.result())\n",
    "\n",
    "    print(\"Epoch {:03d}: Loss: {:.3f}, Accuracy: {:.3%}\".format(epoch,\n",
    "                                                                epoch_loss_avg.result(),\n",
    "                                                                epoch_accuracy.result()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_metric(optimizer_name, alpha):\n",
    "    fig, axes = plt.subplots(2, sharex=True, figsize=(12, 8))\n",
    "    fig.suptitle('Training Metrics')\n",
    "\n",
    "    axes[0].set_ylabel(\"Loss\", fontsize=14)\n",
    "    axes[0].plot(adam_train_loss_results, label = r\"{0}, $\\alpha = {1}$\".format(optimizer_name, alpha))\n",
    "    axes[0].grid()\n",
    "    axes[0].legend()\n",
    "\n",
    "    axes[1].set_ylabel(\"Accuracy\", fontsize=14)\n",
    "    axes[1].set_xlabel(\"Epoch\", fontsize=14)\n",
    "    axes[1].plot(adam_train_accuracy_results, label = r\"{0}, $\\alpha = {1}$\".format(optimizer_name, alpha))\n",
    "    axes[1].grid()\n",
    "    axes[1].legend()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_metric(\"Adam\", 0.001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(\"adam_train_loss_results_alpha_0_01_cnn_cifar_10.txt\", \"w\") as file:\n",
    "    print(*adam_train_loss_results, file=file)\n",
    "    \n",
    "with open(\"adam_train_accuracy_results_alpha_0_01_cnn_cifar_10.txt\", \"w\") as file:\n",
    "    print(*adam_train_accuracy_results, file=file)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Shampoo, lr = 0.001, eps = 0.0001"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 000: Loss: 2.269, Accuracy: 20.180%\n",
      "Epoch 001: Loss: 2.215, Accuracy: 25.853%\n",
      "Epoch 002: Loss: 2.193, Accuracy: 27.869%\n",
      "Epoch 003: Loss: 2.178, Accuracy: 29.307%\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-74-4306b4614866>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     26\u001b[0m         \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtrain_labels\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     27\u001b[0m         \u001b[0;31m# Optimize the model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 28\u001b[0;31m         \u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrads\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     29\u001b[0m         \u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mapply_gradients\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mzip\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgrads\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrainable_weights\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     30\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-59-fa66445214eb>\u001b[0m in \u001b[0;36mgrad\u001b[0;34m(model, inputs, targets)\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtargets\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m     \u001b[0;32mwith\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mGradientTape\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mtape\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 8\u001b[0;31m         \u001b[0mloss_value\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mloss\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtargets\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      9\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtape\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgradient\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrainable_weights\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-59-fa66445214eb>\u001b[0m in \u001b[0;36mloss\u001b[0;34m(model, x, y)\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mloss\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m     \u001b[0my_\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mloss_object\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_true\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_pred\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my_\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/engine/base_layer.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, inputs, *args, **kwargs)\u001b[0m\n\u001b[1;32m    820\u001b[0m           with base_layer_utils.autocast_context_manager(\n\u001b[1;32m    821\u001b[0m               self._compute_dtype):\n\u001b[0;32m--> 822\u001b[0;31m             \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcast_inputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    823\u001b[0m           \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_handle_activity_regularization\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moutputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    824\u001b[0m           \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_set_mask_metadata\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moutputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput_masks\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/engine/sequential.py\u001b[0m in \u001b[0;36mcall\u001b[0;34m(self, inputs, training, mask)\u001b[0m\n\u001b[1;32m    265\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbuilt\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    266\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_init_graph_network\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moutputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 267\u001b[0;31m       \u001b[0;32mreturn\u001b[0m \u001b[0msuper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mSequential\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtraining\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtraining\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmask\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmask\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    268\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    269\u001b[0m     \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0minputs\u001b[0m  \u001b[0;31m# handle the corner case where self.layers is empty\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/engine/network.py\u001b[0m in \u001b[0;36mcall\u001b[0;34m(self, inputs, training, mask)\u001b[0m\n\u001b[1;32m    713\u001b[0m                                 ' implement a `call` method.')\n\u001b[1;32m    714\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 715\u001b[0;31m     return self._run_internal_graph(\n\u001b[0m\u001b[1;32m    716\u001b[0m         \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtraining\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtraining\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmask\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmask\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    717\u001b[0m         convert_kwargs_to_constants=base_layer_utils.call_context().saving)\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/engine/network.py\u001b[0m in \u001b[0;36m_run_internal_graph\u001b[0;34m(self, inputs, training, mask, convert_kwargs_to_constants)\u001b[0m\n\u001b[1;32m    889\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    890\u001b[0m           \u001b[0;31m# Compute outputs.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 891\u001b[0;31m           \u001b[0moutput_tensors\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlayer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcomputed_tensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    892\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    893\u001b[0m           \u001b[0;31m# Update tensor_dict.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/engine/base_layer.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, inputs, *args, **kwargs)\u001b[0m\n\u001b[1;32m    658\u001b[0m     \u001b[0;31m# mode when all inputs can be traced back to `keras.Input()` (when building\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    659\u001b[0m     \u001b[0;31m# models using the functional API).