61_VGG_CAM.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import tensorflow.keras.backend as K\n",
    "\n",
    "from tensorflow.keras.models import Sequential, Model\n",
    "\n",
    "from tensorflow.keras.layers import (\n",
    "    Input,\n",
    "    Dense,\n",
    "    Reshape,\n",
    "    Flatten,\n",
    "    Dropout,\n",
    "    BatchNormalization,\n",
    "    Activation,\n",
    "    ZeroPadding2D,\n",
    "    LeakyReLU,\n",
    "    UpSampling2D,\n",
    "    Conv2D,\n",
    "    Convolution2D,\n",
    "    MaxPooling2D,\n",
    "    Concatenate,\n",
    "    GaussianNoise,\n",
    "    GaussianDropout,\n",
    "    Lambda,\n",
    "    GlobalAveragePooling2D,\n",
    ")\n",
    "\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "\n",
    "import h5py\n",
    "import pickle\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import os\n",
    "import pathlib\n",
    "\n",
    "import time\n",
    "\n",
    "import math"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Tensorflow version:  2.1.0\n",
      "Keras version:  2.2.4-tf\n",
      "Is eager execution enabled:  True\n",
      "WARNING:tensorflow:From <ipython-input-2-177a7f55bd9d>:6: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use `tf.config.list_physical_devices('GPU')` instead.\n",
      "Is there a GPU available:  True\n"
     ]
    }
   ],
   "source": [
    "print(\"Tensorflow version: \", tf.version.VERSION)  # tf2\n",
    "print(\"Keras version: \", tf.keras.__version__)  # 2.2.4-tf\n",
    "\n",
    "# tf.enable_eager_execution()  # tf2\n",
    "print(\"Is eager execution enabled: \", tf.executing_eagerly())\n",
    "print(\"Is there a GPU available: \", tf.test.is_gpu_available())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "path_train_tfrecord = \"fp56_train.tfrecord\"\n",
    "path_test_tfrecord = \"fp56_test.tfrecord\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# model save dir"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "dir_model = \"vgg_cam/\"\n",
    "pathlib.Path(dir_model).mkdir(parents=True, exist_ok=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "fp_dim = (56, 56, 6)\n",
    "\n",
    "\n",
    "def _parse_function(example_proto):\n",
    "    # Create a description of the features.\n",
    "    feature_description = {\n",
    "        \"floorplan\": tf.io.FixedLenFeature(\n",
    "            fp_dim, tf.float32, default_value=tf.zeros(fp_dim, tf.float32)\n",
    "        ),\n",
    "        \"plan_id\": tf.io.FixedLenFeature([], tf.string, default_value=\"\"),\n",
    "        \"year\": tf.io.FixedLenFeature([], tf.int64, default_value=-1),  # 0~9\n",
    "        # \"sido\": tf.FixedLenFeature([], tf.int64, default_value=-1),\n",
    "        # \"norm_area\": tf.FixedLenFeature([], tf.float32, default_value=0.0),\n",
    "        # \"num_rooms\": tf.FixedLenFeature([], tf.int64, default_value=-1),\n",
    "        # \"num_baths\": tf.FixedLenFeature([], tf.int64, default_value=-1),\n",
    "    }\n",
    "\n",
    "    # Parse the input tf.Example proto using the dictionary above.\n",
    "    parsed_example = tf.io.parse_single_example(example_proto, feature_description)\n",
    "\n",
    "    return parsed_example[\"floorplan\"], parsed_example[\"year\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def _onehot_year(fp, year):\n",
    "    year_onehot = tf.one_hot(year, 10)  # 1970~4 -> 0, 2015~9 -> 9\n",
    "    return (fp, year_onehot)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_dataset(filepath):\n",
    "    # This works with arrays as well\n",
    "    dataset = tf.data.TFRecordDataset(filepath, compression_type=\"GZIP\")\n",
    "\n",
    "    # Maps the parser on every filepath in the array. You can set the number of parallel loaders here\n",
    "    dataset = dataset.map(_parse_function, num_parallel_calls=4)\n",
    "\n",
    "    ### preprocess the features\n",
    "\n",
    "    # won't use it. use sparse_categorical_crossentropy instead of categorical_crossentropy.\n",
    "    #     dataset = dataset.map(_onehot_year, num_parallel_calls=4)\n",
    "\n",
    "    return dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def VGG16_convolutions():\n",
    "    if K.image_data_format() == \"channels_last\":\n",
    "        input_shape = (fp_dim[0], fp_dim[1], fp_dim[2])\n",
    "    else:\n",
    "        input_shape = (fp_dim[2], fp_dim[0], fp_dim[1])\n",
    "\n",
    "    model = Sequential()\n",
    "    model.add(GaussianNoise(0.1, input_shape=input_shape))\n",
    "\n",
    "    model.add(Conv2D(64, (3, 3), activation=\"relu\", name=\"conv1_1\", padding=\"same\"))\n",
