live.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/anaconda3/envs/dlearn/lib/python3.5/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import os\n",
    "import pathlib\n",
    "\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import datetime\n",
    "\n",
    "import keras\n",
    "from keras.preprocessing.image import ImageDataGenerator, array_to_img, img_to_array, load_img\n",
    "from keras.models import Sequential, Model\n",
    "from keras.layers import Dense, Dropout, Flatten, BatchNormalization\n",
    "from keras.layers import Conv2D, MaxPooling2D, Activation, Reshape, GlobalAveragePooling2D\n",
    "from keras.models import load_model\n",
    "from keras.backend.tensorflow_backend import set_session\n",
    "from keras.callbacks import TensorBoard\n",
    "\n",
    "import skimage\n",
    "from skimage import io, color, transform"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def set_gpu_mem_alloc(mem_use):\n",
    "    avail  = 4041\n",
    "    percent = mem_use / avail\n",
    "    config = tf.ConfigProto()\n",
    "    config.gpu_options.per_process_gpu_memory_fraction = percent\n",
    "    config.gpu_options.visible_device_list = \"0\"\n",
    "    set_session(tf.Session(config=config))\n",
    "    \n",
    "set_gpu_mem_alloc(2700)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_size = 4\n",
    "h,w,d = 224,224,3\n",
    "epochs = 1000\n",
    "date_str = datetime.datetime.now().strftime(\"%d%m%y_%H%M\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 130 images belonging to 3 classes.\n",
      "Found 28 images belonging to 3 classes.\n"
     ]
    }
   ],
   "source": [
    "# this is the augmentation configuration we will use for training\n",
    "train_datagen = ImageDataGenerator(\n",
    "    rescale=1./255, rotation_range=20, \n",
    "    shear_range=0.2,\n",
    "    width_shift_range=0.1, height_shift_range=0.1,\n",
    "    zoom_range=0.5                            \n",
    ")\n",
    "\n",
    "\n",
    "# this is the augmentation configuration we will use for testing:\n",
    "# only rescaling\n",
    "test_datagen = ImageDataGenerator(\n",
    "    rescale=1./255, rotation_range=20, \n",
    "    shear_range=0.2,\n",
    "    width_shift_range=0.1, height_shift_range=0.1,\n",
    "    zoom_range=0.5                            \n",
    ")\n",
    "\n",
    "\n",
    "# this is a generator that will read pictures found in\n",
    "# subfolers of 'data/train', and indefinitely generate\n",
    "# batches of augmented image data\n",
    "train_generator = train_datagen.flow_from_directory(\n",
    "        'dataset/emotion/train',  # this is the target directory\n",
    "        target_size=(h, w),  # all images will be resized to 150x150\n",
    "        batch_size=batch_size,\n",
    "        class_mode='categorical')  # since we use binary_crossentropy loss, we need binary labels\n",
    "\n",
    "# this is a similar generator, for validation data\n",
    "validation_generator = test_datagen.flow_from_directory(\n",
    "        'dataset/emotion/valid',\n",
    "        target_size=(h, w),\n",
    "        batch_size=batch_size,\n",
    "        class_mode='categorical')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f6d97eff748>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f6d737f7908>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "img = train_generator.next()[0]\n",
    "plt.imshow(img[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NUM Classes =>  3\n",
      "Train Step  =>  130\n",
      "Valid Step  =>  28\n"
     ]
    }
   ],
   "source": [
    "num_class = len(train_generator.class_indices)\n",
    "train_step = len(train_generator.classes)\n",
    "valid_step = len(validation_generator.classes)\n",
    "\n",
    "print('NUM Classes => ', num_class)\n",
    "print('Train Step  => ',  train_step)\n",
    "print('Valid Step  => ',  valid_step)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 224, 224, 32)      896       \n",
      "_________________________________________________________________\n",
      "activation_1 (Activation)    (None, 224, 224, 32)      0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_1 (Batch (None, 224, 224, 32)      128       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 112, 112, 32)      0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 112, 112, 32)      0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 112, 112, 64)      18496     \n",