\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 660\u001b[0;31m     \u001b[0mbuild_graph\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf_utils\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mare_all_symbolic_tensors\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput_list\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    661\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    662\u001b[0m     \u001b[0;31m# Accept NumPy and scalar inputs by converting to Tensors.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/utils/tf_utils.py\u001b[0m in \u001b[0;36mare_all_symbolic_tensors\u001b[0;34m(tensors)\u001b[0m\n\u001b[1;32m    314\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    315\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mare_all_symbolic_tensors\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensors\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 316\u001b[0;31m   \u001b[0;32mreturn\u001b[0m \u001b[0mall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mis_symbolic_tensor\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtensor\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mtensor\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mtensors\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    317\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    318\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "model = models.Sequential()\n",
    "model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.Flatten())\n",
    "model.add(layers.Dense(64, activation='relu'))\n",
    "model.add(layers.Dense(10, activation='softmax'))\n",
    "loss_object = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
    "optimizer = ShampooOptimizer(learning_rate=0.001)\n",
    "\n",
    "shampoo_train_loss_results = []\n",
    "shampoo_train_accuracy_results = []\n",
    "\n",
    "epochs = 250\n",
    "batch_size = 128\n",
    "\n",
    "for epoch in range(epochs):\n",
    "    epoch_loss_avg = tf.keras.metrics.Mean()\n",
    "    epoch_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()\n",
    "\n",
    "    # Training loop - using batches of 128\n",
    "    for i in range(train_images.shape[0] // batch_size):\n",
    "        x = train_images[i:i+batch_size]\n",
    "        y = train_labels[i:i+batch_size]\n",
    "        # Optimize the model\n",
    "        loss_value, grads = grad(model, x, y)\n",
    "        optimizer.apply_gradients(zip(grads, model.trainable_weights))\n",
    "\n",
    "        # Track progress\n",
    "        epoch_loss_avg(loss_value)  # Add current batch loss\n",
    "        # Compare predicted label to actual label\n",
    "        epoch_accuracy(y, model(x))\n",
    "\n",
    "        # End epoch\n",
    "    shampoo_train_loss_results.append(epoch_loss_avg.result())\n",
    "    shampoo_train_accuracy_results.append(epoch_accuracy.result())\n",
    "\n",
    "    print(\"Epoch {:03d}: Loss: {:.3f}, Accuracy: {:.3%}\".format(epoch,\n",
    "                                                                epoch_loss_avg.result(),\n",
    "                                                                epoch_accuracy.result()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Shampoo, lr = 0.1, eps = 0.0001"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 000: Loss: 2.330, Accuracy: 13.047%\n",
      "Epoch 001: Loss: 2.331, Accuracy: 13.041%\n",
      "Epoch 002: Loss: 2.331, Accuracy: 13.041%\n",
      "Epoch 003: Loss: 2.136, Accuracy: 32.382%\n",
      "Epoch 004: Loss: 1.855, Accuracy: 60.471%\n",
      "Epoch 005: Loss: 1.747, Accuracy: 71.092%\n",
      "Epoch 006: Loss: 1.707, Accuracy: 74.988%\n",
      "Epoch 007: Loss: 1.662, Accuracy: 79.651%\n",
      "Epoch 008: Loss: 1.632, Accuracy: 82.500%\n",
      "Epoch 009: Loss: 1.626, Accuracy: 83.049%\n",
      "Epoch 010: Loss: 1.624, Accuracy: 83.171%\n",
      "Epoch 011: Loss: 1.621, Accuracy: 83.442%\n",
      "Epoch 012: Loss: 1.619, Accuracy: 83.730%\n",
      "Epoch 013: Loss: 1.617, Accuracy: 83.840%\n",
      "Epoch 014: Loss: 1.617, Accuracy: 83.928%\n",
      "Epoch 015: Loss: 1.616, Accuracy: 83.928%\n",
      "Epoch 016: Loss: 1.616, Accuracy: 83.928%\n",
      "Epoch 017: Loss: 1.616, Accuracy: 83.934%\n",
      "Epoch 018: Loss: 1.616, Accuracy: 84.008%\n",
      "Epoch 019: Loss: 1.615, Accuracy: 84.016%\n",
      "Epoch 020: Loss: 1.612, Accuracy: 84.371%\n",
      "Epoch 021: Loss: 1.610, Accuracy: 84.567%\n",
      "Epoch 022: Loss: 1.609, Accuracy: 84.685%\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-75-b45027f0b6bf>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     26\u001b[0m         \u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtrain_labels\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     27\u001b[0m         \u001b[0;31m# Optimize the model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 28\u001b[0;31m         \u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrads\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     29\u001b[0m         \u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mapply_gradients\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mzip\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgrads\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrainable_weights\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     30\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-59-fa66445214eb>\u001b[0m in \u001b[0;36mgrad\u001b[0;34m(model, inputs, targets)\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtargets\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m     \u001b[0;32mwith\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mGradientTape\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mtape\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 8\u001b[0;31m         \u001b[0mloss_value\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mloss\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtargets\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      9\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtape\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgradient\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrainable_weights\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-59-fa66445214eb>\u001b[0m in \u001b[0;36mloss\u001b[0;34m(model, x, y)\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mloss\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m     \u001b[0my_\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mloss_object\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my_true\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my_pred\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my_\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/engine/base_layer.