    "    model.add(Conv2D(64, (3, 3), activation=\"relu\", name=\"conv1_2\", padding=\"same\"))\n",
    "    model.add(MaxPooling2D((2, 2), strides=(1, 1), padding=\"same\"))\n",
    "\n",
    "    model.add(Conv2D(128, (3, 3), activation=\"relu\", name=\"conv2_1\", padding=\"same\"))\n",
    "    model.add(Conv2D(128, (3, 3), activation=\"relu\", name=\"conv2_2\", padding=\"same\"))\n",
    "    model.add(MaxPooling2D((2, 2), strides=(1, 1), padding=\"same\"))\n",
    "\n",
    "    model.add(Conv2D(256, (3, 3), activation=\"relu\", name=\"conv3_1\", padding=\"same\"))\n",
    "    model.add(Conv2D(256, (3, 3), activation=\"relu\", name=\"conv3_2\", padding=\"same\"))\n",
    "    model.add(Conv2D(256, (3, 3), activation=\"relu\", name=\"conv3_3\", padding=\"same\"))\n",
    "    model.add(MaxPooling2D((2, 2), strides=(2, 2)))\n",
    "\n",
    "    model.add(Conv2D(512, (3, 3), activation=\"relu\", name=\"conv4_1\", padding=\"same\"))\n",
    "    model.add(Conv2D(512, (3, 3), activation=\"relu\", name=\"conv4_2\", padding=\"same\"))\n",
    "    model.add(Conv2D(512, (3, 3), activation=\"relu\", name=\"conv4_3\", padding=\"same\"))\n",
    "    model.add(MaxPooling2D((2, 2), strides=(2, 2)))\n",
    "\n",
    "    model.add(Conv2D(512, (3, 3), activation=\"relu\", name=\"conv5_1\", padding=\"same\"))\n",
    "    model.add(Conv2D(512, (3, 3), activation=\"relu\", name=\"conv5_2\", padding=\"same\"))\n",
    "    model.add(Conv2D(512, (3, 3), activation=\"relu\", name=\"conv5_3\", padding=\"same\"))\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_classes = 10\n",
    "\n",
    "\n",
    "def create_model():\n",
    "    model = VGG16_convolutions()\n",
    "\n",
    "    model.add(GlobalAveragePooling2D())\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(num_classes, activation=\"softmax\"))\n",
    "\n",
    "    model.compile(\n",
    "        optimizer=\"sgd\", loss=\"sparse_categorical_crossentropy\", metrics=[\"accuracy\"]\n",
    "    )\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "callback_checkpoint = tf.keras.callbacks.ModelCheckpoint(\n",
    "    dir_model + \"model-{epoch:02d}-{val_loss:.2f}-{val_accuracy:.1%}.hdf5\",\n",
    "    # save_weights_only=True,\n",
    "    verbose=1,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "callback_stop = tf.keras.callbacks.EarlyStopping(monitor=\"val_loss\", patience=20)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# run"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(<BatchDataset shapes: ((None, 56, 56, 6), (None,)), types: (tf.float32, tf.int64)>,\n",
       " <BatchDataset shapes: ((None, 56, 56, 6), (None,)), types: (tf.float32, tf.int64)>)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# .repeat().shuffle(4096).batch(8)\n",
    "\n",
    "train_dataset = create_dataset(path_train_tfrecord).shuffle(1024).batch(8)\n",
    "test_dataset = create_dataset(path_test_tfrecord).batch(8)\n",
    "\n",
    "train_dataset, test_dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "gaussian_noise (GaussianNois (None, 56, 56, 6)         0         \n",
      "_________________________________________________________________\n",
      "conv1_1 (Conv2D)             (None, 56, 56, 64)        3520      \n",
      "_________________________________________________________________\n",
      "conv1_2 (Conv2D)             (None, 56, 56, 64)        36928     \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 56, 56, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2_1 (Conv2D)             (None, 56, 56, 128)       73856     \n",
      "_________________________________________________________________\n",
      "conv2_2 (Conv2D)             (None, 56, 56, 128)       147584    \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "conv3_1 (Conv2D)             (None, 56, 56, 256)       295168    \n",
      "_________________________________________________________________\n",
      "conv3_2 (Conv2D)             (None, 56, 56, 256)       590080    \n",
      "_________________________________________________________________\n",
      "conv3_3 (Conv2D)             (None, 56, 56, 256)       590080    \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 28, 28, 256)       0         \n",
      "_________________________________________________________________\n",
      "conv4_1 (Conv2D)             (None, 28, 28, 512)       1180160   \n",
      "_________________________________________________________________\n",
