2장. 3절. 콜백을 이용한 딥러닝 모형 관리 및 모니터링.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "fef2f373",
   "metadata": {},
   "source": [
    "# 콜백을 이용한 딥러닝 모델 관리 및 모니터링"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "fa3816ba",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-11-29T06:45:32.416870Z",
     "start_time": "2021-11-29T06:45:30.250953Z"
    }
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "mnist = tf.keras.datasets.mnist\n",
    "(X_train, y_train),(X_test, y_test) = mnist.load_data()\n",
    "X_train, X_test = X_train / 255.0, X_test / 255.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "b6e12b77",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(60000, 28, 28)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "318e857b",
   "metadata": {},
   "source": [
    "## 체크포인트 콜백"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "7cd943e4",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-11-29T06:45:32.424184Z",
     "start_time": "2021-11-29T06:45:32.417784Z"
    }
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.callbacks import ModelCheckpoint\n",
    "path = 'mnist_keras_model.keras'\n",
    "checkpoint = ModelCheckpoint(filepath=path,\n",
    "                             monitor='val_accuracy',\n",
    "                             save_best_only=True,\n",
    "                             verbose=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "18fc7fd9",
   "metadata": {},
   "source": [
    "## 조기종료 콜백"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "95f63307",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-11-29T06:45:32.427815Z",
     "start_time": "2021-11-29T06:45:32.425911Z"
    }
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "early_stopping=EarlyStopping(monitor='val_accuracy',\n",
    "                             patience=3)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "08403037",
   "metadata": {},
   "source": [
    "## 텐서보드 콜백"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "015fe23d",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-11-29T06:45:32.432337Z",
     "start_time": "2021-11-29T06:45:32.429319Z"
    }
   },
   "outputs": [],
   "source": [
    "from tensorflow.keras.callbacks import TensorBoard\n",
    "tensor_board = TensorBoard(log_dir=\"tensor_log\",\n",
    "                           embeddings_freq=1,\n",
    "                           histogram_freq=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2956e80e",
   "metadata": {},
   "source": [
    "## 모델 정의"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "b20eb56f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Input, Flatten, Dense, Dropout\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Input(shape=(28, 28)))\n",
    "model.add(Flatten())\n",
    "model.add(Dense(360, activation='relu'))\n",
    "model.add(Dense(128, activation='relu'))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(Dense(10, activation='softmax'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "3f4bfacc",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-11-29T06:45:32.476609Z",
     "start_time": "2021-11-29T06:45:32.468937Z"
    }
   },
   "outputs": [],
   "source": [
    "model.compile(optimizer='SGD',\n",
    "              loss='sparse_categorical_crossentropy',\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4138ac45",
   "metadata": {},
   "source": [
    "## 모델 학습"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "7dddad7e",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-11-29T06:46:05.478640Z",
     "start_time": "2021-11-29T06:45:32.477387Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.1139 - loss: 2.3045\n",
      "Epoch 1: val_accuracy improved from -inf to 0.44658, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 21ms/step - accuracy: 0.1152 - loss: 2.3029 - val_accuracy: 0.4466 - val_loss: 2.0496\n",
      "Epoch 2/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 7ms/step - accuracy: 0.4141 - loss: 2.0053\n",
      "Epoch 2: val_accuracy improved from 0.44658 to 0.67533, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.4204 - loss: 1.9976 - val_accuracy: 0.6753 - val_loss: 1.7431\n",
      "Epoch 3/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.6024 - loss: 1.7071\n",
      "Epoch 3: val_accuracy improved from 0.67533 to 0.74733, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.6038 - loss: 1.7027 - val_accuracy: 0.7473 - val_loss: 1.4378\n",
      "Epoch 4/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.6714 - loss: 1.4341\n",
      "Epoch 4: val_accuracy improved from 0.74733 to 0.78658, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.6725 - loss: 1.4290 - val_accuracy: 0.7866 - val_loss: 1.1790\n",
      "Epoch 5/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.7158 - loss: 1.2110\n",
      "Epoch 5: val_accuracy improved from 0.78658 to 0.81258, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.7164 - loss: 1.2078 - val_accuracy: 0.8126 - val_loss: 0.9836\n",
      "Epoch 6/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.7469 - loss: 1.0399\n",
      "Epoch 6: val_accuracy improved from 0.81258 to 0.83283, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.7473 - loss: 1.0377 - val_accuracy: 0.8328 - val_loss: 0.8420\n",