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, inputs, *args, **kwargs)\u001b[0m\n\u001b[1;32m    820\u001b[0m           with base_layer_utils.autocast_context_manager(\n\u001b[1;32m    821\u001b[0m               self._compute_dtype):\n\u001b[0;32m--> 822\u001b[0;31m             \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcast_inputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    823\u001b[0m           \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_handle_activity_regularization\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moutputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    824\u001b[0m           \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_set_mask_metadata\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moutputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput_masks\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/engine/sequential.py\u001b[0m in \u001b[0;36mcall\u001b[0;34m(self, inputs, training, mask)\u001b[0m\n\u001b[1;32m    265\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbuilt\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    266\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_init_graph_network\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0moutputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 267\u001b[0;31m       \u001b[0;32mreturn\u001b[0m \u001b[0msuper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mSequential\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtraining\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtraining\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmask\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmask\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    268\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    269\u001b[0m     \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0minputs\u001b[0m  \u001b[0;31m# handle the corner case where self.layers is empty\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/engine/network.py\u001b[0m in \u001b[0;36mcall\u001b[0;34m(self, inputs, training, mask)\u001b[0m\n\u001b[1;32m    713\u001b[0m                                 ' implement a `call` method.')\n\u001b[1;32m    714\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 715\u001b[0;31m     return self._run_internal_graph(\n\u001b[0m\u001b[1;32m    716\u001b[0m         \u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtraining\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtraining\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmask\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmask\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    717\u001b[0m         convert_kwargs_to_constants=base_layer_utils.call_context().saving)\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/engine/network.py\u001b[0m in \u001b[0;36m_run_internal_graph\u001b[0;34m(self, inputs, training, mask, convert_kwargs_to_constants)\u001b[0m\n\u001b[1;32m    889\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    890\u001b[0m           \u001b[0;31m# Compute outputs.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 891\u001b[0;31m           \u001b[0moutput_tensors\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlayer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcomputed_tensors\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    892\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    893\u001b[0m           \u001b[0;31m# Update tensor_dict.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/engine/base_layer.py\u001b[0m in \u001b[0;36m__call__\u001b[0;34m(self, inputs, *args, **kwargs)\u001b[0m\n\u001b[1;32m    820\u001b[0m           with base_layer_utils.autocast_context_manager(\n\u001b[1;32m    821\u001b[0m               self._compute_dtype):\n\u001b[0;32m--> 822\u001b[0;31m             \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mcast_inputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    823\u001b[0m           \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_handle_activity_regularization\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moutputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    824\u001b[0m           \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_set_mask_metadata\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moutputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minput_masks\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/keras/layers/core.py\u001b[0m in \u001b[0;36mcall\u001b[0;34m(self, inputs)\u001b[0m\n\u001b[1;32m   1087\u001b[0m         \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msparse_ops\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msparse_tensor_dense_matmul\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkernel\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1088\u001b[0m       \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1089\u001b[0;31m         \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgen_math_ops\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmat_mul\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkernel\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1090\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0muse_bias\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1091\u001b[0m       \u001b[0moutputs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbias_add\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moutputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbias\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/ops/gen_math_ops.py\u001b[0m in \u001b[0;36mmat_mul\u001b[0;34m(a, b, transpose_a, transpose_b, name)\u001b[0m\n\u001b[1;32m   5616\u001b[0m   \u001b[0;32mif\u001b[0m \u001b[0mtld\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mis_eager\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   5617\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 5618\u001b[0;31m       _result = _pywrap_tensorflow.TFE_Py_FastPathExecute(\n\u001b[0m\u001b[1;32m   5619\u001b[0m         \u001b[0m_ctx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_context_handle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtld\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdevice_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"MatMul\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   5620\u001b[0m         \u001b[0mtld\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mop_callbacks\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0ma\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mb\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"transpose_a\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtranspose_a\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"transpose_b\"\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "model = models.Sequential()\n",
    "model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.Flatten())\n",
    "model.add(layers.Dense(64, activation='relu'))\n",
    "model.add(layers.Dense(10, activation='softmax'))\n",
    "loss_object = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
    "optimizer = ShampooOptimizer(learning_rate=0.1)\n",
    "\n",
    "shampoo_train_loss_results = []\n",
    "shampoo_train_accuracy_results = []\n",
    "\n",
    "epochs = 250\n",
    "batch_size = 128\n",
    "\n",
    "for epoch in range(epochs):\n",
    "    epoch_loss_avg = tf.keras.metrics.Mean()\n",
    "    epoch_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()\n",
    "\n",
    "    # Training loop - using batches of 128\n",
    "    for i in range(train_images.shape[0] // batch_size):\n",
    "        x = train_images[i:i+batch_size]\n",
    "        y = train_labels[i:i+batch_size]\n",
    "        # Optimize the model\n",
    "        loss_value, grads = grad(model, x, y)\n",
    "        optimizer.apply_gradients(zip(grads, model.trainable_weights))\n",
    "\n",
    "        # Track progress\n",
    "        epoch_loss_avg(loss_value)  # Add current batch loss\n",
    "        # Compare predicted label to actual label\n",
    "        epoch_accuracy(y, model(x))\n",
    "\n",
    "        # End epoch\n",
    "    shampoo_train_loss_results.append(epoch_loss_avg.result())\n",
    "    shampoo_train_accuracy_results.append(epoch_accuracy.result())\n",
    "\n",
    "    print(\"Epoch {:03d}: Loss: {:.3f}, Accuracy: {:.3%}\".format(epoch,\n",
    "                                                                epoch_loss_avg.result(),\n",
    "                                                                epoch_accuracy.result()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Shampoo, lr = 0.01, eps = 0.00001"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 000: Loss: 2.047, Accuracy: 43.169%\n",
      "Epoch 001: Loss: 1.945, Accuracy: 51.871%\n",
      "Epoch 002: Loss: 1.883, Accuracy: 58.011%\n",