      "conv4_2 (Conv2D)             (None, 28, 28, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "conv4_3 (Conv2D)             (None, 28, 28, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 14, 14, 512)       0         \n",
      "_________________________________________________________________\n",
      "conv5_1 (Conv2D)             (None, 14, 14, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "conv5_2 (Conv2D)             (None, 14, 14, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "conv5_3 (Conv2D)             (None, 14, 14, 512)       2359808   \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d (Gl (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 10)                5130      \n",
      "=================================================================\n",
      "Total params: 14,721,546\n",
      "Trainable params: 14,721,546\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = create_model()\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "   5075/Unknown - 1155s 228ms/step - loss: 1.8163 - accuracy: 0.2780\n",
      "Epoch 00001: saving model to vgg_cam/model-01-1.83-31.9%.hdf5\n",
      "5075/5075 [==============================] - 1262s 249ms/step - loss: 1.8163 - accuracy: 0.2780 - val_loss: 1.8276 - val_accuracy: 0.3195\n",
      "Epoch 2/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 1.4637 - accuracy: 0.4164\n",
      "Epoch 00002: saving model to vgg_cam/model-02-1.48-39.9%.hdf5\n",
      "5075/5075 [==============================] - 1272s 251ms/step - loss: 1.4636 - accuracy: 0.4164 - val_loss: 1.4779 - val_accuracy: 0.3992\n",
      "Epoch 3/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 1.3363 - accuracy: 0.4566\n",
      "Epoch 00003: saving model to vgg_cam/model-03-1.50-39.2%.hdf5\n",
      "5075/5075 [==============================] - 1272s 251ms/step - loss: 1.3362 - accuracy: 0.4567 - val_loss: 1.4997 - val_accuracy: 0.3915\n",
      "Epoch 4/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 1.2741 - accuracy: 0.4780\n",
      "Epoch 00004: saving model to vgg_cam/model-04-1.42-43.8%.hdf5\n",
      "5075/5075 [==============================] - 1272s 251ms/step - loss: 1.2742 - accuracy: 0.4780 - val_loss: 1.4247 - val_accuracy: 0.4380\n",
      "Epoch 5/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 1.2266 - accuracy: 0.4984\n",
      "Epoch 00005: saving model to vgg_cam/model-05-1.47-44.3%.hdf5\n",
      "5075/5075 [==============================] - 1271s 250ms/step - loss: 1.2265 - accuracy: 0.4984 - val_loss: 1.4676 - val_accuracy: 0.4425\n",
      "Epoch 6/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 1.1861 - accuracy: 0.5167\n",
      "Epoch 00006: saving model to vgg_cam/model-06-1.34-47.0%.hdf5\n",
      "5075/5075 [==============================] - 1270s 250ms/step - loss: 1.1860 - accuracy: 0.5168 - val_loss: 1.3386 - val_accuracy: 0.4696\n",
      "Epoch 7/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 1.1509 - accuracy: 0.5252\n",
      "Epoch 00007: saving model to vgg_cam/model-07-1.37-47.8%.hdf5\n",
      "5075/5075 [==============================] - 1280s 252ms/step - loss: 1.1507 - accuracy: 0.5252 - val_loss: 1.3712 - val_accuracy: 0.4781\n",
      "Epoch 8/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 1.1193 - accuracy: 0.5405\n",
      "Epoch 00008: saving model to vgg_cam/model-08-1.31-48.0%.hdf5\n",
      "5075/5075 [==============================] - 1284s 253ms/step - loss: 1.1193 - accuracy: 0.5404 - val_loss: 1.3089 - val_accuracy: 0.4798\n",
      "Epoch 9/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 1.0878 - accuracy: 0.5535\n",
      "Epoch 00009: saving model to vgg_cam/model-09-1.29-47.8%.hdf5\n",
      "5075/5075 [==============================] - 1296s 255ms/step - loss: 1.0877 - accuracy: 0.5536 - val_loss: 1.2914 - val_accuracy: 0.4778\n",
      "Epoch 10/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 1.0571 - accuracy: 0.5646\n",
      "Epoch 00010: saving model to vgg_cam/model-10-1.23-50.5%.hdf5\n",
      "5075/5075 [==============================] - 1294s 255ms/step - loss: 1.0570 - accuracy: 0.5646 - val_loss: 1.2324 - val_accuracy: 0.5051\n",
      "Epoch 11/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 1.0257 - accuracy: 0.5773\n",
      "Epoch 00011: saving model to vgg_cam/model-11-1.18-52.3%.hdf5\n",
      "5075/5075 [==============================] - 1302s 257ms/step - loss: 1.0256 - accuracy: 0.5774 - val_loss: 1.1782 - val_accuracy: 0.5233\n",
      "Epoch 12/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.9991 - accuracy: 0.5898\n",
      "Epoch 00012: saving model to vgg_cam/model-12-1.26-50.3%.hdf5\n",