      "Epoch 7/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 7ms/step - accuracy: 0.7719 - loss: 0.9104\n",
      "Epoch 7: val_accuracy improved from 0.83283 to 0.84708, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 12ms/step - accuracy: 0.7721 - loss: 0.9095 - val_accuracy: 0.8471 - val_loss: 0.7391\n",
      "Epoch 8/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 14ms/step - accuracy: 0.7885 - loss: 0.8246\n",
      "Epoch 8: val_accuracy improved from 0.84708 to 0.85867, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 21ms/step - accuracy: 0.7886 - loss: 0.8236 - val_accuracy: 0.8587 - val_loss: 0.6632\n",
      "Epoch 9/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.8043 - loss: 0.7516\n",
      "Epoch 9: val_accuracy improved from 0.85867 to 0.86550, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.8044 - loss: 0.7508 - val_accuracy: 0.8655 - val_loss: 0.6056\n",
      "Epoch 10/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.8154 - loss: 0.6970\n",
      "Epoch 10: val_accuracy improved from 0.86550 to 0.87242, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.8155 - loss: 0.6966 - val_accuracy: 0.8724 - val_loss: 0.5610\n",
      "Epoch 11/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 7ms/step - accuracy: 0.8233 - loss: 0.6535\n",
      "Epoch 11: val_accuracy improved from 0.87242 to 0.87775, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.8235 - loss: 0.6528 - val_accuracy: 0.8777 - val_loss: 0.5252\n",
      "Epoch 12/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.8291 - loss: 0.6241\n",
      "Epoch 12: val_accuracy improved from 0.87775 to 0.88258, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.8291 - loss: 0.6238 - val_accuracy: 0.8826 - val_loss: 0.4966\n",
      "Epoch 13/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.8362 - loss: 0.5944\n",
      "Epoch 13: val_accuracy improved from 0.88258 to 0.88683, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.8363 - loss: 0.5940 - val_accuracy: 0.8868 - val_loss: 0.4727\n",
      "Epoch 14/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.8453 - loss: 0.5617\n",
      "Epoch 14: val_accuracy improved from 0.88683 to 0.88925, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.8452 - loss: 0.5617 - val_accuracy: 0.8892 - val_loss: 0.4529\n",
      "Epoch 15/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.8486 - loss: 0.5429\n",
      "Epoch 15: val_accuracy improved from 0.88925 to 0.89175, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.8489 - loss: 0.5425 - val_accuracy: 0.8917 - val_loss: 0.4360\n",
      "Epoch 16/1000\n",
      "\u001b[1m42/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.8556 - loss: 0.5244\n",
      "Epoch 16: val_accuracy improved from 0.89175 to 0.89400, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.8555 - loss: 0.5239 - val_accuracy: 0.8940 - val_loss: 0.4208\n",
      "Epoch 17/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.8577 - loss: 0.5112\n",
      "Epoch 17: val_accuracy improved from 0.89400 to 0.89558, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 20ms/step - accuracy: 0.8577 - loss: 0.5109 - val_accuracy: 0.8956 - val_loss: 0.4083\n",
      "Epoch 18/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.8612 - loss: 0.4952\n",
      "Epoch 18: val_accuracy improved from 0.89558 to 0.89808, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.8613 - loss: 0.4949 - val_accuracy: 0.8981 - val_loss: 0.3970\n",
      "Epoch 19/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.8660 - loss: 0.4852\n",
      "Epoch 19: val_accuracy improved from 0.89808 to 0.89975, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.8661 - loss: 0.4850 - val_accuracy: 0.8997 - val_loss: 0.3872\n",
      "Epoch 20/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.8667 - loss: 0.4677\n",
      "Epoch 20: val_accuracy improved from 0.89975 to 0.90100, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.8667 - loss: 0.4677 - val_accuracy: 0.9010 - val_loss: 0.3782\n",
      "Epoch 21/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.8688 - loss: 0.4574\n",
      "Epoch 21: val_accuracy improved from 0.90100 to 0.90242, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.8689 - loss: 0.4575 - val_accuracy: 0.9024 - val_loss: 0.3699\n",
      "Epoch 22/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.8696 - loss: 0.4586\n",
      "Epoch 22: val_accuracy improved from 0.90242 to 0.90492, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.8697 - loss: 0.4582 - val_accuracy: 0.9049 - val_loss: 0.3627\n",
      "Epoch 23/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.8754 - loss: 0.4415\n",
      "Epoch 23: val_accuracy improved from 0.90492 to 0.90625, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.8754 - loss: 0.4415 - val_accuracy: 0.9062 - val_loss: 0.3559\n",
      "Epoch 24/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.8745 - loss: 0.4434\n",
      "Epoch 24: val_accuracy improved from 0.90625 to 0.90750, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.8748 - loss: 0.4426 - val_accuracy: 0.9075 - val_loss: 0.3497\n",
      "Epoch 25/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.8772 - loss: 0.4280\n",
      "Epoch 25: val_accuracy improved from 0.90750 to 0.90783, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.8775 - loss: 0.4275 - val_accuracy: 0.9078 - val_loss: 0.3441\n",