      "Epoch 003: Loss: 1.839, Accuracy: 62.077%\n",
      "Epoch 004: Loss: 1.820, Accuracy: 63.758%\n",
      "Epoch 005: Loss: 1.811, Accuracy: 64.357%\n",
      "Epoch 006: Loss: 1.804, Accuracy: 65.000%\n",
      "Epoch 007: Loss: 1.793, Accuracy: 66.094%\n",
      "Epoch 008: Loss: 1.789, Accuracy: 66.418%\n",
      "Epoch 009: Loss: 1.787, Accuracy: 66.436%\n",
      "Epoch 010: Loss: 1.787, Accuracy: 66.440%\n",
      "Epoch 011: Loss: 1.786, Accuracy: 66.442%\n",
      "Epoch 012: Loss: 1.785, Accuracy: 66.617%\n",
      "Epoch 013: Loss: 1.781, Accuracy: 67.073%\n",
      "Epoch 014: Loss: 1.776, Accuracy: 67.544%\n",
      "Epoch 015: Loss: 1.775, Accuracy: 67.626%\n",
      "Epoch 016: Loss: 1.774, Accuracy: 67.652%\n",
      "Epoch 017: Loss: 1.773, Accuracy: 67.756%\n",
      "Epoch 018: Loss: 1.769, Accuracy: 68.173%\n",
      "Epoch 019: Loss: 1.768, Accuracy: 68.243%\n",
      "Epoch 020: Loss: 1.767, Accuracy: 68.371%\n",
      "Epoch 021: Loss: 1.766, Accuracy: 68.379%\n",
      "Epoch 022: Loss: 1.765, Accuracy: 68.504%\n",
      "Epoch 023: Loss: 1.762, Accuracy: 68.754%\n",
      "Epoch 024: Loss: 1.761, Accuracy: 68.882%\n",
      "Epoch 025: Loss: 1.760, Accuracy: 68.990%\n",
      "Epoch 026: Loss: 1.760, Accuracy: 68.990%\n",
      "Epoch 027: Loss: 1.760, Accuracy: 68.992%\n",
      "Epoch 028: Loss: 1.759, Accuracy: 68.992%\n",
      "Epoch 029: Loss: 1.759, Accuracy: 68.992%\n",
      "Epoch 030: Loss: 1.759, Accuracy: 68.992%\n",
      "Epoch 031: Loss: 1.759, Accuracy: 68.992%\n",
      "Epoch 032: Loss: 1.759, Accuracy: 68.992%\n",
      "Epoch 033: Loss: 1.759, Accuracy: 68.992%\n",
      "Epoch 034: Loss: 1.759, Accuracy: 68.992%\n",
      "Epoch 035: Loss: 1.759, Accuracy: 68.992%\n",
      "Epoch 036: Loss: 1.759, Accuracy: 68.992%\n",
      "Epoch 037: Loss: 1.757, Accuracy: 69.183%\n",
      "Epoch 038: Loss: 1.753, Accuracy: 69.615%\n",
      "Epoch 039: Loss: 1.752, Accuracy: 69.762%\n",
      "Epoch 040: Loss: 1.752, Accuracy: 69.762%\n",
      "Epoch 041: Loss: 1.752, Accuracy: 69.780%\n",
      "Epoch 042: Loss: 1.748, Accuracy: 70.208%\n",
      "Epoch 043: Loss: 1.744, Accuracy: 70.587%\n",
      "Epoch 044: Loss: 1.744, Accuracy: 70.591%\n",
      "Epoch 045: Loss: 1.743, Accuracy: 70.661%\n",
      "Epoch 046: Loss: 1.740, Accuracy: 70.950%\n",
      "Epoch 047: Loss: 1.739, Accuracy: 71.104%\n",
      "Epoch 048: Loss: 1.739, Accuracy: 71.106%\n",
      "Epoch 049: Loss: 1.739, Accuracy: 71.108%\n",
      "Epoch 050: Loss: 1.738, Accuracy: 71.108%\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-76-9b0f4e953d6f>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m     27\u001b[0m         \u001b[0;31m# Optimize the model\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     28\u001b[0m         \u001b[0mloss_value\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrads\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 29\u001b[0;31m         \u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mapply_gradients\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mzip\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgrads\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrainable_weights\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     30\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     31\u001b[0m         \u001b[0;31m# Track progress\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/training/optimizer.py\u001b[0m in \u001b[0;36mapply_gradients\u001b[0;34m(self, grads_and_vars, global_step, name)\u001b[0m\n\u001b[1;32m    612\u001b[0m           \u001b[0mscope_name\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mvar\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mop\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    613\u001b[0m         \u001b[0;32mwith\u001b[0m \u001b[0mops\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mname_scope\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"update_\"\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0mscope_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mops\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcolocate_with\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mvar\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 614\u001b[0;31m           \u001b[0mupdate_ops\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mprocessor\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mupdate_op\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    615\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0mglobal_step\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    616\u001b[0m         \u001b[0mapply_updates\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_finish\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mupdate_ops\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/training/optimizer.py\u001b[0m in \u001b[0;36mupdate_op\u001b[0;34m(self, optimizer, g)\u001b[0m\n\u001b[1;32m    169\u001b[0m       return optimizer._resource_apply_sparse_duplicate_indices(\n\u001b[1;32m    170\u001b[0m           g.values, self._v, g.indices)\n\u001b[0;32m--> 171\u001b[0;31m     \u001b[0mupdate_op\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_resource_apply_dense\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mg\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_v\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    172\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_v\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconstraint\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    173\u001b[0m       \u001b[0;32mwith\u001b[0m \u001b[0mops\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcontrol_dependencies\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mupdate_op\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-2-ea6605aa264a>\u001b[0m in \u001b[0;36m_resource_apply_dense\u001b[0;34m(self, grad, var)\u001b[0m\n\u001b[1;32m    123\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    124\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_resource_apply_dense\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvar\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 125\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_apply_dense\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgrad\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvar\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    126\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    127\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_apply_dense\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvar\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-2-ea6605aa264a>\u001b[0m in \u001b[0;36m_apply_dense\u001b[0;34m(self, grad, var)\u001b[0m\n\u001b[1;32m    126\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    127\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_apply_dense\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvar\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 128\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_apply_gradient\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mgrad\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvar\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    129\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    130\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_resource_apply_sparse\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad_values\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mvar\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgrad_indices\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m<ipython-input-2-ea6605aa264a>\u001b[0m in \u001b[0;36m_apply_gradient\u001b[0;34m(self, grad, var, indices)\u001b[0m\n\u001b[1;32m    395\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    396\u001b[0m             var_updated = state_ops.assign_sub(var,\n\u001b[0;32m--> 