      "5075/5075 [==============================] - 1308s 258ms/step - loss: 0.9991 - accuracy: 0.5898 - val_loss: 1.2603 - val_accuracy: 0.5029\n",
      "Epoch 13/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.9689 - accuracy: 0.6051\n",
      "Epoch 00013: saving model to vgg_cam/model-13-1.19-51.5%.hdf5\n",
      "5075/5075 [==============================] - 1310s 258ms/step - loss: 0.9688 - accuracy: 0.6051 - val_loss: 1.1886 - val_accuracy: 0.5148\n",
      "Epoch 14/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.9403 - accuracy: 0.6145\n",
      "Epoch 00014: saving model to vgg_cam/model-14-1.27-49.9%.hdf5\n",
      "5075/5075 [==============================] - 1307s 257ms/step - loss: 0.9401 - accuracy: 0.6145 - val_loss: 1.2741 - val_accuracy: 0.4994\n",
      "Epoch 15/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.9103 - accuracy: 0.6272\n",
      "Epoch 00015: saving model to vgg_cam/model-15-1.19-52.8%.hdf5\n",
      "5075/5075 [==============================] - 1285s 253ms/step - loss: 0.9103 - accuracy: 0.6272 - val_loss: 1.1858 - val_accuracy: 0.5283\n",
      "Epoch 16/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.8782 - accuracy: 0.6428\n",
      "Epoch 00016: saving model to vgg_cam/model-16-1.26-52.8%.hdf5\n",
      "5075/5075 [==============================] - 1282s 253ms/step - loss: 0.8781 - accuracy: 0.6428 - val_loss: 1.2559 - val_accuracy: 0.5281\n",
      "Epoch 17/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.8454 - accuracy: 0.6528\n",
      "Epoch 00017: saving model to vgg_cam/model-17-1.17-53.3%.hdf5\n",
      "5075/5075 [==============================] - 1275s 251ms/step - loss: 0.8455 - accuracy: 0.6527 - val_loss: 1.1735 - val_accuracy: 0.5327\n",
      "Epoch 18/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.8093 - accuracy: 0.6724\n",
      "Epoch 00018: saving model to vgg_cam/model-18-1.22-53.1%.hdf5\n",
      "5075/5075 [==============================] - 1278s 252ms/step - loss: 0.8093 - accuracy: 0.6724 - val_loss: 1.2236 - val_accuracy: 0.5311\n",
      "Epoch 19/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.7753 - accuracy: 0.6869\n",
      "Epoch 00019: saving model to vgg_cam/model-19-1.22-52.7%.hdf5\n",
      "5075/5075 [==============================] - 1264s 249ms/step - loss: 0.7753 - accuracy: 0.6869 - val_loss: 1.2217 - val_accuracy: 0.5271\n",
      "Epoch 20/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.7371 - accuracy: 0.7053\n",
      "Epoch 00020: saving model to vgg_cam/model-20-1.28-52.4%.hdf5\n",
      "5075/5075 [==============================] - 1270s 250ms/step - loss: 0.7371 - accuracy: 0.7053 - val_loss: 1.2832 - val_accuracy: 0.5235\n",
      "Epoch 21/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.6986 - accuracy: 0.7198\n",
      "Epoch 00021: saving model to vgg_cam/model-21-1.30-53.0%.hdf5\n",
      "5075/5075 [==============================] - 1276s 251ms/step - loss: 0.6986 - accuracy: 0.7198 - val_loss: 1.3000 - val_accuracy: 0.5298\n",
      "Epoch 22/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.6581 - accuracy: 0.7366\n",
      "Epoch 00022: saving model to vgg_cam/model-22-1.51-50.4%.hdf5\n",
      "5075/5075 [==============================] - 1278s 252ms/step - loss: 0.6581 - accuracy: 0.7366 - val_loss: 1.5058 - val_accuracy: 0.5037\n",
      "Epoch 23/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.6146 - accuracy: 0.7564\n",
      "Epoch 00023: saving model to vgg_cam/model-23-1.95-50.4%.hdf5\n",
      "5075/5075 [==============================] - 1279s 252ms/step - loss: 0.6145 - accuracy: 0.7564 - val_loss: 1.9534 - val_accuracy: 0.5037\n",
      "Epoch 24/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.5768 - accuracy: 0.7736\n",
      "Epoch 00024: saving model to vgg_cam/model-24-1.55-51.4%.hdf5\n",
      "5075/5075 [==============================] - 1280s 252ms/step - loss: 0.5769 - accuracy: 0.7736 - val_loss: 1.5468 - val_accuracy: 0.5136\n",
      "Epoch 25/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.5351 - accuracy: 0.7895\n",
      "Epoch 00025: saving model to vgg_cam/model-25-1.53-53.0%.hdf5\n",
      "5075/5075 [==============================] - 1280s 252ms/step - loss: 0.5350 - accuracy: 0.7895 - val_loss: 1.5311 - val_accuracy: 0.5298\n",
      "Epoch 26/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.4968 - accuracy: 0.8089\n",
      "Epoch 00026: saving model to vgg_cam/model-26-1.64-51.1%.hdf5\n",
      "5075/5075 [==============================] - 1280s 252ms/step - loss: 0.4968 - accuracy: 0.8089 - val_loss: 1.6422 - val_accuracy: 0.5106\n",