      "_________________________________________________________________\n",
      "activation_2 (Activation)    (None, 112, 112, 64)      0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_2 (Batch (None, 112, 112, 64)      256       \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 112, 112, 64)      36928     \n",
      "_________________________________________________________________\n",
      "activation_3 (Activation)    (None, 112, 112, 64)      0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_3 (Batch (None, 112, 112, 64)      256       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 56, 56, 64)        0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 56, 56, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 56, 56, 128)       73856     \n",
      "_________________________________________________________________\n",
      "activation_4 (Activation)    (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_4 (Batch (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 56, 56, 128)       147584    \n",
      "_________________________________________________________________\n",
      "activation_5 (Activation)    (None, 56, 56, 128)       0         \n",
      "_________________________________________________________________\n",
      "batch_normalization_5 (Batch (None, 56, 56, 128)       512       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 28, 28, 128)       0         \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 28, 28, 128)       0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 100352)            0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 3)                 301059    \n",
      "_________________________________________________________________\n",
      "activation_6 (Activation)    (None, 3)                 0         \n",
      "=================================================================\n",
      "Total params: 580,483\n",
      "Trainable params: 579,651\n",
      "Non-trainable params: 832\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "ksize = (3,3)\n",
    "psize = (2,2)\n",
    "\n",
    "input_shape = (h,w,d)\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Conv2D(32, kernel_size=ksize, padding='same', input_shape=input_shape))\n",
    "model.add(Activation('relu'))\n",
    "model.add(BatchNormalization())\n",
    "model.add(MaxPooling2D(pool_size=psize))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Conv2D(64, kernel_size=ksize, padding=\"same\"))\n",
    "model.add(Activation('relu'))\n",
    "model.add(BatchNormalization())\n",
    "model.add(Conv2D(64, kernel_size=ksize, padding=\"same\"))\n",
    "model.add(Activation('relu'))\n",
    "model.add(BatchNormalization())\n",
    "model.add(MaxPooling2D(pool_size=psize))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "model.add(Conv2D(128,kernel_size=ksize, padding=\"same\"))\n",
    "model.add(Activation('relu'))\n",
    "model.add(BatchNormalization())\n",
    "model.add(Conv2D(128,kernel_size=ksize, padding=\"same\"))\n",
    "model.add(Activation('relu'))\n",
    "model.add(BatchNormalization())\n",
    "model.add(MaxPooling2D(pool_size=psize))\n",
    "model.add(Dropout(0.25))\n",
    "\n",
    "# flatten and classifier\n",
    "model.add(Flatten())\n",
    "model.add(Dense(num_class))\n",
    "model.add(Activation(\"softmax\"))\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "adam = keras.optimizers.Adam(lr=0.00001)\n",
    "model.compile(loss='categorical_crossentropy', optimizer=adam, metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# untuk liat log history\n",
    "tb1_path = 'log/'+date_str\n",
    "pathlib.Path(tb1_path).mkdir(parents=True, exist_ok=True) \n",
    "#training\n",
    "tensorboard = TensorBoard( log_dir=tb1_path, histogram_freq=0, \n",
    "                            write_graph=True, write_images=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/1000\n",
      " - 10s - loss: 2.4110 - acc: 0.3595 - val_loss: 1.0942 - val_acc: 0.5000\n",
      "Epoch 2/1000\n",
      " - 4s - loss: 2.1819 - acc: 0.3593 - val_loss: 1.4027 - val_acc: 0.3214\n",
      "Epoch 3/1000\n",
      " - 4s - loss: 1.6692 - acc: 0.4531 - val_loss: 1.4429 - val_acc: 0.4286\n",