397\u001b[0;31m                                              learning_rate_t * preconditioned_grad)\n\u001b[0m\u001b[1;32m    398\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mvar_updated\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/ops/math_ops.py\u001b[0m in \u001b[0;36mr_binary_op_wrapper\u001b[0;34m(y, x)\u001b[0m\n\u001b[1;32m    923\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    924\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0mr_binary_op_wrapper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 925\u001b[0;31m     \u001b[0;32mwith\u001b[0m \u001b[0mops\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mname_scope\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mop_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    926\u001b[0m       \u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mops\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconvert_to_tensor\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdtype\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdtype\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbase_dtype\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m\"x\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    927\u001b[0m       \u001b[0;32mreturn\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mname\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/tensorflow_core/python/framework/ops.py\u001b[0m in \u001b[0;36m__enter__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   6375\u001b[0m         \u001b[0mbut\u001b[0m\u001b[0;31m \u001b[0m\u001b[0;31m`\u001b[0m\u001b[0mvalues\u001b[0m\u001b[0;31m`\u001b[0m \u001b[0mare\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   6376\u001b[0m     \"\"\"\n\u001b[0;32m-> 6377\u001b[0;31m     \u001b[0mctx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcontext\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcontext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   6378\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mctx\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mexecuting_eagerly\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   6379\u001b[0m       \u001b[0mscope_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mold_scope_name\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0menter_eager_name_scope\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mctx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "model = models.Sequential()\n",
    "model.add(layers.Conv2D(32, (3, 3), activation='relu', input_shape=(32, 32, 3)))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.MaxPooling2D((2, 2)))\n",
    "model.add(layers.Conv2D(64, (3, 3), activation='relu'))\n",
    "model.add(layers.Flatten())\n",
    "model.add(layers.Dense(64, activation='relu'))\n",
    "model.add(layers.Dense(10, activation='softmax'))\n",
    "loss_object = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
    "optimizer = ShampooOptimizer(learning_rate=0.01, epsilon=1e-5)\n",
    "\n",
    "shampoo_train_loss_results = []\n",
    "shampoo_train_accuracy_results = []\n",
    "\n",
    "epochs = 250\n",
    "batch_size = 128\n",
    "\n",
    "for epoch in range(epochs):\n",
    "    epoch_loss_avg = tf.keras.metrics.Mean()\n",
    "    epoch_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()\n",
    "\n",
    "    # Training loop - using batches of 128\n",
    "    for i in range(train_images.shape[0] // batch_size):\n",
    "        x = train_images[i:i+batch_size]\n",
    "        y = train_labels[i:i+batch_size]\n",
    "        # Optimize the model\n",
    "        loss_value, grads = grad(model, x, y)\n",
    "        optimizer.apply_gradients(zip(grads, model.trainable_weights))\n",
    "\n",
    "        # Track progress\n",
    "        epoch_loss_avg(loss_value)  # Add current batch loss\n",
    "        # Compare predicted label to actual label\n",
    "        epoch_accuracy(y, model(x))\n",
    "\n",
    "        # End epoch\n",
    "    shampoo_train_loss_results.append(epoch_loss_avg.result())\n",
    "    shampoo_train_accuracy_results.append(epoch_accuracy.result())\n",
    "\n",
    "    print(\"Epoch {:03d}: Loss: {:.3f}, Accuracy: {:.3%}\".format(epoch,\n",
    "                                                                epoch_loss_avg.result(),\n",
    "                                                                epoch_accuracy.result()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# We try to train VGG net, but it works too long :("
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"vgg16\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_12 (InputLayer)        [(None, 32, 32, 3)]       0         \n",
      "_________________________________________________________________\n",
      "block1_conv1 (Conv2D)        (None, 32, 32, 64)        1792      \n",
      "_________________________________________________________________\n",
      "block1_conv2 (Conv2D)        (None, 32, 32, 64)        36928     \n",
      "_________________________________________________________________\n",
      "block1_pool (MaxPooling2D)   (None, 16, 16, 64)        0         \n",
      "_________________________________________________________________\n",
      "block2_conv1 (Conv2D)        (None, 16, 16, 128)       73856     \n",
      "_________________________________________________________________\n",
      "block2_conv2 (Conv2D)        (None, 16, 16, 128)       147584    \n",
      "_________________________________________________________________\n",
      "block2_pool (MaxPooling2D)   (None, 8, 8, 128)         0         \n",
      "_________________________________________________________________\n",
      "block3_conv1 (Conv2D)        (None, 8, 8, 256)         295168    \n",
      "_________________________________________________________________\n",
      "block3_conv2 (Conv2D)        (None, 8, 8, 256)         590080    \n",
      "_________________________________________________________________\n",
      "block3_conv3 (Conv2D)        (None, 8, 8, 256)         590080    \n",
      "_________________________________________________________________\n",
      "block3_pool (MaxPooling2D)   (None, 4, 4, 256)         0         \n",
      "_________________________________________________________________\n",
      "block4_conv1 (Conv2D)        (None, 4, 4, 512)         1180160   \n",
      "_________________________________________________________________\n",
      "block4_conv2 (Conv2D)        (None, 4, 4, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block4_conv3 (Conv2D)        (None, 4, 4, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block4_pool (MaxPooling2D)   (None, 2, 2, 512)         0         \n",
      "_________________________________________________________________\n",
      "block5_conv1 (Conv2D)        (None, 2, 2, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv2 (Conv2D)        (None, 2, 2, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_conv3 (Conv2D)        (None, 2, 2, 512)         2359808   \n",
      "_________________________________________________________________\n",
      "block5_pool (MaxPooling2D)   (None, 1, 1, 512)         0         \n",
      "_________________________________________________________________\n",
      "flatten (Flatten)            (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "fc1 (Dense)                  (None, 4096)              2101248   \n",
      "_________________________________________________________________\n",
      "fc2 (Dense)                  (None, 4096)              16781312  \n",
      "_________________________________________________________________\n",
      "predictions (Dense)          (None, 10)                40970     \n",
      "=================================================================\n",
      "Total params: 33,638,218\n",