      "Epoch 27/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.4606 - accuracy: 0.8202\n",
      "Epoch 00027: saving model to vgg_cam/model-27-1.66-53.8%.hdf5\n",
      "5075/5075 [==============================] - 1281s 252ms/step - loss: 0.4605 - accuracy: 0.8202 - val_loss: 1.6612 - val_accuracy: 0.5378\n",
      "Epoch 28/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.4267 - accuracy: 0.8377\n",
      "Epoch 00028: saving model to vgg_cam/model-28-1.70-53.3%.hdf5\n",
      "5075/5075 [==============================] - 1281s 252ms/step - loss: 0.4267 - accuracy: 0.8377 - val_loss: 1.7048 - val_accuracy: 0.5326\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 29/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.3955 - accuracy: 0.8482\n",
      "Epoch 00029: saving model to vgg_cam/model-29-2.03-52.1%.hdf5\n",
      "5075/5075 [==============================] - 1281s 252ms/step - loss: 0.3955 - accuracy: 0.8483 - val_loss: 2.0285 - val_accuracy: 0.5214\n",
      "Epoch 30/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.3683 - accuracy: 0.8567\n",
      "Epoch 00030: saving model to vgg_cam/model-30-1.85-51.2%.hdf5\n",
      "5075/5075 [==============================] - 1281s 252ms/step - loss: 0.3683 - accuracy: 0.8567 - val_loss: 1.8516 - val_accuracy: 0.5123\n",
      "Epoch 31/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.3414 - accuracy: 0.8712\n",
      "Epoch 00031: saving model to vgg_cam/model-31-1.67-54.6%.hdf5\n",
      "5075/5075 [==============================] - 1281s 252ms/step - loss: 0.3414 - accuracy: 0.8712 - val_loss: 1.6691 - val_accuracy: 0.5456\n",
      "Epoch 32/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.3097 - accuracy: 0.8837\n",
      "Epoch 00032: saving model to vgg_cam/model-32-1.93-52.8%.hdf5\n",
      "5075/5075 [==============================] - 1281s 252ms/step - loss: 0.3097 - accuracy: 0.8837 - val_loss: 1.9304 - val_accuracy: 0.5281\n",
      "Epoch 33/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.2885 - accuracy: 0.8938\n",
      "Epoch 00033: saving model to vgg_cam/model-33-2.14-52.6%.hdf5\n",
      "5075/5075 [==============================] - 1281s 252ms/step - loss: 0.2885 - accuracy: 0.8938 - val_loss: 2.1392 - val_accuracy: 0.5260\n",
      "Epoch 34/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.2693 - accuracy: 0.8992\n",
      "Epoch 00034: saving model to vgg_cam/model-34-2.16-53.0%.hdf5\n",
      "5075/5075 [==============================] - 1281s 252ms/step - loss: 0.2692 - accuracy: 0.8993 - val_loss: 2.1564 - val_accuracy: 0.5299\n",
      "Epoch 35/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.2442 - accuracy: 0.9114\n",
      "Epoch 00035: saving model to vgg_cam/model-35-2.26-53.3%.hdf5\n",
      "5075/5075 [==============================] - 1281s 252ms/step - loss: 0.2442 - accuracy: 0.9114 - val_loss: 2.2560 - val_accuracy: 0.5335\n",
      "Epoch 36/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.2287 - accuracy: 0.9168\n",
      "Epoch 00036: saving model to vgg_cam/model-36-2.37-52.9%.hdf5\n",
      "5075/5075 [==============================] - 1282s 253ms/step - loss: 0.2287 - accuracy: 0.9168 - val_loss: 2.3739 - val_accuracy: 0.5287\n",
      "Epoch 37/50\n",
      "5074/5075 [============================>.] - ETA: 0s - loss: 0.2106 - accuracy: 0.9225\n",
      "Epoch 00037: saving model to vgg_cam/model-37-2.62-53.3%.hdf5\n",
      "5075/5075 [==============================] - 1281s 252ms/step - loss: 0.2105 - accuracy: 0.9225 - val_loss: 2.6229 - val_accuracy: 0.5333\n"
     ]
    }
   ],
   "source": [
    "path_best = dir_model + \"model-15-1.36.hdf5\"\n",
    "epoch_best = 0  # 0 if starting from fresh\n",
    "\n",
    "if epoch_best and os.path.exists(path_best):\n",
    "    model.load_weights(path_best)\n",
    "    history = model.fit(\n",
    "        train_dataset,\n",
    "        epochs=50,\n",
    "        initial_epoch=epoch_best,\n",
    "        validation_data=test_dataset,\n",
    "        callbacks=[callback_checkpoint, callback_stop],\n",
    "    )\n",
    "else:\n",
    "    history = model.fit(\n",
    "        train_dataset,\n",
    "        epochs=50,\n",
    "        validation_data=test_dataset,\n",
    "        callbacks=[callback_checkpoint, callback_stop],\n",
    "    )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'loss': [1.8162978943348331,\n",
       "  1.4635991291833268,\n",
       "  1.3362321493349507,\n",
       "  1.2741757292742564,\n",
       "  1.2264919695851133,\n",
       "  1.186024851397086,\n",