      "Epoch 26/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 7ms/step - accuracy: 0.8830 - loss: 0.4212\n",
      "Epoch 26: val_accuracy improved from 0.90783 to 0.90883, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 12ms/step - accuracy: 0.8829 - loss: 0.4209 - val_accuracy: 0.9088 - val_loss: 0.3388\n",
      "Epoch 27/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.8860 - loss: 0.4084\n",
      "Epoch 27: val_accuracy improved from 0.90883 to 0.91050, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 20ms/step - accuracy: 0.8858 - loss: 0.4086 - val_accuracy: 0.9105 - val_loss: 0.3337\n",
      "Epoch 28/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.8854 - loss: 0.4081\n",
      "Epoch 28: val_accuracy improved from 0.91050 to 0.91133, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.8853 - loss: 0.4080 - val_accuracy: 0.9113 - val_loss: 0.3294\n",
      "Epoch 29/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.8857 - loss: 0.3994\n",
      "Epoch 29: val_accuracy improved from 0.91133 to 0.91167, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.8857 - loss: 0.3993 - val_accuracy: 0.9117 - val_loss: 0.3250\n",
      "Epoch 30/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.8857 - loss: 0.3958\n",
      "Epoch 30: val_accuracy improved from 0.91167 to 0.91308, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.8858 - loss: 0.3958 - val_accuracy: 0.9131 - val_loss: 0.3210\n",
      "Epoch 31/1000\n",
      "\u001b[1m41/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.8893 - loss: 0.3894 \n",
      "Epoch 31: val_accuracy improved from 0.91308 to 0.91383, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.8893 - loss: 0.3892 - val_accuracy: 0.9138 - val_loss: 0.3172\n",
      "Epoch 32/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.8903 - loss: 0.3837\n",
      "Epoch 32: val_accuracy improved from 0.91383 to 0.91433, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.8903 - loss: 0.3838 - val_accuracy: 0.9143 - val_loss: 0.3136\n",
      "Epoch 33/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.8909 - loss: 0.3808\n",
      "Epoch 33: val_accuracy improved from 0.91433 to 0.91483, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.8910 - loss: 0.3806 - val_accuracy: 0.9148 - val_loss: 0.3100\n",
      "Epoch 34/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.8939 - loss: 0.3781\n",
      "Epoch 34: val_accuracy improved from 0.91483 to 0.91617, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.8939 - loss: 0.3780 - val_accuracy: 0.9162 - val_loss: 0.3069\n",
      "Epoch 35/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.8919 - loss: 0.3731\n",
      "Epoch 35: val_accuracy improved from 0.91617 to 0.91633, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.8920 - loss: 0.3731 - val_accuracy: 0.9163 - val_loss: 0.3037\n",
      "Epoch 36/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.8939 - loss: 0.3688\n",
      "Epoch 36: val_accuracy improved from 0.91633 to 0.91700, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 12ms/step - accuracy: 0.8940 - loss: 0.3687 - val_accuracy: 0.9170 - val_loss: 0.3008\n",
      "Epoch 37/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.8953 - loss: 0.3649\n",
      "Epoch 37: val_accuracy improved from 0.91700 to 0.91842, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.8953 - loss: 0.3649 - val_accuracy: 0.9184 - val_loss: 0.2980\n",
      "Epoch 38/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.8979 - loss: 0.3623\n",
      "Epoch 38: val_accuracy improved from 0.91842 to 0.91875, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.8978 - loss: 0.3622 - val_accuracy: 0.9187 - val_loss: 0.2952\n",
      "Epoch 39/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.8964 - loss: 0.3551\n",
      "Epoch 39: val_accuracy improved from 0.91875 to 0.91908, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.8965 - loss: 0.3551 - val_accuracy: 0.9191 - val_loss: 0.2927\n",
      "Epoch 40/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.8971 - loss: 0.3581\n",
      "Epoch 40: val_accuracy improved from 0.91908 to 0.91992, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.8971 - loss: 0.3579 - val_accuracy: 0.9199 - val_loss: 0.2900\n",
      "Epoch 41/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9024 - loss: 0.3460\n",
      "Epoch 41: val_accuracy improved from 0.91992 to 0.92033, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9023 - loss: 0.3462 - val_accuracy: 0.9203 - val_loss: 0.2876\n",
      "Epoch 42/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9012 - loss: 0.3461\n",
      "Epoch 42: val_accuracy improved from 0.92033 to 0.92067, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9012 - loss: 0.3461 - val_accuracy: 0.9207 - val_loss: 0.2852\n",
      "Epoch 43/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9007 - loss: 0.3451\n",
      "Epoch 43: val_accuracy improved from 0.92067 to 0.92217, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9008 - loss: 0.3450 - val_accuracy: 0.9222 - val_loss: 0.2829\n",
      "Epoch 44/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9038 - loss: 0.3361\n",