      "Epoch 4/1000\n",
      " - 4s - loss: 1.6756 - acc: 0.4841 - val_loss: 1.3978 - val_acc: 0.3929\n",
      "Epoch 5/1000\n",
      " - 4s - loss: 1.6417 - acc: 0.4140 - val_loss: 1.2906 - val_acc: 0.4286\n",
      "Epoch 6/1000\n",
      " - 4s - loss: 1.4536 - acc: 0.4609 - val_loss: 1.1433 - val_acc: 0.5357\n",
      "Epoch 7/1000\n",
      " - 4s - loss: 1.4120 - acc: 0.4841 - val_loss: 1.3354 - val_acc: 0.5000\n",
      "Epoch 8/1000\n",
      " - 4s - loss: 1.4227 - acc: 0.5156 - val_loss: 1.1409 - val_acc: 0.6429\n",
      "Epoch 9/1000\n",
      " - 4s - loss: 1.2793 - acc: 0.5623 - val_loss: 1.3289 - val_acc: 0.4643\n",
      "Epoch 10/1000\n",
      " - 4s - loss: 1.5488 - acc: 0.4766 - val_loss: 1.1381 - val_acc: 0.5357\n",
      "Epoch 11/1000\n",
      " - 4s - loss: 1.3370 - acc: 0.5396 - val_loss: 1.5061 - val_acc: 0.5000\n",
      "Epoch 12/1000\n",
      " - 4s - loss: 1.3358 - acc: 0.5547 - val_loss: 1.2170 - val_acc: 0.5714\n",
      "Epoch 13/1000\n",
      " - 4s - loss: 1.3746 - acc: 0.5159 - val_loss: 1.4584 - val_acc: 0.3571\n",
      "Epoch 14/1000\n",
      " - 4s - loss: 1.3862 - acc: 0.5393 - val_loss: 1.1730 - val_acc: 0.5714\n",
      "Epoch 15/1000\n",
      " - 4s - loss: 1.3401 - acc: 0.5857 - val_loss: 1.1225 - val_acc: 0.4286\n",
      "Epoch 16/1000\n",
      " - 4s - loss: 1.2621 - acc: 0.5938 - val_loss: 1.2120 - val_acc: 0.5714\n",
      "Epoch 17/1000\n",
      " - 4s - loss: 1.0654 - acc: 0.6326 - val_loss: 1.2692 - val_acc: 0.5000\n",
      "Epoch 18/1000\n",
      " - 4s - loss: 1.2050 - acc: 0.5703 - val_loss: 1.2198 - val_acc: 0.6071\n",
      "Epoch 19/1000\n",
      " - 4s - loss: 1.2951 - acc: 0.6173 - val_loss: 1.4127 - val_acc: 0.5357\n",
      "Epoch 20/1000\n",
      " - 4s - loss: 1.1833 - acc: 0.5782 - val_loss: 1.2806 - val_acc: 0.5357\n",
      "Epoch 21/1000\n",
      " - 4s - loss: 1.1715 - acc: 0.6250 - val_loss: 1.1722 - val_acc: 0.4643\n",
      "Epoch 22/1000\n",
      " - 4s - loss: 1.4497 - acc: 0.5078 - val_loss: 1.2948 - val_acc: 0.5000\n",
      "Epoch 23/1000\n",
      " - 4s - loss: 1.0269 - acc: 0.6248 - val_loss: 1.3846 - val_acc: 0.4643\n",
      "Epoch 24/1000\n",
      " - 4s - loss: 1.1826 - acc: 0.6094 - val_loss: 1.4594 - val_acc: 0.5000\n",
      "Epoch 25/1000\n",
      " - 4s - loss: 1.0458 - acc: 0.6019 - val_loss: 1.2860 - val_acc: 0.5714\n",
      "Epoch 26/1000\n",
      " - 4s - loss: 1.4093 - acc: 0.5706 - val_loss: 1.5159 - val_acc: 0.4643\n",
      "Epoch 27/1000\n",
      " - 4s - loss: 1.1428 - acc: 0.6326 - val_loss: 1.4058 - val_acc: 0.4643\n",
      "Epoch 28/1000\n",
      " - 4s - loss: 1.1698 - acc: 0.6016 - val_loss: 1.3791 - val_acc: 0.5714\n",
      "Epoch 29/1000\n",
      " - 4s - loss: 1.2010 - acc: 0.6251 - val_loss: 1.4418 - val_acc: 0.3571\n",
      "Epoch 30/1000\n",
      " - 4s - loss: 1.1480 - acc: 0.6953 - val_loss: 1.4574 - val_acc: 0.3929\n",
      "Epoch 31/1000\n",
      " - 4s - loss: 1.0330 - acc: 0.6407 - val_loss: 1.5858 - val_acc: 0.5357\n",
      "Epoch 32/1000\n",
      " - 4s - loss: 1.1897 - acc: 0.5860 - val_loss: 1.3967 - val_acc: 0.4643\n",
      "Epoch 33/1000\n",
      " - 4s - loss: 0.7823 - acc: 0.7108 - val_loss: 1.4240 - val_acc: 0.5357\n",
      "Epoch 34/1000\n",
      " - 4s - loss: 1.1656 - acc: 0.6248 - val_loss: 1.4990 - val_acc: 0.5000\n",
      "Epoch 35/1000\n",
      " - 4s - loss: 1.2375 - acc: 0.5860 - val_loss: 1.5008 - val_acc: 0.5000\n",
      "Epoch 36/1000\n",
      " - 4s - loss: 1.2294 - acc: 0.6092 - val_loss: 1.5113 - val_acc: 0.5714\n",
      "Epoch 37/1000\n",
      " - 4s - loss: 1.1361 - acc: 0.6483 - val_loss: 1.4537 - val_acc: 0.6429\n",
      "Epoch 38/1000\n",
      " - 4s - loss: 1.2159 - acc: 0.6331 - val_loss: 1.5676 - val_acc: 0.4643\n",
      "Epoch 39/1000\n",
      " - 4s - loss: 0.7748 - acc: 0.7189 - val_loss: 1.7171 - val_acc: 0.4643\n",
      "Epoch 40/1000\n",
      " - 4s - loss: 0.9521 - acc: 0.6875 - val_loss: 1.5261 - val_acc: 0.3929\n",
      "Epoch 41/1000\n",
      " - 4s - loss: 0.9562 - acc: 0.7028 - val_loss: 1.5503 - val_acc: 0.4643\n",
      "Epoch 42/1000\n",
      " - 4s - loss: 1.1748 - acc: 0.6406 - val_loss: 1.1393 - val_acc: 0.5357\n",
      "Epoch 43/1000\n",
      " - 4s - loss: 0.9279 - acc: 0.6331 - val_loss: 1.6297 - val_acc: 0.5357\n",
      "Epoch 44/1000\n",