       "  1.1507806714265016,\n",
       "  1.119309566991013,\n",
       "  1.0876969628553919,\n",
       "  1.0570432796616984,\n",
       "  1.0256374557767984,\n",
       "  0.9991143535520004,\n",
       "  0.9688518497502137,\n",
       "  0.9401483448623548,\n",
       "  0.9102872164239202,\n",
       "  0.8780909352328626,\n",
       "  0.8454820460959358,\n",
       "  0.8092848961224338,\n",
       "  0.7753078669178949,\n",
       "  0.7370844810395072,\n",
       "  0.6985614552739348,\n",
       "  0.6580675742025299,\n",
       "  0.6145530962747002,\n",
       "  0.5768494394976398,\n",
       "  0.5350521280238432,\n",
       "  0.4967964432391811,\n",
       "  0.46053187873514906,\n",
       "  0.42668259218678334,\n",
       "  0.39550364380784503,\n",
       "  0.3682810392729146,\n",
       "  0.34136750161174495,\n",
       "  0.3096840356207254,\n",
       "  0.28847593629735113,\n",
       "  0.2692262731716117,\n",
       "  0.24416366183208096,\n",
       "  0.22865911675981732,\n",
       "  0.21053600871395625],\n",
       " 'accuracy': [0.27796936,\n",
       "  0.4164491,\n",
       "  0.45667273,\n",
       "  0.47802848,\n",
       "  0.4984482,\n",
       "  0.51677424,\n",
       "  0.5251983,\n",
       "  0.5404453,\n",
       "  0.5535987,\n",
       "  0.5646337,\n",
       "  0.5773683,\n",
       "  0.589832,\n",
       "  0.60507905,\n",
       "  0.61453766,\n",
       "  0.62724763,\n",
       "  0.64281493,\n",
       "  0.6527415,\n",
       "  0.6723977,\n",
       "  0.6869304,\n",
       "  0.7052564,\n",
       "  0.71976453,\n",
       "  0.7366126,\n",
       "  0.75641656,\n",
       "  0.7736095,\n",
       "  0.78952163,\n",
       "  0.80888224,\n",
       "  0.82018816,\n",
       "  0.8377014,\n",
       "  0.8482684,\n",
       "  0.85671705,\n",
       "  0.87117594,\n",
       "  0.8837381,\n",
       "  0.89383715,\n",
       "  0.8992561,\n",
       "  0.9113503,\n",
       "  0.9168432,\n",
       "  0.92248386],\n",
       " 'val_loss': [1.8276090505405607,\n",
       "  1.4778723998215935,\n",
       "  1.4996511626806552,\n",
       "  1.4247261519114045,\n",
       "  1.46764188774035,\n",
       "  1.3385755635902916,\n",
       "  1.3711647873045298,\n",
       "  1.308884818505036,\n",
       "  1.2914354372476387,\n",
       "  1.2324052438288062,\n",
       "  1.1782387287387595,\n",
       "  1.2603415379913365,\n",
       "  1.1885713476999396,\n",
       "  1.2741120515118083,\n",
       "  1.1857899964726109,\n",
       "  1.2558605213462073,\n",
       "  1.1734699930103731,\n",
       "  1.2236379873911853,\n",
       "  1.2216807436738117,\n",
       "  1.283174069360664,\n",
       "  1.2999641219464184,\n",
       "  1.5057814580460898,\n",
       "  1.9534249508955193,\n",
       "  1.546836470084785,\n",
       "  1.5311400687126033,\n",
       "  1.6422120019756474,\n",
       "  1.6612404686921178,\n",
       "  1.704755053038353,\n",
       "  2.0284658899139565,\n",
       "  1.8515558746850154,\n",
       "  1.6690863017158701,\n",
       "  1.9304001867860268,\n",
       "  2.139214891654477,\n",
       "  2.1564442400906203,\n",
       "  2.255996604095231,\n",
       "  2.3738542128211177,\n",
       "  2.6229495651902406],\n",
       " 'val_accuracy': [0.3194531,\n",
       "  0.39921275,\n",
       "  0.39154753,\n",
       "  0.43795317,\n",
       "  0.44251087,\n",
       "  0.46964988,\n",
       "  0.47814378,\n",
       "  0.47980112,\n",
       "  0.47783303,\n",
       "  0.50507563,\n",
       "  0.5233064,\n",
       "  0.50290036,\n",
       "  0.5148125,\n",
       "  0.4993785,\n",
       "  0.5282784,\n",
       "  0.5280713,\n",
       "  0.53273255,\n",
       "  0.5310752,\n",
       "  0.527139,\n",
       "  0.52351356,\n",
       "  0.5298322,\n",
       "  0.50372905,\n",
       "  0.50372905,\n",
       "  0.51356953,\n",
       "  0.5298322,\n",
       "  0.5105656,\n",
       "  0.5378082,\n",
       "  0.53262895,\n",
       "  0.5214419,\n",
       "  0.5123265,\n",
       "  0.545577,\n",
       "  0.5280713,\n",
       "  0.5259996,\n",
       "  0.5299358,\n",
       "  0.53345764,\n",
       "  0.5286928,\n",
       "  0.53325045]}"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "history.history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_hist = pd.DataFrame(\n",
    "    history.history,\n",
    "    index=range(epoch_best + 1, epoch_best + len(history.history[\"loss\"]) + 1),\n",
    ")\n",
    "df_hist.index.name = \"epoch\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>val_loss</th>\n",