      "Epoch 44: val_accuracy improved from 0.92217 to 0.92242, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9038 - loss: 0.3364 - val_accuracy: 0.9224 - val_loss: 0.2806\n",
      "Epoch 45/1000\n",
      "\u001b[1m41/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9034 - loss: 0.3352\n",
      "Epoch 45: val_accuracy improved from 0.92242 to 0.92283, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.9032 - loss: 0.3355 - val_accuracy: 0.9228 - val_loss: 0.2784\n",
      "Epoch 46/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9057 - loss: 0.3351\n",
      "Epoch 46: val_accuracy improved from 0.92283 to 0.92383, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9057 - loss: 0.3350 - val_accuracy: 0.9238 - val_loss: 0.2763\n",
      "Epoch 47/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9049 - loss: 0.3309\n",
      "Epoch 47: val_accuracy improved from 0.92383 to 0.92400, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9048 - loss: 0.3310 - val_accuracy: 0.9240 - val_loss: 0.2743\n",
      "Epoch 48/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9044 - loss: 0.3350\n",
      "Epoch 48: val_accuracy improved from 0.92400 to 0.92417, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9044 - loss: 0.3349 - val_accuracy: 0.9242 - val_loss: 0.2723\n",
      "Epoch 49/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9039 - loss: 0.3327\n",
      "Epoch 49: val_accuracy improved from 0.92417 to 0.92533, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9040 - loss: 0.3324 - val_accuracy: 0.9253 - val_loss: 0.2703\n",
      "Epoch 50/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9062 - loss: 0.3285\n",
      "Epoch 50: val_accuracy improved from 0.92533 to 0.92617, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9064 - loss: 0.3282 - val_accuracy: 0.9262 - val_loss: 0.2687\n",
      "Epoch 51/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9084 - loss: 0.3211\n",
      "Epoch 51: val_accuracy did not improve from 0.92617\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9083 - loss: 0.3212 - val_accuracy: 0.9259 - val_loss: 0.2669\n",
      "Epoch 52/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9114 - loss: 0.3146\n",
      "Epoch 52: val_accuracy improved from 0.92617 to 0.92625, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.9113 - loss: 0.3149 - val_accuracy: 0.9262 - val_loss: 0.2651\n",
      "Epoch 53/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9099 - loss: 0.3170\n",
      "Epoch 53: val_accuracy improved from 0.92625 to 0.92708, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9099 - loss: 0.3170 - val_accuracy: 0.9271 - val_loss: 0.2633\n",
      "Epoch 54/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9107 - loss: 0.3177\n",
      "Epoch 54: val_accuracy improved from 0.92708 to 0.92725, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9107 - loss: 0.3176 - val_accuracy: 0.9273 - val_loss: 0.2615\n",
      "Epoch 55/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9108 - loss: 0.3151\n",
      "Epoch 55: val_accuracy improved from 0.92725 to 0.92792, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.9108 - loss: 0.3151 - val_accuracy: 0.9279 - val_loss: 0.2600\n",
      "Epoch 56/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9132 - loss: 0.3079\n",
      "Epoch 56: val_accuracy improved from 0.92792 to 0.92842, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9131 - loss: 0.3079 - val_accuracy: 0.9284 - val_loss: 0.2583\n",
      "Epoch 57/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9127 - loss: 0.3056\n",
      "Epoch 57: val_accuracy improved from 0.92842 to 0.92867, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9126 - loss: 0.3057 - val_accuracy: 0.9287 - val_loss: 0.2567\n",
      "Epoch 58/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9124 - loss: 0.3088\n",
      "Epoch 58: val_accuracy did not improve from 0.92867\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9124 - loss: 0.3086 - val_accuracy: 0.9286 - val_loss: 0.2551\n",
      "Epoch 59/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9102 - loss: 0.3125\n",
      "Epoch 59: val_accuracy improved from 0.92867 to 0.92967, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.9104 - loss: 0.3118 - val_accuracy: 0.9297 - val_loss: 0.2535\n",
      "Epoch 60/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9128 - loss: 0.3051\n",
      "Epoch 60: val_accuracy did not improve from 0.92967\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.9128 - loss: 0.3050 - val_accuracy: 0.9292 - val_loss: 0.2520\n",
      "Epoch 61/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9134 - loss: 0.3027\n",
      "Epoch 61: val_accuracy improved from 0.92967 to 0.93000, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9134 - loss: 0.3026 - val_accuracy: 0.9300 - val_loss: 0.2504\n",
      "Epoch 62/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9140 - loss: 0.2992\n",
      "Epoch 62: val_accuracy improved from 0.93000 to 0.93025, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9140 - loss: 0.2991 - val_accuracy: 0.9302 - val_loss: 0.2489\n",
      "Epoch 63/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9148 - loss: 0.2971\n",
      "Epoch 63: val_accuracy improved from 0.93025 to 0.93083, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9148 - loss: 0.2972 - val_accuracy: 0.9308 - val_loss: 0.2477\n",