      "Trainable params: 33,638,218\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from tensorflow.keras.applications.vgg16 import VGG16\n",
    "\n",
    "model = VGG16(\n",
    "    weights=None, \n",
    "    include_top=True, \n",
    "    classes=10,\n",
    "    input_shape=(32,32,3)\n",
    ")\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "i = 0\n",
      "i = 1\n",
      "i = 2\n",
      "i = 3\n",
      "i = 4\n",
      "i = 5\n",
      "i = 6\n",
      "i = 7\n",
      "i = 8\n",
      "i = 9\n",
      "i = 10\n",
      "i = 11\n",
      "i = 12\n",
      "i = 13\n",
      "i = 14\n",
      "i = 15\n",
      "i = 16\n",
      "i = 17\n",
      "i = 18\n",
      "i = 19\n",
      "i = 20\n",
      "i = 21\n",
      "i = 22\n",
      "i = 23\n",
      "i = 24\n",
      "i = 25\n",
      "i = 26\n",
      "i = 27\n",
      "i = 28\n",
      "i = 29\n",
      "i = 30\n",
      "i = 31\n",
      "i = 32\n",
      "i = 33\n",
      "i = 34\n",
      "i = 35\n",
      "i = 36\n",
      "i = 37\n",
      "i = 38\n",
      "i = 39\n",
      "i = 40\n",
      "i = 41\n",
      "i = 42\n",
      "i = 43\n",
      "i = 44\n",
      "i = 45\n",
      "i = 46\n",
      "i = 47\n",
      "i = 48\n",
      "i = 49\n",
      "i = 50\n",
      "i = 51\n",
      "i = 52\n",
      "i = 53\n",
      "i = 54\n",
      "i = 55\n",
      "i = 56\n",
      "i = 57\n",
      "i = 58\n",
      "i = 59\n",
      "i = 60\n",
      "i = 61\n",
      "i = 62\n",
      "i = 63\n",
      "i = 64\n",
      "i = 65\n",
      "i = 66\n",
      "i = 67\n",
      "i = 68\n",
      "i = 69\n",
      "i = 70\n",
      "i = 71\n",
      "i = 72\n",
      "i = 73\n",
      "i = 74\n",
      "i = 75\n",
      "i = 76\n",
      "i = 77\n",
      "i = 78\n",
      "i = 79\n",
      "i = 80\n",
      "i = 81\n",
      "i = 82\n",
      "i = 83\n",
      "i = 84\n",
      "i = 85\n",
      "i = 86\n",
      "i = 87\n",
      "i = 88\n",
      "i = 89\n",
      "i = 90\n",
      "i = 91\n",
      "i = 92\n",
      "i = 93\n",
      "i = 94\n",
      "i = 95\n",
      "i = 96\n",
      "i = 97\n",
      "i = 98\n",
      "i = 99\n",
      "i = 100\n",
      "i = 101\n",
      "i = 102\n",
      "i = 103\n",
      "i = 104\n",
      "i = 105\n",
      "i = 106\n",
      "i = 107\n",
      "i = 108\n",
      "i = 109\n",
      "i = 110\n",
      "i = 111\n",
      "i = 112\n",
      "i = 113\n",
      "i = 114\n",
      "i = 115\n",
      "i = 116\n",
      "i = 117\n",
      "i = 118\n",
      "i = 119\n",
      "i = 120\n",
      "i = 121\n",
      "i = 122\n",
      "i = 123\n",
      "i = 124\n",
      "i = 125\n",
      "i = 126\n",
      "i = 127\n",
      "i = 128\n",
      "i = 129\n",
      "i = 130\n",
      "i = 131\n",
      "i = 132\n",
      "i = 133\n",
      "i = 134\n",
      "i = 135\n",
      "i = 136\n",
      "i = 137\n",
      "i = 138\n",
      "i = 139\n",
      "i = 140\n",
      "i = 141\n",
      "i = 142\n",
      "i = 143\n",
      "i = 144\n",
      "i = 145\n",
      "i = 146\n",
      "i = 147\n",
      "i = 148\n",
      "i = 149\n",
      "i = 150\n",
      "i = 151\n",
      "i = 152\n",
      "i = 153\n",
      "i = 154\n",
      "i = 155\n",
      "i = 156\n",
      "i = 157\n",
      "i = 158\n",
      "i = 159\n",
      "i = 160\n",
      "i = 161\n",
      "i = 162\n",
      "i = 163\n",
      "i = 164\n",
      "i = 165\n",
      "i = 166\n",
      "i = 167\n",
      "i = 168\n",
      "i = 169\n",
      "i = 170\n",
      "i = 171\n",
      "i = 172\n",
      "i = 173\n",
      "i = 174\n",
      "i = 175\n",
      "i = 176\n",
      "i = 177\n",
      "i = 178\n",
      "i = 179\n",
      "i = 180\n",
      "i = 181\n",
      "i = 182\n",
      "i = 183\n",
      "i = 184\n",
      "i = 185\n",
      "i = 186\n",
      "i = 187\n",
      "i = 188\n",
      "i = 189\n",
      "i = 190\n",
      "i = 191\n",
      "i = 192\n",
      "i = 193\n",
      "i = 194\n",
      "i = 195\n",
      "i = 196\n",
      "i = 197\n",
      "i = 198\n",
      "i = 199\n",
      "i = 200\n",
      "i = 201\n",
      "i = 202\n",
      "i = 203\n",
      "i = 204\n",
      "i = 205\n",
      "i = 206\n",
      "i = 207\n",
      "i = 208\n",
      "i = 209\n",
      "i = 210\n",
      "i = 211\n",
      "i = 212\n",
      "i = 213\n",
      "i = 214\n",
      "i = 215\n",
      "i = 216\n",
      "i = 217\n",
      "i = 218\n",
      "i = 219\n",
      "i = 220\n",
      "i = 221\n",
      "i = 222\n",
      "i = 223\n",
      "i = 224\n",
      "i = 225\n",
      "i = 226\n",
      "i = 227\n",
      "i = 228\n",
      "i = 229\n",
      "i = 230\n",
      "i = 231\n",
      "i = 232\n",
      "i = 233\n",
      "i = 234\n",
      "i = 235\n",
      "i = 236\n",
      "i = 237\n",
      "i = 238\n",
      "i = 239\n",
      "i = 240\n",
      "i = 241\n",
      "i = 242\n",
      "i = 243\n",
      "i = 244\n",
      "i = 245\n",
      "i = 246\n",
      "i = 247\n",
      "i = 248\n",
      "i = 249\n",
      "i = 250\n",
      "i = 251\n",
      "i = 252\n",
      "i = 253\n",
      "i = 254\n",
      "i = 255\n",
      "i = 256\n",
      "i = 257\n",
      "i = 258\n",
      "i = 259\n",
      "i = 260\n",
      "i = 261\n",
      "i = 262\n",
      "i = 263\n",
      "i = 264\n",
      "i = 265\n",
      "i = 266\n",
      "i = 267\n",
      "i = 268\n",
      "i = 269\n",
      "i = 270\n",
      "i = 271\n",
      "i = 272\n",
      "i = 273\n",
      "i = 274\n",
      "i = 275\n",
      "i = 276\n",
      "i = 277\n",
      "i = 278\n",
      "i = 279\n",
      "i = 280\n",
      "i = 281\n",
      "i = 282\n",
      "i = 283\n",
      "i = 284\n",
      "i = 285\n",
      "i = 286\n",
      "i = 287\n",
      "i = 288\n",
      "i = 289\n",
      "i = 290\n",
      "i = 291\n",
      "i = 292\n",
      "i = 293\n",
      "i = 294\n",
      "i = 295\n",
      "i = 296\n",
      "i = 297\n",
      "i = 298\n",
      "i = 299\n",
      "i = 300\n",
      "i = 301\n",
      "i = 302\n",
      "i = 303\n",
      "i = 304\n",
      "i = 305\n",
      "i = 306\n",
      "i = 307\n",
      "i = 308\n",
      "i = 309\n",
      "i = 310\n",
      "i = 311\n",
      "i = 312\n",
      "i = 313\n",
      "i = 314\n",
      "i = 315\n",
      "i = 316\n",
      "i = 317\n",
      "i = 318\n",
      "i = 319\n",
      "i = 320\n",
      "i = 321\n",
      "i = 322\n",
      "i = 323\n",
      "i = 324\n",
      "i = 325\n",
      "i = 326\n",
      "i = 327\n",
      "i = 328\n",
      "i = 329\n",
      "i = 330\n",
      "i = 331\n",
      "i = 332\n",
      "i = 333\n",
      "i = 334\n",
      "i = 335\n",
      "i = 336\n",
      "i = 337\n",
      "i = 338\n",
      "i = 339\n",
      "i = 340\n",
      "i = 341\n",
      "i = 342\n",
      "i = 343\n",
      "i = 344\n",
      "i = 345\n",
      "i = 346\n",
      "i = 347\n",
      "i = 348\n",
      "i = 349\n",
      "i = 350\n",
      "i = 351\n",
      "i = 352\n",
      "i = 353\n",
      "i = 354\n",
      "i = 355\n",
      "i = 356\n",
      "i = 357\n",
      "i = 358\n",
      "i = 359\n",
      "i = 360\n",
      "i = 361\n",
      "i = 362\n",
      "i = 363\n",
      "i = 364\n",
      "i = 365\n",
      "i = 366\n",
      "i = 367\n",
      "i = 368\n",
      "i = 369\n",
      "i = 370\n",
      "i = 371\n",
      "i = 372\n",
      "i = 373\n",
      "i = 374\n",
      "i = 375\n",
      "i = 376\n",
      "i = 377\n",
      "i = 378\n",
      "i = 379\n",
      "i = 380\n",
      "i = 381\n",
      "i = 382\n",
      "i = 383\n",
      "i = 384\n",
      "i = 385\n",
      "i = 386\n",
      "i = 387\n",
      "i = 388\n",
      "i = 389\n",
      "Epoch 000: Loss: 2.331, Accuracy: 13.041%\n",
      "i = 0\n",
      "i = 1\n",
      "i = 2\n",
      "i = 3\n",
      "i = 4\n",
      "i = 5\n",
      "i = 6\n",
      "i = 7\n",
      "i = 8\n",
      "i = 9\n",
      "i = 10\n",
      "i = 11\n",
      "i = 12\n",
      "i = 13\n",
      "i = 14\n",