      " - 4s - loss: 0.9403 - acc: 0.6874 - val_loss: 1.7410 - val_acc: 0.4643\n",
      "Epoch 45/1000\n",
      " - 4s - loss: 0.8940 - acc: 0.7109 - val_loss: 1.5110 - val_acc: 0.4286\n",
      "Epoch 46/1000\n",
      " - 4s - loss: 1.0095 - acc: 0.6799 - val_loss: 1.6500 - val_acc: 0.5000\n",
      "Epoch 47/1000\n",
      " - 4s - loss: 1.0461 - acc: 0.6875 - val_loss: 1.4492 - val_acc: 0.3929\n",
      "Epoch 48/1000\n",
      " - 4s - loss: 0.9727 - acc: 0.6876 - val_loss: 1.6950 - val_acc: 0.5357\n",
      "Epoch 49/1000\n",
      " - 4s - loss: 0.9185 - acc: 0.7186 - val_loss: 1.4470 - val_acc: 0.5000\n",
      "Epoch 50/1000\n",
      " - 4s - loss: 0.9331 - acc: 0.6720 - val_loss: 1.5374 - val_acc: 0.5000\n",
      "Epoch 51/1000\n",
      " - 4s - loss: 0.8161 - acc: 0.7577 - val_loss: 1.5028 - val_acc: 0.4286\n",
      "Epoch 52/1000\n",
      " - 4s - loss: 0.9308 - acc: 0.6876 - val_loss: 1.0990 - val_acc: 0.4286\n",
      "Epoch 53/1000\n",
      " - 4s - loss: 1.1610 - acc: 0.6406 - val_loss: 1.2953 - val_acc: 0.5000\n",
      "Epoch 54/1000\n",
      " - 4s - loss: 0.9961 - acc: 0.7422 - val_loss: 1.3759 - val_acc: 0.4286\n",
      "Epoch 55/1000\n",
      " - 4s - loss: 0.9504 - acc: 0.6250 - val_loss: 1.5409 - val_acc: 0.5357\n",
      "Epoch 56/1000\n",
      " - 4s - loss: 0.7897 - acc: 0.7421 - val_loss: 1.4630 - val_acc: 0.5357\n",
      "Epoch 57/1000\n",
      " - 4s - loss: 1.0281 - acc: 0.6253 - val_loss: 1.6469 - val_acc: 0.5357\n",
      "Epoch 58/1000\n",
      " - 4s - loss: 0.9529 - acc: 0.6639 - val_loss: 1.3981 - val_acc: 0.5000\n",
      "Epoch 59/1000\n",
      " - 4s - loss: 1.0647 - acc: 0.6483 - val_loss: 1.1473 - val_acc: 0.5000\n",
      "Epoch 60/1000\n",
      " - 4s - loss: 0.8526 - acc: 0.7264 - val_loss: 1.3896 - val_acc: 0.5714\n",
      "Epoch 61/1000\n",
      " - 4s - loss: 0.7662 - acc: 0.7577 - val_loss: 1.4676 - val_acc: 0.4643\n",
      "Epoch 62/1000\n",
      " - 4s - loss: 0.8784 - acc: 0.6796 - val_loss: 1.2932 - val_acc: 0.5000\n",
      "Epoch 63/1000\n",
      " - 4s - loss: 1.0594 - acc: 0.6720 - val_loss: 1.4880 - val_acc: 0.5714\n",
      "Epoch 64/1000\n",
      " - 4s - loss: 0.8636 - acc: 0.7031 - val_loss: 1.4442 - val_acc: 0.5000\n",
      "Epoch 65/1000\n",
      " - 4s - loss: 0.9650 - acc: 0.6952 - val_loss: 1.7892 - val_acc: 0.3214\n",
      "Epoch 66/1000\n",
      " - 4s - loss: 0.8467 - acc: 0.7421 - val_loss: 1.5012 - val_acc: 0.5714\n",
      "Epoch 67/1000\n",
      " - 4s - loss: 0.8306 - acc: 0.7504 - val_loss: 1.7125 - val_acc: 0.3929\n",
      "Epoch 68/1000\n",
      " - 4s - loss: 0.8159 - acc: 0.7032 - val_loss: 1.6822 - val_acc: 0.4286\n",
      "Epoch 69/1000\n",
      " - 4s - loss: 0.7669 - acc: 0.7108 - val_loss: 1.6612 - val_acc: 0.4643\n",
      "Epoch 70/1000\n",
      " - 4s - loss: 0.9652 - acc: 0.6720 - val_loss: 2.0602 - val_acc: 0.3929\n",
      "Epoch 71/1000\n",
      " - 4s - loss: 0.6684 - acc: 0.7579 - val_loss: 1.6237 - val_acc: 0.4643\n",
      "Epoch 72/1000\n",
      " - 4s - loss: 0.7877 - acc: 0.6720 - val_loss: 1.9243 - val_acc: 0.4643\n",
      "Epoch 73/1000\n",
      " - 4s - loss: 1.1089 - acc: 0.6641 - val_loss: 1.7248 - val_acc: 0.4286\n",
      "Epoch 74/1000\n",
      " - 4s - loss: 0.8550 - acc: 0.6953 - val_loss: 1.6330 - val_acc: 0.5000\n",
      "Epoch 75/1000\n",
      " - 4s - loss: 0.5954 - acc: 0.8123 - val_loss: 1.6747 - val_acc: 0.4286\n",
      "Epoch 76/1000\n",
      " - 4s - loss: 0.7562 - acc: 0.6875 - val_loss: 1.7178 - val_acc: 0.5714\n",
      "Epoch 77/1000\n",
      " - 4s - loss: 0.8163 - acc: 0.7422 - val_loss: 1.6112 - val_acc: 0.4286\n",
      "Epoch 78/1000\n",
      " - 4s - loss: 0.9037 - acc: 0.6641 - val_loss: 1.8697 - val_acc: 0.5357\n",
      "Epoch 79/1000\n",
      " - 4s - loss: 0.7934 - acc: 0.6953 - val_loss: 1.5531 - val_acc: 0.5000\n",
      "Epoch 80/1000\n",
      " - 4s - loss: 0.6753 - acc: 0.7815 - val_loss: 1.5818 - val_acc: 0.5000\n",
      "Epoch 81/1000\n",
      " - 4s - loss: 0.9913 - acc: 0.6641 - val_loss: 1.7303 - val_acc: 0.4286\n",
      "Epoch 82/1000\n",
      " - 4s - loss: 0.8127 - acc: 0.7032 - val_loss: 1.7118 - val_acc: 0.5357\n",
      "Epoch 83/1000\n",
      " - 4s - loss: 0.7827 - acc: 0.7186 - val_loss: 1.7498 - val_acc: 0.5714\n",
      "Epoch 84/1000\n",