      "Epoch 64/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9129 - loss: 0.3002\n",
      "Epoch 64: val_accuracy improved from 0.93083 to 0.93108, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9129 - loss: 0.3001 - val_accuracy: 0.9311 - val_loss: 0.2462\n",
      "Epoch 65/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9165 - loss: 0.2930\n",
      "Epoch 65: val_accuracy improved from 0.93108 to 0.93158, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.9165 - loss: 0.2930 - val_accuracy: 0.9316 - val_loss: 0.2448\n",
      "Epoch 66/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9189 - loss: 0.2860\n",
      "Epoch 66: val_accuracy did not improve from 0.93158\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9188 - loss: 0.2866 - val_accuracy: 0.9316 - val_loss: 0.2435\n",
      "Epoch 67/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9142 - loss: 0.2925\n",
      "Epoch 67: val_accuracy improved from 0.93158 to 0.93258, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9143 - loss: 0.2924 - val_accuracy: 0.9326 - val_loss: 0.2421\n",
      "Epoch 68/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9173 - loss: 0.2893\n",
      "Epoch 68: val_accuracy improved from 0.93258 to 0.93292, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9174 - loss: 0.2891 - val_accuracy: 0.9329 - val_loss: 0.2409\n",
      "Epoch 69/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9183 - loss: 0.2837\n",
      "Epoch 69: val_accuracy did not improve from 0.93292\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9183 - loss: 0.2837 - val_accuracy: 0.9327 - val_loss: 0.2395\n",
      "Epoch 70/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9150 - loss: 0.2912\n",
      "Epoch 70: val_accuracy improved from 0.93292 to 0.93350, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9153 - loss: 0.2904 - val_accuracy: 0.9335 - val_loss: 0.2384\n",
      "Epoch 71/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9185 - loss: 0.2788\n",
      "Epoch 71: val_accuracy improved from 0.93350 to 0.93458, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9186 - loss: 0.2793 - val_accuracy: 0.9346 - val_loss: 0.2371\n",
      "Epoch 72/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9170 - loss: 0.2896\n",
      "Epoch 72: val_accuracy did not improve from 0.93458\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.9171 - loss: 0.2895 - val_accuracy: 0.9342 - val_loss: 0.2359\n",
      "Epoch 73/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9208 - loss: 0.2791\n",
      "Epoch 73: val_accuracy improved from 0.93458 to 0.93533, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9208 - loss: 0.2791 - val_accuracy: 0.9353 - val_loss: 0.2347\n",
      "Epoch 74/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9198 - loss: 0.2844\n",
      "Epoch 74: val_accuracy improved from 0.93533 to 0.93558, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9198 - loss: 0.2840 - val_accuracy: 0.9356 - val_loss: 0.2334\n",
      "Epoch 75/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9237 - loss: 0.2732\n",
      "Epoch 75: val_accuracy improved from 0.93558 to 0.93608, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9235 - loss: 0.2733 - val_accuracy: 0.9361 - val_loss: 0.2323\n",
      "Epoch 76/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9214 - loss: 0.2714\n",
      "Epoch 76: val_accuracy did not improve from 0.93608\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9214 - loss: 0.2716 - val_accuracy: 0.9357 - val_loss: 0.2311\n",
      "Epoch 77/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.9220 - loss: 0.2760\n",
      "Epoch 77: val_accuracy improved from 0.93608 to 0.93625, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 20ms/step - accuracy: 0.9220 - loss: 0.2759 - val_accuracy: 0.9362 - val_loss: 0.2300\n",
      "Epoch 78/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9216 - loss: 0.2731\n",
      "Epoch 78: val_accuracy improved from 0.93625 to 0.93675, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9216 - loss: 0.2730 - val_accuracy: 0.9367 - val_loss: 0.2289\n",
      "Epoch 79/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9220 - loss: 0.2727\n",
      "Epoch 79: val_accuracy improved from 0.93675 to 0.93683, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9221 - loss: 0.2726 - val_accuracy: 0.9368 - val_loss: 0.2279\n",
      "Epoch 80/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9237 - loss: 0.2643\n",
      "Epoch 80: val_accuracy improved from 0.93683 to 0.93725, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.9237 - loss: 0.2644 - val_accuracy: 0.9373 - val_loss: 0.2266\n",
      "Epoch 81/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9255 - loss: 0.2615\n",
      "Epoch 81: val_accuracy improved from 0.93725 to 0.93800, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9253 - loss: 0.2618 - val_accuracy: 0.9380 - val_loss: 0.2256\n",
      "Epoch 82/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9273 - loss: 0.2601\n",
      "Epoch 82: val_accuracy did not improve from 0.93800\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9272 - loss: 0.2603 - val_accuracy: 0.9379 - val_loss: 0.2246\n",