      "i = 15\n",
      "i = 16\n",
      "i = 17\n",
      "i = 18\n",
      "i = 19\n",
      "i = 20\n",
      "i = 21\n",
      "i = 22\n",
      "i = 23\n",
      "i = 24\n",
      "i = 25\n",
      "i = 26\n",
      "i = 27\n",
      "i = 28\n",
      "i = 29\n",
      "i = 30\n",
      "i = 31\n",
      "i = 32\n",
      "i = 33\n",
      "i = 34\n",
      "i = 35\n",
      "i = 36\n",
      "i = 37\n",
      "i = 38\n",
      "i = 39\n",
      "i = 40\n",
      "i = 41\n",
      "i = 42\n",
      "i = 43\n",
      "i = 44\n",
      "i = 45\n",
      "i = 46\n",
      "i = 47\n",
      "i = 48\n",
      "i = 49\n",
      "i = 50\n",
      "i = 51\n",
      "i = 52\n",
      "i = 53\n",
      "i = 54\n",
      "i = 55\n",
      "i = 56\n",
      "i = 57\n",
      "i = 58\n",
      "i = 59\n",
      "i = 60\n",
      "i = 61\n",
      "i = 62\n",
      "i = 63\n",
      "i = 64\n",
      "i = 65\n",
      "i = 66\n",
      "i = 67\n",
      "i = 68\n",
      "i = 69\n",
      "i = 70\n",
      "i = 71\n",
      "i = 72\n",
      "i = 73\n",
      "i = 74\n",
      "i = 75\n",
      "i = 76\n",
      "i = 77\n",
      "i = 78\n",
      "i = 79\n",
      "i = 80\n",
      "i = 81\n",
      "i = 82\n",
      "i = 83\n",
      "i = 84\n",
      "i = 85\n",
      "i = 86\n",
      "i = 87\n",
      "i = 88\n",
      "i = 89\n",
      "i = 90\n",
      "i = 91\n",
      "i = 92\n",
      "i = 93\n",
      "i = 94\n",
      "i = 95\n",
      "i = 96\n",
      "i = 97\n",
      "i = 98\n",
      "i = 99\n",
      "i = 100\n",
      "i = 101\n",
      "i = 102\n",
      "i = 103\n",
      "i = 104\n",
      "i = 105\n",
      "i = 106\n",
      "i = 107\n",
      "i = 108\n",
      "i = 109\n",
      "i = 110\n",
      "i = 111\n",
      "i = 112\n",
      "i = 113\n",
      "i = 114\n",
      "i = 115\n",
      "i = 116\n",
      "i = 117\n",
      "i = 118\n",
      "i = 119\n",
      "i = 120\n",
      "i = 121\n",
      "i = 122\n",
      "i = 123\n",
      "i = 124\n",
      "i = 125\n",
      "i = 126\n",
      "i = 127\n",
      "i = 128\n",
      "i = 129\n",
      "i = 130\n",
      "i = 131\n",
      "i = 132\n",
      "i = 133\n",
      "i = 134\n",
      "i = 135\n",
      "i = 136\n",
      "i = 137\n",
      "i = 138\n",
      "i = 139\n",
      "i = 140\n",
      "i = 141\n",
      "i = 142\n",
      "i = 143\n",
      "i = 144\n",
      "i = 145\n",
      "i = 146\n",
      "i = 147\n",
      "i = 148\n",
      "i = 149\n",
      "i = 150\n",
      "i = 151\n",
      "i = 152\n",
      "i = 153\n",
      "i = 154\n",
      "i = 155\n",
      "i = 156\n",
      "i = 157\n",
      "i = 158\n",
      "i = 159\n",
      "i = 160\n",
      "i = 161\n",
      "i = 162\n",
      "i = 163\n",
      "i = 164\n",
      "i = 165\n",
      "i = 166\n",
      "i = 167\n",
      "i = 168\n",
      "i = 169\n",
      "i = 170\n",
      "i = 171\n",
      "i = 172\n",
      "i = 173\n",
      "i = 174\n",
      "i = 175\n",
      "i = 176\n",
      "i = 177\n",
      "i = 178\n",
      "i = 179\n",
      "i = 180\n",
      "i = 181\n",
      "i = 182\n",
      "i = 183\n",
      "i = 184\n",
      "i = 185\n",
      "i = 186\n",
      "i = 187\n",
      "i = 188\n",
      "i = 189\n",
      "i = 190\n",
      "i = 191\n",
      "i = 192\n",
      "i = 193\n",
      "i = 194\n",
      "i = 195\n",
      "i = 196\n",
      "i = 197\n",
      "i = 198\n",
      "i = 199\n",
      "i = 200\n",
      "i = 201\n",
      "i = 202\n",
      "i = 203\n",
      "i = 204\n",
      "i = 205\n",
      "i = 206\n",
      "i = 207\n",
      "i = 208\n",
      "i = 209\n",
      "i = 210\n",
      "i = 211\n",
      "i = 212\n",
      "i = 213\n",
      "i = 214\n",
      "i = 215\n",
      "i = 216\n",
      "i = 217\n",
      "i = 218\n",
      "i = 219\n",
      "i = 220\n",
      "i = 221\n",
      "i = 222\n",
      "i = 223\n",
      "i = 224\n",
      "i = 225\n",
      "i = 226\n",
      "i = 227\n",
      "i = 228\n",
      "i = 229\n",
      "i = 230\n",
      "i = 231\n",
      "i = 232\n",
      "i = 233\n",
      "i = 234\n",
      "i = 235\n",
      "i = 236\n",
      "i = 237\n",
      "i = 238\n",
      "i = 239\n",
      "i = 240\n",
      "i = 241\n",
      "i = 242\n",
      "i = 243\n",
      "i = 244\n",
      "i = 245\n",
      "i = 246\n",
      "i = 247\n",
      "i = 248\n",
      "i = 249\n",
      "i = 250\n",
      "i = 251\n",
      "i = 252\n",
      "i = 253\n",
      "i = 254\n",
      "i = 255\n",
      "i = 256\n",
      "i = 257\n",
      "i = 258\n",
      "i = 259\n",
      "i = 260\n",
      "i = 261\n",
      "i = 262\n",
      "i = 263\n",
      "i = 264\n",
      "i = 265\n",
      "i = 266\n",
      "i = 267\n",
      "i = 268\n",
      "i = 269\n",
      "i = 270\n",
      "i = 271\n",
      "i = 272\n",
      "i = 273\n",
      "i = 274\n",
      "i = 275\n",
      "i = 276\n",
      "i = 277\n",
      "i = 278\n",
      "i = 279\n",
      "i = 280\n",
      "i = 281\n",
      "i = 282\n",
      "i = 283\n",
      "i = 284\n",
      "i = 285\n",
      "i = 286\n",
      "i = 287\n",
      "i = 288\n",
      "i = 289\n",
      "i = 290\n",
      "i = 291\n",
      "i = 292\n",
      "i = 293\n",
      "i = 294\n",
      "i = 295\n",
      "i = 296\n",
      "i = 297\n",
      "i = 298\n",
      "i = 299\n",
      "i = 300\n",
      "i = 301\n",
      "i = 302\n",
      "i = 303\n",
      "i = 304\n",
      "i = 305\n",
      "i = 306\n",
      "i = 307\n",
      "i = 308\n",
      "i = 309\n",
      "i = 310\n",
      "i = 311\n",
      "i = 312\n",
      "i = 313\n",
      "i = 314\n",
      "i = 315\n",
      "i = 316\n",
      "i = 317\n",
      "i = 318\n",
      "i = 319\n",
      "i = 320\n",
      "i = 321\n",
      "i = 322\n",
      "i = 323\n",
      "i = 324\n",
      "i = 325\n",
      "i = 326\n",
      "i = 327\n",
      "i = 328\n",
      "i = 329\n",
      "i = 330\n",
      "i = 331\n",
      "i = 332\n",
      "i = 333\n",
      "i = 334\n",
      "i = 335\n",
      "i = 336\n",
      "i = 337\n",
      "i = 338\n",
      "i = 339\n",
      "i = 340\n",
      "i = 341\n",
      "i = 342\n",
      "i = 343\n",
      "i = 344\n",
      "i = 345\n",
      "i = 346\n",
      "i = 347\n",
      "i = 348\n",
      "i = 349\n",
      "i = 350\n",
      "i = 351\n",
      "i = 352\n",
      "i = 353\n",
      "i = 354\n",
      "i = 355\n",
      "i = 356\n",
      "i = 357\n",
      "i = 358\n",
      "i = 359\n",
      "i = 360\n",
      "i = 361\n",
      "i = 362\n",
      "i = 363\n",
      "i = 364\n",
      "i = 365\n",
      "i = 366\n",
      "i = 367\n",
      "i = 368\n",
      "i = 369\n",
      "i = 370\n",
      "i = 371\n",
      "i = 372\n",
      "i = 373\n",
      "i = 374\n",
      "i = 375\n",
      "i = 376\n",
      "i = 377\n",
      "i = 378\n",
      "i = 379\n",
      "i = 380\n",
      "i = 381\n",
      "i = 382\n",
      "i = 383\n",
      "i = 384\n",
      "i = 385\n",
      "i = 386\n",
      "i = 387\n",
      "i = 388\n",
      "i = 389\n",
      "Epoch 001: Loss: 2.331, Accuracy: 13.041%\n",
      "i = 0\n",
      "i = 1\n",
      "i = 2\n",
      "i = 3\n",
      "i = 4\n",
      "i = 5\n",
      "i = 6\n",
      "i = 7\n",
      "i = 8\n",
      "i = 9\n",
      "i = 10\n",
      "i = 11\n",
      "i = 12\n",
      "i = 13\n",
      "i = 14\n",
      "i = 15\n",
      "i = 16\n",
      "i = 17\n",
      "i = 18\n",
      "i = 19\n",
      "i = 20\n",
      "i = 21\n",
      "i = 22\n",
      "i = 23\n",
      "i = 24\n",
      "i = 25\n",
      "i = 26\n",
      "i = 27\n",
      "i = 28\n",
      "i = 29\n",
      "i = 30\n",
      "i = 31\n",
      "i = 32\n",
      "i = 33\n",
      "i = 34\n",
      "i = 35\n",
      "i = 36\n",
      "i = 37\n",
      "i = 38\n",
      "i = 39\n",
      "i = 40\n",
      "i = 41\n",
      "i = 42\n",
      "i = 43\n",
      "i = 44\n",
      "i = 45\n",
      "i = 46\n",