      " - 4s - loss: 0.9075 - acc: 0.7107 - val_loss: 1.8324 - val_acc: 0.5000\n",
      "Epoch 85/1000\n",
      " - 4s - loss: 0.6756 - acc: 0.7422 - val_loss: 1.8911 - val_acc: 0.5000\n",
      "Epoch 86/1000\n",
      " - 4s - loss: 0.9218 - acc: 0.6877 - val_loss: 1.8619 - val_acc: 0.4286\n",
      "Epoch 87/1000\n",
      " - 4s - loss: 1.0405 - acc: 0.6953 - val_loss: 1.5953 - val_acc: 0.4643\n",
      "Epoch 88/1000\n",
      " - 4s - loss: 0.7482 - acc: 0.7499 - val_loss: 1.6648 - val_acc: 0.4643\n",
      "Epoch 89/1000\n",
      " - 4s - loss: 0.7916 - acc: 0.7032 - val_loss: 1.6608 - val_acc: 0.5357\n",
      "Epoch 90/1000\n",
      " - 4s - loss: 0.8706 - acc: 0.7346 - val_loss: 1.7562 - val_acc: 0.5000\n",
      "Epoch 91/1000\n",
      " - 4s - loss: 0.7143 - acc: 0.7266 - val_loss: 1.5676 - val_acc: 0.5357\n",
      "Epoch 92/1000\n",
      " - 4s - loss: 0.7561 - acc: 0.7890 - val_loss: 1.5060 - val_acc: 0.6071\n",
      "Epoch 93/1000\n",
      " - 4s - loss: 0.5365 - acc: 0.8124 - val_loss: 1.6006 - val_acc: 0.5714\n",
      "Epoch 94/1000\n",
      " - 4s - loss: 0.6580 - acc: 0.7811 - val_loss: 2.1048 - val_acc: 0.3929\n",
      "Epoch 95/1000\n",
      " - 4s - loss: 0.6080 - acc: 0.7890 - val_loss: 1.8513 - val_acc: 0.5000\n",
      "Epoch 96/1000\n",
      " - 4s - loss: 0.6329 - acc: 0.7733 - val_loss: 2.1655 - val_acc: 0.5000\n",
      "Epoch 97/1000\n",
      " - 4s - loss: 0.8659 - acc: 0.6799 - val_loss: 1.8449 - val_acc: 0.3929\n",
      "Epoch 98/1000\n",
      " - 4s - loss: 0.8077 - acc: 0.6953 - val_loss: 1.4593 - val_acc: 0.5357\n",
      "Epoch 99/1000\n",
      " - 4s - loss: 0.7976 - acc: 0.7342 - val_loss: 1.5936 - val_acc: 0.4643\n",
      "Epoch 100/1000\n",
      " - 4s - loss: 0.7139 - acc: 0.7736 - val_loss: 1.7280 - val_acc: 0.4643\n",
      "Epoch 101/1000\n",
      " - 4s - loss: 0.9238 - acc: 0.7267 - val_loss: 1.7842 - val_acc: 0.3929\n",
      "Epoch 102/1000\n",
      " - 4s - loss: 0.6086 - acc: 0.7655 - val_loss: 1.9245 - val_acc: 0.5000\n",
      "Epoch 103/1000\n",
      " - 4s - loss: 0.8533 - acc: 0.7108 - val_loss: 1.8812 - val_acc: 0.6071\n",
      "Epoch 104/1000\n",
      " - 4s - loss: 0.5936 - acc: 0.8280 - val_loss: 1.8147 - val_acc: 0.4643\n",
      "Epoch 105/1000\n",
      " - 4s - loss: 0.8427 - acc: 0.7811 - val_loss: 2.0333 - val_acc: 0.3571\n",
      "Epoch 106/1000\n",
      " - 4s - loss: 0.5645 - acc: 0.7968 - val_loss: 1.4931 - val_acc: 0.5714\n",
      "Epoch 107/1000\n",
      " - 4s - loss: 0.6370 - acc: 0.7968 - val_loss: 1.4602 - val_acc: 0.5000\n",
      "Epoch 108/1000\n",
      " - 4s - loss: 0.6805 - acc: 0.7733 - val_loss: 2.0280 - val_acc: 0.3214\n",
      "Epoch 109/1000\n",
      " - 3s - loss: 0.7777 - acc: 0.7501 - val_loss: 1.4125 - val_acc: 0.5357\n",
      "Epoch 110/1000\n",
      " - 4s - loss: 0.8237 - acc: 0.7500 - val_loss: 1.4462 - val_acc: 0.4643\n",
      "Epoch 111/1000\n",
      " - 4s - loss: 1.0032 - acc: 0.7500 - val_loss: 1.8240 - val_acc: 0.4643\n",
      "Epoch 112/1000\n",
      " - 4s - loss: 0.7331 - acc: 0.7422 - val_loss: 1.8842 - val_acc: 0.4643\n",
      "Epoch 113/1000\n",
      " - 4s - loss: 0.8278 - acc: 0.7578 - val_loss: 1.7000 - val_acc: 0.5000\n",
      "Epoch 114/1000\n",
      " - 4s - loss: 0.6502 - acc: 0.7579 - val_loss: 1.6996 - val_acc: 0.4286\n",
      "Epoch 115/1000\n",
      " - 4s - loss: 0.7363 - acc: 0.7577 - val_loss: 1.8143 - val_acc: 0.3214\n",
      "Epoch 116/1000\n",
      " - 4s - loss: 0.8605 - acc: 0.7109 - val_loss: 1.5961 - val_acc: 0.5000\n",
      "Epoch 117/1000\n",
      " - 4s - loss: 0.8614 - acc: 0.7578 - val_loss: 1.7562 - val_acc: 0.4643\n",
      "Epoch 118/1000\n",
      " - 4s - loss: 0.5742 - acc: 0.7968 - val_loss: 1.8059 - val_acc: 0.5000\n",
      "Epoch 119/1000\n",
      " - 4s - loss: 1.0003 - acc: 0.6641 - val_loss: 2.1656 - val_acc: 0.4286\n",
      "Epoch 120/1000\n",
      " - 4s - loss: 0.8007 - acc: 0.7577 - val_loss: 1.9810 - val_acc: 0.5000\n",
      "Epoch 121/1000\n",
      " - 4s - loss: 0.6167 - acc: 0.7735 - val_loss: 2.2759 - val_acc: 0.5000\n",
      "Epoch 122/1000\n",
      " - 4s - loss: 0.6657 - acc: 0.7109 - val_loss: 2.0421 - val_acc: 0.3929\n",
      "Epoch 123/1000\n",
      " - 4s - loss: 0.7736 - acc: 0.7736 - val_loss: 2.1401 - val_acc: 0.3571\n",
      "Epoch 124/1000\n",