       "      <th>val_accuracy</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>epoch</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1.816298</td>\n",
       "      <td>0.277969</td>\n",
       "      <td>1.827609</td>\n",
       "      <td>0.319453</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1.463599</td>\n",
       "      <td>0.416449</td>\n",
       "      <td>1.477872</td>\n",
       "      <td>0.399213</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1.336232</td>\n",
       "      <td>0.456673</td>\n",
       "      <td>1.499651</td>\n",
       "      <td>0.391548</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1.274176</td>\n",
       "      <td>0.478028</td>\n",
       "      <td>1.424726</td>\n",
       "      <td>0.437953</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1.226492</td>\n",
       "      <td>0.498448</td>\n",
       "      <td>1.467642</td>\n",
       "      <td>0.442511</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>1.186025</td>\n",
       "      <td>0.516774</td>\n",
       "      <td>1.338576</td>\n",
       "      <td>0.469650</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>1.150781</td>\n",
       "      <td>0.525198</td>\n",
       "      <td>1.371165</td>\n",
       "      <td>0.478144</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1.119310</td>\n",
       "      <td>0.540445</td>\n",
       "      <td>1.308885</td>\n",
       "      <td>0.479801</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>1.087697</td>\n",
       "      <td>0.553599</td>\n",
       "      <td>1.291435</td>\n",
       "      <td>0.477833</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>1.057043</td>\n",
       "      <td>0.564634</td>\n",
       "      <td>1.232405</td>\n",
       "      <td>0.505076</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>1.025637</td>\n",
       "      <td>0.577368</td>\n",
       "      <td>1.178239</td>\n",
       "      <td>0.523306</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.999114</td>\n",
       "      <td>0.589832</td>\n",
       "      <td>1.260342</td>\n",
       "      <td>0.502900</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.968852</td>\n",
       "      <td>0.605079</td>\n",
       "      <td>1.188571</td>\n",
       "      <td>0.514813</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.940148</td>\n",
       "      <td>0.614538</td>\n",
       "      <td>1.274112</td>\n",
       "      <td>0.499379</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.910287</td>\n",
       "      <td>0.627248</td>\n",
       "      <td>1.185790</td>\n",
       "      <td>0.528278</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.878091</td>\n",
       "      <td>0.642815</td>\n",
       "      <td>1.255861</td>\n",
       "      <td>0.528071</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0.845482</td>\n",
       "      <td>0.652741</td>\n",
       "      <td>1.173470</td>\n",
       "      <td>0.532733</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0.809285</td>\n",
       "      <td>0.672398</td>\n",
       "      <td>1.223638</td>\n",
       "      <td>0.531075</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0.775308</td>\n",
       "      <td>0.686930</td>\n",
       "      <td>1.221681</td>\n",
       "      <td>0.527139</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.737084</td>\n",
       "      <td>0.705256</td>\n",
       "      <td>1.283174</td>\n",
       "      <td>0.523514</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>0.698561</td>\n",
       "      <td>0.719765</td>\n",
       "      <td>1.299964</td>\n",
       "      <td>0.529832</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.658068</td>\n",
       "      <td>0.736613</td>\n",
       "      <td>1.505781</td>\n",
       "      <td>0.503729</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.614553</td>\n",
       "      <td>0.756417</td>\n",
       "      <td>1.953425</td>\n",
       "      <td>0.503729</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0.576849</td>\n",
       "      <td>0.773610</td>\n",
       "      <td>1.546836</td>\n",
       "      <td>0.513570</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.535052</td>\n",
       "      <td>0.789522</td>\n",
       "      <td>1.531140</td>\n",
       "      <td>0.529832</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.496796</td>\n",
       "      <td>0.808882</td>\n",
       "      <td>1.642212</td>\n",
       "      <td>0.510566</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.460532</td>\n",