      "i = 47\n",
      "i = 48\n",
      "i = 49\n",
      "i = 50\n",
      "i = 51\n",
      "i = 52\n",
      "i = 53\n",
      "i = 54\n",
      "i = 55\n",
      "i = 56\n",
      "i = 57\n",
      "i = 58\n",
      "i = 59\n",
      "i = 60\n",
      "i = 61\n",
      "i = 62\n",
      "i = 63\n",
      "i = 64\n",
      "i = 65\n",
      "i = 66\n",
      "i = 67\n",
      "i = 68\n",
      "i = 69\n",
      "i = 70\n",
      "i = 71\n",
      "i = 72\n",
      "i = 73\n",
      "i = 74\n",
      "i = 75\n",
      "i = 76\n",
      "i = 77\n",
      "i = 78\n",
      "i = 79\n",
      "i = 80\n",
      "i = 81\n",
      "i = 82\n",
      "i = 83\n",
      "i = 84\n",
      "i = 85\n",
      "i = 86\n",
      "i = 87\n",
      "i = 88\n",
      "i = 89\n",
      "i = 90\n",
      "i = 91\n",
      "i = 92\n",
      "i = 93\n",
      "i = 94\n",
      "i = 95\n",
      "i = 96\n",
      "i = 97\n",
      "i = 98\n",
      "i = 99\n",
      "i = 100\n",
      "i = 101\n",
      "i = 102\n",
      "i = 103\n",
      "i = 104\n",
      "i = 105\n",
      "i = 106\n",
      "i = 107\n",
      "i = 108\n",
      "i = 109\n",
      "i = 110\n",
      "i = 111\n",
      "i = 112\n",
      "i = 113\n",
      "i = 114\n",
      "i = 115\n",
      "i = 116\n",
      "i = 117\n",
      "i = 118\n",
      "i = 119\n",
      "i = 120\n",
      "i = 121\n",
      "i = 122\n",
      "i = 123\n",
      "i = 124\n",
      "i = 125\n",
      "i = 126\n",
      "i = 127\n",
      "i = 128\n",
      "i = 129\n",
      "i = 130\n",
      "i = 131\n",
      "i = 132\n",
      "i = 133\n",
      "i = 134\n",
      "i = 135\n",
      "i = 136\n",
      "i = 137\n",
      "i = 138\n",
      "i = 139\n",
      "i = 140\n",
      "i = 141\n",
      "i = 142\n",
      "i = 143\n",
      "i = 144\n",
      "i = 145\n",
      "i = 146\n",
      "i = 147\n",
      "i = 148\n",
      "i = 149\n",
      "i = 150\n",
      "i = 151\n",
      "i = 152\n",
      "i = 153\n",
      "i = 154\n",
      "i = 155\n",
      "i = 156\n",
      "i = 157\n",
      "i = 158\n",
      "i = 159\n",
      "i = 160\n",
      "i = 161\n",
      "i = 162\n",
      "i = 163\n",
      "i = 164\n",
      "i = 165\n",
      "i = 166\n",
      "i = 167\n",
      "i = 168\n",
      "i = 169\n",
      "i = 170\n",
      "i = 171\n",
      "i = 172\n",
      "i = 173\n",
      "i = 174\n",
      "i = 175\n",
      "i = 176\n",
      "i = 177\n",
      "i = 178\n",
      "i = 179\n",
      "i = 180\n",
      "i = 181\n",
      "i = 182\n",
      "i = 183\n",
      "i = 184\n",
      "i = 185\n",
      "i = 186\n",
      "i = 187\n",
      "i = 188\n",
      "i = 189\n",
      "i = 190\n",
      "i = 191\n",
      "i = 192\n",
      "i = 193\n",
      "i = 194\n",
      "i = 195\n",
      "i = 196\n",
      "i = 197\n",
      "i = 198\n",
      "i = 199\n",
      "i = 200\n",
      "i = 201\n",
      "i = 202\n",
      "i = 203\n",
      "i = 204\n",
      "i = 205\n",
      "i = 206\n",
      "i = 207\n",
      "i = 208\n",
      "i = 209\n",
      "i = 210\n",
      "i = 211\n",
      "i = 212\n",
      "i = 213\n",
      "i = 214\n",
      "i = 215\n",
      "i = 216\n",
      "i = 217\n",
      "i = 218\n",
      "i = 219\n",
      "i = 220\n",
      "i = 221\n",
      "i = 222\n",
      "i = 223\n",
      "i = 224\n",
      "i = 225\n",
      "i = 226\n",
      "i = 227\n",
      "i = 228\n",
      "i = 229\n",
      "i = 230\n",
      "i = 231\n",
      "i = 232\n",
      "i = 233\n",
      "i = 234\n",
      "i = 235\n",
      "i = 236\n",
      "i = 237\n",
      "i = 238\n",
      "i = 239\n",
      "i = 240\n",
      "i = 241\n",
      "i = 242\n",
      "i = 243\n",
      "i = 244\n",
      "i = 245\n",
      "i = 246\n",
      "i = 247\n",
      "i = 248\n",
      "i = 249\n",
      "i = 250\n",
      "i = 251\n",
      "i = 252\n",
      "i = 253\n",
      "i = 254\n",
      "i = 255\n",
      "i = 256\n",
      "i = 257\n",
      "i = 258\n",
      "i = 259\n",
      "i = 260\n",
      "i = 261\n",
      "i = 262\n",
      "i = 263\n",
      "i = 264\n",
      "i = 265\n",
      "i = 266\n",
      "i = 267\n",
      "i = 268\n",
      "i = 269\n",
      "i = 270\n",
      "i = 271\n",
      "i = 272\n",
      "i = 273\n",
      "i = 274\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "i = 275\n",
      "i = 276\n",
      "i = 277\n",
      "i = 278\n",
      "i = 279\n",
      "i = 280\n",
      "i = 281\n",
      "i = 282\n",
      "i = 283\n",
      "i = 284\n",
      "i = 285\n",
      "i = 286\n",
      "i = 287\n",
      "i = 288\n",
      "i = 289\n",
      "i = 290\n",
      "i = 291\n",
      "i = 292\n",
      "i = 293\n",
      "i = 294\n",
      "i = 295\n",
      "i = 296\n",
      "i = 297\n",
      "i = 298\n",
      "i = 299\n",
      "i = 300\n",
      "i = 301\n",
      "i = 302\n",
      "i = 303\n",
      "i = 304\n",
      "i = 305\n",
      "i = 306\n",
      "i = 307\n",
      "i = 308\n",
      "i = 309\n",
      "i = 310\n",
      "i = 311\n",
      "i = 312\n",
      "i = 313\n",
      "i = 314\n",
      "i = 315\n",
      "i = 316\n",
      "i = 317\n",
      "i = 318\n",
      "i = 319\n",
      "i = 320\n",
      "i = 321\n",
      "i = 322\n",
      "i = 323\n",
      "i = 324\n",
      "i = 325\n",
      "i = 326\n",
      "i = 327\n",
      "i = 328\n",
      "i = 329\n",
      "i = 330\n",
      "i = 331\n",
      "i = 332\n",
      "i = 333\n",
      "i = 334\n",
      "i = 335\n",
      "i = 336\n",
      "i = 337\n",
      "i = 338\n",
      "i = 339\n",
      "i = 340\n",
      "i = 341\n",
      "i = 342\n",
      "i = 343\n",
      "i = 344\n",
      "i = 345\n",
      "i = 346\n",
      "i = 347\n",
      "i = 348\n",
      "i = 349\n",
      "i = 350\n",
      "i = 351\n",
      "i = 352\n",
      "i = 353\n",
      "i = 354\n",
      "i = 355\n",
      "i = 356\n",
      "i = 357\n",
      "i = 358\n",
      "i = 359\n",
      "i = 360\n",
      "i = 361\n",
      "i = 362\n",
      "i = 363\n",
      "i = 364\n",
      "i = 365\n",
      "i = 366\n",
      "i = 367\n",
      "i = 368\n",
      "i = 369\n",
      "i = 370\n"
     ]
    }
   ],
   "source": [
    "loss_object = tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True)\n",
    "optimizer = tf.keras.optimizers.Adam(learning_rate=0.01)\n",
    "\n",
    "adam_train_loss_results = []\n",
    "adam_train_accuracy_results = []\n",
    "\n",
    "epochs = 250\n",
    "batch_size = 128\n",
    "\n",
    "for epoch in range(epochs):\n",
    "    epoch_loss_avg = tf.keras.metrics.Mean()\n",
    "    epoch_accuracy = tf.keras.metrics.SparseCategoricalAccuracy()\n",
    "\n",
    "    # Training loop - using batches of 128\n",
    "    for i in range(train_images.shape[0] // batch_size):\n",
    "        print(\"i = {0}\".format(i))\n",
    "        x = train_images[i:i+batch_size]\n",
    "        y = train_labels[i:i+batch_size]\n",
    "        # Optimize the model\n",
    "        loss_value, grads = grad(model, x, y)\n",
    "        optimizer.apply_gradients(zip(grads, model.trainable_weights))\n",
    "        \n",
    "\n",
    "        # Track progress\n",
    "        epoch_loss_avg(loss_value)  # Add current batch loss\n",
    "        # Compare predicted label to actual label\n",
    "        epoch_accuracy(y, model(x))\n",
    "\n",
    "        # End epoch\n",
    "    adam_train_loss_results.append(epoch_loss_avg.result())\n",
    "    adam_train_accuracy_results.append(epoch_accuracy.result())\n",
    "\n",
    "    print(\"Epoch {:03d}: Loss: {:.3f}, Accuracy: {:.3%}\".format(epoch,\n",
    "                                                                epoch_loss_avg.result(),\n",
    "                                                                epoch_accuracy.result()))"
   ]
  }
 ],
 "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": "3.8.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}