      " - 4s - loss: 0.6036 - acc: 0.7970 - val_loss: 2.1345 - val_acc: 0.3929\n",
      "Epoch 125/1000\n",
      " - 4s - loss: 0.5565 - acc: 0.7655 - val_loss: 1.9236 - val_acc: 0.3929\n",
      "Epoch 126/1000\n",
      " - 4s - loss: 0.7044 - acc: 0.7736 - val_loss: 2.2871 - val_acc: 0.5000\n",
      "Epoch 127/1000\n",
      " - 4s - loss: 0.6814 - acc: 0.7655 - val_loss: 1.8448 - val_acc: 0.5357\n",
      "Epoch 128/1000\n",
      " - 4s - loss: 0.7764 - acc: 0.7421 - val_loss: 1.8378 - val_acc: 0.5357\n",
      "Epoch 129/1000\n",
      " - 4s - loss: 0.9536 - acc: 0.7579 - val_loss: 1.4839 - val_acc: 0.5714\n",
      "Epoch 130/1000\n",
      " - 4s - loss: 0.7443 - acc: 0.7655 - val_loss: 2.1830 - val_acc: 0.5357\n",
      "Epoch 131/1000\n",
      " - 4s - loss: 0.7741 - acc: 0.7579 - val_loss: 1.7631 - val_acc: 0.6071\n",
      "Epoch 132/1000\n",
      " - 4s - loss: 0.6304 - acc: 0.7577 - val_loss: 1.7531 - val_acc: 0.3571\n",
      "Epoch 133/1000\n",
      " - 4s - loss: 0.6325 - acc: 0.7968 - val_loss: 1.9373 - val_acc: 0.4643\n",
      "Epoch 134/1000\n",
      " - 4s - loss: 0.9573 - acc: 0.7108 - val_loss: 1.9118 - val_acc: 0.4286\n",
      "Epoch 135/1000\n",
      " - 4s - loss: 0.7744 - acc: 0.7811 - val_loss: 1.6799 - val_acc: 0.5357\n",
      "Epoch 136/1000\n",
      " - 4s - loss: 0.7144 - acc: 0.7499 - val_loss: 1.9477 - val_acc: 0.4643\n",
      "Epoch 137/1000\n",
      " - 4s - loss: 0.6427 - acc: 0.8046 - val_loss: 2.1772 - val_acc: 0.5714\n",
      "Epoch 138/1000\n",
      " - 4s - loss: 0.8082 - acc: 0.7031 - val_loss: 2.2809 - val_acc: 0.5000\n",
      "Epoch 139/1000\n",
      " - 4s - loss: 0.7332 - acc: 0.7502 - val_loss: 1.5444 - val_acc: 0.5714\n",
      "Epoch 140/1000\n",
      " - 4s - loss: 0.5440 - acc: 0.7970 - val_loss: 2.1949 - val_acc: 0.6071\n",
      "Epoch 141/1000\n",
      " - 4s - loss: 0.8232 - acc: 0.7422 - val_loss: 1.9434 - val_acc: 0.4643\n",
      "Epoch 142/1000\n",
      " - 4s - loss: 0.4930 - acc: 0.8436 - val_loss: 1.2988 - val_acc: 0.5000\n",
      "Epoch 143/1000\n",
      " - 4s - loss: 0.9359 - acc: 0.7347 - val_loss: 2.1808 - val_acc: 0.3571\n",
      "Epoch 144/1000\n",
      " - 4s - loss: 0.8404 - acc: 0.7656 - val_loss: 2.1458 - val_acc: 0.5000\n",
      "Epoch 145/1000\n",
      " - 4s - loss: 0.7351 - acc: 0.8126 - val_loss: 1.9985 - val_acc: 0.5000\n",
      "Epoch 146/1000\n",
      " - 4s - loss: 0.6058 - acc: 0.7811 - val_loss: 2.2171 - val_acc: 0.4286\n",
      "Epoch 147/1000\n",
      " - 4s - loss: 0.8034 - acc: 0.7577 - val_loss: 1.8403 - val_acc: 0.5000\n",
      "Epoch 148/1000\n",
      " - 4s - loss: 0.5531 - acc: 0.7890 - val_loss: 2.0733 - val_acc: 0.5000\n",
      "Epoch 149/1000\n",
      " - 4s - loss: 0.8382 - acc: 0.7501 - val_loss: 1.9990 - val_acc: 0.4286\n",
      "Epoch 150/1000\n",
      " - 4s - loss: 0.7220 - acc: 0.7109 - val_loss: 2.1609 - val_acc: 0.6071\n",
      "Epoch 151/1000\n",
      " - 4s - loss: 0.7213 - acc: 0.7578 - val_loss: 2.3263 - val_acc: 0.3929\n",
      "Epoch 152/1000\n",
      " - 4s - loss: 0.5504 - acc: 0.8359 - val_loss: 1.7942 - val_acc: 0.5000\n",
      "Epoch 153/1000\n",
      " - 4s - loss: 0.7356 - acc: 0.7736 - val_loss: 1.9262 - val_acc: 0.5000\n",
      "Epoch 154/1000\n",
      " - 4s - loss: 0.8303 - acc: 0.7423 - val_loss: 2.0709 - val_acc: 0.5357\n",
      "Epoch 155/1000\n",
      " - 4s - loss: 0.5326 - acc: 0.7968 - val_loss: 2.0041 - val_acc: 0.5357\n",
      "Epoch 156/1000\n",
      " - 4s - loss: 0.4020 - acc: 0.8517 - val_loss: 1.5557 - val_acc: 0.5000\n",
      "Epoch 157/1000\n",
      " - 4s - loss: 0.9669 - acc: 0.7342 - val_loss: 1.6526 - val_acc: 0.4643\n",
      "Epoch 158/1000\n",
      " - 4s - loss: 0.7697 - acc: 0.7814 - val_loss: 1.9820 - val_acc: 0.5000\n",
      "Epoch 159/1000\n",
      " - 4s - loss: 0.5303 - acc: 0.8437 - val_loss: 1.5645 - val_acc: 0.5357\n",
      "Epoch 160/1000\n",
      " - 4s - loss: 0.7369 - acc: 0.7267 - val_loss: 1.7042 - val_acc: 0.5357\n",
      "Epoch 161/1000\n",
      " - 4s - loss: 0.6917 - acc: 0.7811 - val_loss: 1.9444 - val_acc: 0.4643\n",
      "Epoch 162/1000\n",
      " - 4s - loss: 0.7575 - acc: 0.7736 - val_loss: 1.7283 - val_acc: 0.5000\n",
      "Epoch 163/1000\n",
      " - 4s - loss: 0.7483 - acc: 0.7579 - val_loss: 2.1739 - val_acc: 0.4643\n",
      "Epoch 164/1000\n",