      "Epoch 83/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9247 - loss: 0.2637\n",
      "Epoch 83: val_accuracy improved from 0.93800 to 0.93867, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9248 - loss: 0.2637 - val_accuracy: 0.9387 - val_loss: 0.2234\n",
      "Epoch 84/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9260 - loss: 0.2610\n",
      "Epoch 84: val_accuracy improved from 0.93867 to 0.93925, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9259 - loss: 0.2612 - val_accuracy: 0.9392 - val_loss: 0.2224\n",
      "Epoch 85/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9243 - loss: 0.2672\n",
      "Epoch 85: val_accuracy did not improve from 0.93925\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9244 - loss: 0.2670 - val_accuracy: 0.9389 - val_loss: 0.2214\n",
      "Epoch 86/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9239 - loss: 0.2631\n",
      "Epoch 86: val_accuracy did not improve from 0.93925\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.9239 - loss: 0.2628 - val_accuracy: 0.9388 - val_loss: 0.2204\n",
      "Epoch 87/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9261 - loss: 0.2569\n",
      "Epoch 87: val_accuracy improved from 0.93925 to 0.93967, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.9260 - loss: 0.2570 - val_accuracy: 0.9397 - val_loss: 0.2194\n",
      "Epoch 88/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9260 - loss: 0.2595\n",
      "Epoch 88: val_accuracy did not improve from 0.93967\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9260 - loss: 0.2594 - val_accuracy: 0.9396 - val_loss: 0.2183\n",
      "Epoch 89/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9260 - loss: 0.2566\n",
      "Epoch 89: val_accuracy improved from 0.93967 to 0.94008, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9261 - loss: 0.2564 - val_accuracy: 0.9401 - val_loss: 0.2174\n",
      "Epoch 90/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9292 - loss: 0.2514 \n",
      "Epoch 90: val_accuracy did not improve from 0.94008\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9290 - loss: 0.2518 - val_accuracy: 0.9401 - val_loss: 0.2163\n",
      "Epoch 91/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9283 - loss: 0.2531\n",
      "Epoch 91: val_accuracy improved from 0.94008 to 0.94092, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.9283 - loss: 0.2531 - val_accuracy: 0.9409 - val_loss: 0.2156\n",
      "Epoch 92/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9289 - loss: 0.2508\n",
      "Epoch 92: val_accuracy did not improve from 0.94092\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.9288 - loss: 0.2509 - val_accuracy: 0.9407 - val_loss: 0.2146\n",
      "Epoch 93/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9287 - loss: 0.2499\n",
      "Epoch 93: val_accuracy improved from 0.94092 to 0.94100, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9287 - loss: 0.2499 - val_accuracy: 0.9410 - val_loss: 0.2136\n",
      "Epoch 94/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9287 - loss: 0.2512\n",
      "Epoch 94: val_accuracy improved from 0.94100 to 0.94125, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.9287 - loss: 0.2512 - val_accuracy: 0.9413 - val_loss: 0.2126\n",
      "Epoch 95/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9287 - loss: 0.2485\n",
      "Epoch 95: val_accuracy improved from 0.94125 to 0.94142, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9287 - loss: 0.2485 - val_accuracy: 0.9414 - val_loss: 0.2116\n",
      "Epoch 96/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9259 - loss: 0.2523\n",
      "Epoch 96: val_accuracy improved from 0.94142 to 0.94233, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9261 - loss: 0.2521 - val_accuracy: 0.9423 - val_loss: 0.2108\n",
      "Epoch 97/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9302 - loss: 0.2428\n",
      "Epoch 97: val_accuracy improved from 0.94233 to 0.94250, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9301 - loss: 0.2432 - val_accuracy: 0.9425 - val_loss: 0.2100\n",
      "Epoch 98/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9297 - loss: 0.2424\n",
      "Epoch 98: val_accuracy improved from 0.94250 to 0.94267, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9297 - loss: 0.2425 - val_accuracy: 0.9427 - val_loss: 0.2092\n",
      "Epoch 99/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9294 - loss: 0.2461\n",
      "Epoch 99: val_accuracy improved from 0.94267 to 0.94283, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.9294 - loss: 0.2461 - val_accuracy: 0.9428 - val_loss: 0.2082\n",
      "Epoch 100/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9307 - loss: 0.2435\n",
      "Epoch 100: val_accuracy improved from 0.94283 to 0.94325, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9307 - loss: 0.2434 - val_accuracy: 0.9433 - val_loss: 0.2074\n",
      "Epoch 101/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9333 - loss: 0.2390\n",
      "Epoch 101: val_accuracy improved from 0.94325 to 0.94358, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9333 - loss: 0.2390 - val_accuracy: 0.9436 - val_loss: 0.2065\n",
      "Epoch 102/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9327 - loss: 0.2415\n",