       "      <td>0.820188</td>\n",
       "      <td>1.661240</td>\n",
       "      <td>0.537808</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0.426683</td>\n",
       "      <td>0.837701</td>\n",
       "      <td>1.704755</td>\n",
       "      <td>0.532629</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>0.395504</td>\n",
       "      <td>0.848268</td>\n",
       "      <td>2.028466</td>\n",
       "      <td>0.521442</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>0.368281</td>\n",
       "      <td>0.856717</td>\n",
       "      <td>1.851556</td>\n",
       "      <td>0.512326</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>0.341368</td>\n",
       "      <td>0.871176</td>\n",
       "      <td>1.669086</td>\n",
       "      <td>0.545577</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>0.309684</td>\n",
       "      <td>0.883738</td>\n",
       "      <td>1.930400</td>\n",
       "      <td>0.528071</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>0.288476</td>\n",
       "      <td>0.893837</td>\n",
       "      <td>2.139215</td>\n",
       "      <td>0.526000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>0.269226</td>\n",
       "      <td>0.899256</td>\n",
       "      <td>2.156444</td>\n",
       "      <td>0.529936</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>0.244164</td>\n",
       "      <td>0.911350</td>\n",
       "      <td>2.255997</td>\n",
       "      <td>0.533458</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>0.228659</td>\n",
       "      <td>0.916843</td>\n",
       "      <td>2.373854</td>\n",
       "      <td>0.528693</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>0.210536</td>\n",
       "      <td>0.922484</td>\n",
       "      <td>2.622950</td>\n",
       "      <td>0.533250</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           loss  accuracy  val_loss  val_accuracy\n",
       "epoch                                            \n",
       "1      1.816298  0.277969  1.827609      0.319453\n",
       "2      1.463599  0.416449  1.477872      0.399213\n",
       "3      1.336232  0.456673  1.499651      0.391548\n",
       "4      1.274176  0.478028  1.424726      0.437953\n",
       "5      1.226492  0.498448  1.467642      0.442511\n",
       "6      1.186025  0.516774  1.338576      0.469650\n",
       "7      1.150781  0.525198  1.371165      0.478144\n",
       "8      1.119310  0.540445  1.308885      0.479801\n",
       "9      1.087697  0.553599  1.291435      0.477833\n",
       "10     1.057043  0.564634  1.232405      0.505076\n",
       "11     1.025637  0.577368  1.178239      0.523306\n",
       "12     0.999114  0.589832  1.260342      0.502900\n",
       "13     0.968852  0.605079  1.188571      0.514813\n",
       "14     0.940148  0.614538  1.274112      0.499379\n",
       "15     0.910287  0.627248  1.185790      0.528278\n",
       "16     0.878091  0.642815  1.255861      0.528071\n",
       "17     0.845482  0.652741  1.173470      0.532733\n",
       "18     0.809285  0.672398  1.223638      0.531075\n",
       "19     0.775308  0.686930  1.221681      0.527139\n",
       "20     0.737084  0.705256  1.283174      0.523514\n",
       "21     0.698561  0.719765  1.299964      0.529832\n",
       "22     0.658068  0.736613  1.505781      0.503729\n",
       "23     0.614553  0.756417  1.953425      0.503729\n",
       "24     0.576849  0.773610  1.546836      0.513570\n",
       "25     0.535052  0.789522  1.531140      0.529832\n",
       "26     0.496796  0.808882  1.642212      0.510566\n",
       "27     0.460532  0.820188  1.661240      0.537808\n",
       "28     0.426683  0.837701  1.704755      0.532629\n",
       "29     0.395504  0.848268  2.028466      0.521442\n",
       "30     0.368281  0.856717  1.851556      0.512326\n",
       "31     0.341368  0.871176  1.669086      0.545577\n",
       "32     0.309684  0.883738  1.930400      0.528071\n",
       "33     0.288476  0.893837  2.139215      0.526000\n",
       "34     0.269226  0.899256  2.156444      0.529936\n",
       "35     0.244164  0.911350  2.255997      0.533458\n",
       "36     0.228659  0.916843  2.373854      0.528693\n",
       "37     0.210536  0.922484  2.622950      0.533250"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_hist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "path_hist = dir_model + \"history.csv\"\n",
    "df_hist.to_csv(path_hist, mode=\"a\", header=not os.path.exists(path_hist))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}