      " - 4s - loss: 0.4397 - acc: 0.8202 - val_loss: 2.3786 - val_acc: 0.4643\n",
      "Epoch 165/1000\n",
      " - 4s - loss: 0.6260 - acc: 0.8202 - val_loss: 1.6017 - val_acc: 0.6071\n",
      "Epoch 166/1000\n",
      " - 3s - loss: 0.7720 - acc: 0.7733 - val_loss: 1.7967 - val_acc: 0.5714\n",
      "Epoch 167/1000\n",
      " - 4s - loss: 0.5370 - acc: 0.8280 - val_loss: 1.8067 - val_acc: 0.3929\n",
      "Epoch 168/1000\n",
      " - 4s - loss: 0.8040 - acc: 0.7264 - val_loss: 1.6861 - val_acc: 0.5714\n",
      "Epoch 169/1000\n",
      " - 4s - loss: 0.5777 - acc: 0.8127 - val_loss: 1.5762 - val_acc: 0.6786\n",
      "Epoch 170/1000\n",
      " - 4s - loss: 0.6529 - acc: 0.7499 - val_loss: 2.1837 - val_acc: 0.5714\n",
      "Epoch 171/1000\n",
      " - 4s - loss: 0.5654 - acc: 0.7814 - val_loss: 2.0330 - val_acc: 0.4286\n",
      "Epoch 172/1000\n",
      " - 4s - loss: 0.7181 - acc: 0.7968 - val_loss: 2.4153 - val_acc: 0.4286\n",
      "Epoch 173/1000\n",
      " - 4s - loss: 0.4718 - acc: 0.7970 - val_loss: 1.8650 - val_acc: 0.6071\n",
      "Epoch 174/1000\n",
      " - 4s - loss: 0.8104 - acc: 0.7970 - val_loss: 1.9120 - val_acc: 0.4643\n",
      "Epoch 175/1000\n",
      " - 4s - loss: 0.6175 - acc: 0.7890 - val_loss: 2.0744 - val_acc: 0.5000\n",
      "Epoch 176/1000\n",
      " - 4s - loss: 0.6707 - acc: 0.8203 - val_loss: 1.8969 - val_acc: 0.4643\n",
      "Epoch 177/1000\n",
      " - 4s - loss: 0.5641 - acc: 0.8123 - val_loss: 1.7957 - val_acc: 0.5000\n",
      "Epoch 178/1000\n",
      " - 4s - loss: 0.5684 - acc: 0.8124 - val_loss: 2.0042 - val_acc: 0.4643\n",
      "Epoch 179/1000\n",
      " - 4s - loss: 0.6759 - acc: 0.7814 - val_loss: 2.2222 - val_acc: 0.3571\n",
      "Epoch 180/1000\n",
      " - 4s - loss: 0.6308 - acc: 0.7892 - val_loss: 1.8705 - val_acc: 0.5714\n",
      "Epoch 181/1000\n",
      " - 4s - loss: 0.5158 - acc: 0.8047 - val_loss: 2.0134 - val_acc: 0.5000\n",
      "Epoch 182/1000\n",
      " - 4s - loss: 0.7638 - acc: 0.7505 - val_loss: 1.7044 - val_acc: 0.5714\n",
      "Epoch 183/1000\n",
      " - 4s - loss: 0.8179 - acc: 0.7655 - val_loss: 2.1033 - val_acc: 0.4643\n",
      "Epoch 184/1000\n",
      " - 4s - loss: 0.8972 - acc: 0.7656 - val_loss: 1.5128 - val_acc: 0.5000\n",
      "Epoch 185/1000\n",
      " - 4s - loss: 0.8022 - acc: 0.7031 - val_loss: 2.3566 - val_acc: 0.4643\n",
      "Epoch 186/1000\n",
      " - 4s - loss: 0.8038 - acc: 0.7890 - val_loss: 1.7215 - val_acc: 0.5357\n",
      "Epoch 187/1000\n",
      " - 4s - loss: 0.4489 - acc: 0.8438 - val_loss: 1.8790 - val_acc: 0.4643\n",
      "Epoch 188/1000\n",
      " - 4s - loss: 0.5552 - acc: 0.8437 - val_loss: 2.4945 - val_acc: 0.4286\n",
      "Epoch 189/1000\n",
      " - 4s - loss: 0.6794 - acc: 0.8201 - val_loss: 2.0937 - val_acc: 0.5714\n",
      "Epoch 190/1000\n",
      " - 4s - loss: 0.7802 - acc: 0.7422 - val_loss: 2.0148 - val_acc: 0.3929\n",
      "Epoch 191/1000\n",
      " - 4s - loss: 0.6172 - acc: 0.7735 - val_loss: 1.5607 - val_acc: 0.6071\n",
      "Epoch 192/1000\n",
      " - 4s - loss: 0.5881 - acc: 0.7891 - val_loss: 1.8427 - val_acc: 0.3929\n",
      "Epoch 193/1000\n",
      " - 4s - loss: 0.5978 - acc: 0.7814 - val_loss: 2.7179 - val_acc: 0.4643\n",
      "Epoch 194/1000\n",
      " - 4s - loss: 0.5447 - acc: 0.8124 - val_loss: 1.7696 - val_acc: 0.5000\n",
      "Epoch 195/1000\n",
      " - 4s - loss: 0.6472 - acc: 0.7345 - val_loss: 1.9795 - val_acc: 0.5000\n",
      "Epoch 196/1000\n",
      " - 4s - loss: 0.7463 - acc: 0.7810 - val_loss: 2.0227 - val_acc: 0.4643\n",
      "Epoch 197/1000\n",
      " - 4s - loss: 0.5837 - acc: 0.7814 - val_loss: 2.1548 - val_acc: 0.4286\n",
      "Epoch 198/1000\n",
      " - 4s - loss: 0.4682 - acc: 0.8594 - val_loss: 2.3399 - val_acc: 0.5000\n",
      "Epoch 199/1000\n",
      " - 4s - loss: 0.6316 - acc: 0.7892 - val_loss: 2.3281 - val_acc: 0.5000\n",
      "Epoch 200/1000\n",
      " - 4s - loss: 0.5552 - acc: 0.8280 - val_loss: 2.0517 - val_acc: 0.3929\n",
      "Epoch 201/1000\n",
      " - 4s - loss: 0.5597 - acc: 0.7733 - val_loss: 2.5735 - val_acc: 0.4643\n",
      "Epoch 202/1000\n",
      " - 4s - loss: 0.4630 - acc: 0.8124 - val_loss: 2.1003 - val_acc: 0.5000\n",
      "Epoch 203/1000\n",
      " - 4s - loss: 0.6747 - acc: 0.7658 - val_loss: 2.4075 - val_acc: 0.3571\n",
      "Epoch 204/1000\n",