      "Epoch 102: val_accuracy did not improve from 0.94358\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9327 - loss: 0.2415 - val_accuracy: 0.9435 - val_loss: 0.2057\n",
      "Epoch 103/1000\n",
      "\u001b[1m42/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9316 - loss: 0.2386\n",
      "Epoch 103: val_accuracy improved from 0.94358 to 0.94450, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9314 - loss: 0.2389 - val_accuracy: 0.9445 - val_loss: 0.2049\n",
      "Epoch 104/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9301 - loss: 0.2423\n",
      "Epoch 104: val_accuracy improved from 0.94450 to 0.94458, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9302 - loss: 0.2420 - val_accuracy: 0.9446 - val_loss: 0.2041\n",
      "Epoch 105/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9310 - loss: 0.2399\n",
      "Epoch 105: val_accuracy improved from 0.94458 to 0.94492, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9311 - loss: 0.2398 - val_accuracy: 0.9449 - val_loss: 0.2033\n",
      "Epoch 106/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9290 - loss: 0.2480\n",
      "Epoch 106: val_accuracy improved from 0.94492 to 0.94500, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9291 - loss: 0.2475 - val_accuracy: 0.9450 - val_loss: 0.2025\n",
      "Epoch 107/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9319 - loss: 0.2412\n",
      "Epoch 107: val_accuracy improved from 0.94500 to 0.94550, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9320 - loss: 0.2407 - val_accuracy: 0.9455 - val_loss: 0.2016\n",
      "Epoch 108/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9348 - loss: 0.2281\n",
      "Epoch 108: val_accuracy improved from 0.94550 to 0.94575, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9345 - loss: 0.2289 - val_accuracy: 0.9457 - val_loss: 0.2010\n",
      "Epoch 109/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9331 - loss: 0.2294\n",
      "Epoch 109: val_accuracy improved from 0.94575 to 0.94617, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.9331 - loss: 0.2297 - val_accuracy: 0.9462 - val_loss: 0.2000\n",
      "Epoch 110/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9365 - loss: 0.2253\n",
      "Epoch 110: val_accuracy did not improve from 0.94617\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9364 - loss: 0.2255 - val_accuracy: 0.9460 - val_loss: 0.1993\n",
      "Epoch 111/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.9334 - loss: 0.2338\n",
      "Epoch 111: val_accuracy improved from 0.94617 to 0.94625, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 20ms/step - accuracy: 0.9334 - loss: 0.2336 - val_accuracy: 0.9463 - val_loss: 0.1987\n",
      "Epoch 112/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9364 - loss: 0.2241\n",
      "Epoch 112: val_accuracy improved from 0.94625 to 0.94642, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9363 - loss: 0.2243 - val_accuracy: 0.9464 - val_loss: 0.1978\n",
      "Epoch 113/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9371 - loss: 0.2232 \n",
      "Epoch 113: val_accuracy improved from 0.94642 to 0.94692, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9370 - loss: 0.2234 - val_accuracy: 0.9469 - val_loss: 0.1971\n",
      "Epoch 114/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9352 - loss: 0.2288\n",
      "Epoch 114: val_accuracy improved from 0.94692 to 0.94717, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9352 - loss: 0.2288 - val_accuracy: 0.9472 - val_loss: 0.1963\n",
      "Epoch 115/1000\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.9358 - loss: 0.2193\n",
      "Epoch 115: val_accuracy improved from 0.94717 to 0.94725, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 20ms/step - accuracy: 0.9358 - loss: 0.2194 - val_accuracy: 0.9473 - val_loss: 0.1955\n",
      "Epoch 116/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9361 - loss: 0.2246\n",
      "Epoch 116: val_accuracy improved from 0.94725 to 0.94750, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9360 - loss: 0.2247 - val_accuracy: 0.9475 - val_loss: 0.1947\n",
      "Epoch 117/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9333 - loss: 0.2282\n",
      "Epoch 117: val_accuracy improved from 0.94750 to 0.94783, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9334 - loss: 0.2280 - val_accuracy: 0.9478 - val_loss: 0.1941\n",
      "Epoch 118/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9338 - loss: 0.2272\n",
      "Epoch 118: val_accuracy did not improve from 0.94783\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 13ms/step - accuracy: 0.9340 - loss: 0.2269 - val_accuracy: 0.9476 - val_loss: 0.1935\n",
      "Epoch 119/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9390 - loss: 0.2177\n",
      "Epoch 119: val_accuracy improved from 0.94783 to 0.94800, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 20ms/step - accuracy: 0.9388 - loss: 0.2179 - val_accuracy: 0.9480 - val_loss: 0.1926\n",
      "Epoch 120/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9349 - loss: 0.2259\n",
      "Epoch 120: val_accuracy improved from 0.94800 to 0.94858, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9350 - loss: 0.2258 - val_accuracy: 0.9486 - val_loss: 0.1920\n",
      "Epoch 121/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9361 - loss: 0.2213\n",
      "Epoch 121: val_accuracy improved from 0.94858 to 0.94908, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9361 - loss: 0.2212 - val_accuracy: 0.9491 - val_loss: 0.1912\n",
      "Epoch 122/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 9ms/step - accuracy: 0.9372 - loss: 0.2164\n",
      "Epoch 122: val_accuracy did not improve from 0.94908\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9371 - loss: 0.2166 - val_accuracy: 0.9490 - val_loss: 0.1906\n",
      "Epoch 123/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9370 - loss: 0.2147\n",
      "Epoch 123: val_accuracy improved from 0.94908 to 0.94950, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 15ms/step - accuracy: 0.9370 - loss: 0.2151 - val_accuracy: 0.9495 - val_loss: 0.1899\n",
      "Epoch 124/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9367 - loss: 0.2219\n",
      "Epoch 124: val_accuracy did not improve from 0.94950\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9368 - loss: 0.2216 - val_accuracy: 0.9492 - val_loss: 0.1894\n",
      "Epoch 125/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 11ms/step - accuracy: 0.9379 - loss: 0.2123\n",
      "Epoch 125: val_accuracy did not improve from 0.94950\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9379 - loss: 0.2128 - val_accuracy: 0.9495 - val_loss: 0.1887\n",
      "Epoch 126/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9380 - loss: 0.2193\n",
      "Epoch 126: val_accuracy improved from 0.94950 to 0.94983, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.9380 - loss: 0.2191 - val_accuracy: 0.9498 - val_loss: 0.1880\n",
      "Epoch 127/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 8ms/step - accuracy: 0.9382 - loss: 0.2160\n",
      "Epoch 127: val_accuracy did not improve from 0.94983\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 14ms/step - accuracy: 0.9382 - loss: 0.2160 - val_accuracy: 0.9497 - val_loss: 0.1875\n",
      "Epoch 128/1000\n",
      "\u001b[1m44/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9374 - loss: 0.2152\n",
      "Epoch 128: val_accuracy improved from 0.94983 to 0.95050, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 17ms/step - accuracy: 0.9375 - loss: 0.2151 - val_accuracy: 0.9505 - val_loss: 0.1866\n",
      "Epoch 129/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.9416 - loss: 0.2093\n",
      "Epoch 129: val_accuracy did not improve from 0.95050\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9414 - loss: 0.2096 - val_accuracy: 0.9505 - val_loss: 0.1859\n",
      "Epoch 130/1000\n",
      "\u001b[1m45/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.9398 - loss: 0.2113\n",
      "Epoch 130: val_accuracy improved from 0.95050 to 0.95067, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9397 - loss: 0.2115 - val_accuracy: 0.9507 - val_loss: 0.1852\n",
      "Epoch 131/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.9390 - loss: 0.2145\n",
      "Epoch 131: val_accuracy improved from 0.95067 to 0.95083, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9390 - loss: 0.2145 - val_accuracy: 0.9508 - val_loss: 0.1848\n",
      "Epoch 132/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 10ms/step - accuracy: 0.9390 - loss: 0.2134\n",
      "Epoch 132: val_accuracy improved from 0.95083 to 0.95133, saving model to mnist_keras_model.keras\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 16ms/step - accuracy: 0.9391 - loss: 0.2133 - val_accuracy: 0.9513 - val_loss: 0.1840\n",
      "Epoch 133/1000\n",
      "\u001b[1m46/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 13ms/step - accuracy: 0.9397 - loss: 0.2134\n",
      "Epoch 133: val_accuracy did not improve from 0.95133\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 20ms/step - accuracy: 0.9398 - loss: 0.2133 - val_accuracy: 0.9509 - val_loss: 0.1835\n",
      "Epoch 134/1000\n",
      "\u001b[1m47/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9389 - loss: 0.2122\n",
      "Epoch 134: val_accuracy did not improve from 0.95133\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 19ms/step - accuracy: 0.9389 - loss: 0.2121 - val_accuracy: 0.9512 - val_loss: 0.1829\n",
      "Epoch 135/1000\n",
      "\u001b[1m43/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m━━━\u001b[0m \u001b[1m0s\u001b[0m 12ms/step - accuracy: 0.9419 - loss: 0.2047\n",
      "Epoch 135: val_accuracy did not improve from 0.95133\n",
      "\u001b[1m48/48\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m1s\u001b[0m 18ms/step - accuracy: 0.9417 - loss: 0.2050 - val_accuracy: 0.9512 - val_loss: 0.1822\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.history.History at 0x2c1f77af310>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "text/plain": [
       "<keras.src.callbacks.History at 0x21d235d9f90>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, y_train, validation_split=0.2,\n",
    "          batch_size=1000, epochs=1000, \n",
    "          callbacks=[checkpoint, early_stopping, tensor_board])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "94a6c468",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.11.7"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": true
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}