      " - 4s - loss: 0.4733 - acc: 0.8202 - val_loss: 1.9487 - val_acc: 0.5357\n",
      "Epoch 205/1000\n",
      " - 4s - loss: 0.6778 - acc: 0.7814 - val_loss: 2.6080 - val_acc: 0.3571\n",
      "Epoch 206/1000\n",
      " - 4s - loss: 0.3673 - acc: 0.8515 - val_loss: 2.8004 - val_acc: 0.3929\n",
      "Epoch 207/1000\n",
      " - 4s - loss: 0.4963 - acc: 0.8202 - val_loss: 2.2574 - val_acc: 0.4643\n",
      "Epoch 208/1000\n",
      " - 4s - loss: 0.5993 - acc: 0.8124 - val_loss: 2.0330 - val_acc: 0.6071\n",
      "Epoch 209/1000\n",
      " - 4s - loss: 0.5870 - acc: 0.7811 - val_loss: 2.5803 - val_acc: 0.4643\n",
      "Epoch 210/1000\n",
      " - 4s - loss: 0.4972 - acc: 0.8359 - val_loss: 1.7944 - val_acc: 0.5000\n",
      "Epoch 211/1000\n",
      " - 4s - loss: 0.7324 - acc: 0.7578 - val_loss: 2.5339 - val_acc: 0.5000\n",
      "Epoch 212/1000\n",
      " - 4s - loss: 0.6849 - acc: 0.8202 - val_loss: 2.0999 - val_acc: 0.5714\n",
      "Epoch 213/1000\n",
      " - 4s - loss: 0.8592 - acc: 0.7423 - val_loss: 2.5213 - val_acc: 0.5000\n",
      "Epoch 214/1000\n",
      " - 4s - loss: 0.6512 - acc: 0.7736 - val_loss: 2.3606 - val_acc: 0.5000\n",
      "Epoch 215/1000\n",
      " - 4s - loss: 0.7955 - acc: 0.7815 - val_loss: 2.0207 - val_acc: 0.5000\n",
      "Epoch 216/1000\n",
      " - 4s - loss: 0.6598 - acc: 0.7968 - val_loss: 1.8111 - val_acc: 0.5357\n",
      "Epoch 217/1000\n",
      " - 4s - loss: 0.6674 - acc: 0.7969 - val_loss: 2.0459 - val_acc: 0.5000\n",
      "Epoch 218/1000\n",
      " - 4s - loss: 0.5648 - acc: 0.8129 - val_loss: 2.2003 - val_acc: 0.5357\n",
      "Epoch 219/1000\n",
      " - 4s - loss: 0.6914 - acc: 0.7733 - val_loss: 2.3920 - val_acc: 0.3571\n",
      "Epoch 220/1000\n",
      " - 4s - loss: 0.5311 - acc: 0.8124 - val_loss: 2.2922 - val_acc: 0.4286\n",
      "Epoch 221/1000\n",
      " - 4s - loss: 0.6326 - acc: 0.8127 - val_loss: 2.6342 - val_acc: 0.4643\n",
      "Epoch 222/1000\n",
      " - 4s - loss: 0.7116 - acc: 0.7970 - val_loss: 2.3363 - val_acc: 0.5357\n",
      "Epoch 223/1000\n",
      " - 4s - loss: 0.4920 - acc: 0.8359 - val_loss: 1.9883 - val_acc: 0.5000\n",
      "Epoch 224/1000\n",
      " - 4s - loss: 0.6841 - acc: 0.7814 - val_loss: 2.3482 - val_acc: 0.4643\n",
      "Epoch 225/1000\n",
      " - 4s - loss: 0.3994 - acc: 0.8517 - val_loss: 2.1469 - val_acc: 0.3929\n",
      "Epoch 226/1000\n",
      " - 4s - loss: 0.6801 - acc: 0.7577 - val_loss: 2.8247 - val_acc: 0.5357\n",
      "Epoch 227/1000\n",
      " - 4s - loss: 0.5843 - acc: 0.8048 - val_loss: 2.5248 - val_acc: 0.5714\n",
      "Epoch 228/1000\n",
      " - 4s - loss: 0.7148 - acc: 0.7425 - val_loss: 2.1823 - val_acc: 0.5357\n",
      "Epoch 229/1000\n",
      " - 4s - loss: 0.7067 - acc: 0.7733 - val_loss: 2.1776 - val_acc: 0.5357\n",
      "Epoch 230/1000\n",
      " - 4s - loss: 0.3641 - acc: 0.8830 - val_loss: 2.4135 - val_acc: 0.4286\n",
      "Epoch 231/1000\n",
      " - 4s - loss: 0.5089 - acc: 0.8359 - val_loss: 1.8423 - val_acc: 0.5357\n",
      "Epoch 232/1000\n",
      " - 4s - loss: 0.5037 - acc: 0.7812 - val_loss: 2.3493 - val_acc: 0.5000\n",
      "Epoch 233/1000\n",
      " - 4s - loss: 0.7044 - acc: 0.7891 - val_loss: 2.1409 - val_acc: 0.5714\n",
      "Epoch 234/1000\n",
      " - 4s - loss: 0.5840 - acc: 0.8127 - val_loss: 1.7102 - val_acc: 0.5000\n",
      "Epoch 235/1000\n",
      " - 4s - loss: 0.6219 - acc: 0.7814 - val_loss: 1.9445 - val_acc: 0.5000\n",
      "Epoch 236/1000\n",
      " - 4s - loss: 0.6902 - acc: 0.7658 - val_loss: 2.3938 - val_acc: 0.5714\n",
      "Epoch 237/1000\n",
      " - 4s - loss: 0.4600 - acc: 0.7968 - val_loss: 2.6079 - val_acc: 0.5000\n",
      "Epoch 238/1000\n",
      " - 4s - loss: 0.5188 - acc: 0.8124 - val_loss: 2.9683 - val_acc: 0.3571\n",
      "Epoch 239/1000\n",
      " - 4s - loss: 0.5427 - acc: 0.8047 - val_loss: 2.3515 - val_acc: 0.4286\n",
      "Epoch 240/1000\n",
      " - 4s - loss: 0.7721 - acc: 0.7661 - val_loss: 2.4819 - val_acc: 0.5357\n",
      "Epoch 241/1000\n",
      " - 4s - loss: 0.7201 - acc: 0.8127 - val_loss: 1.7378 - val_acc: 0.5357\n",
      "Epoch 242/1000\n",
      " - 4s - loss: 0.5128 - acc: 0.8127 - val_loss: 2.2512 - val_acc: 0.4643\n",
      "Epoch 243/1000\n",
      " - 4s - loss: 0.6681 - acc: 0.7812 - val_loss: 2.5209 - val_acc: 0.3929\n",
      "Epoch 244/1000\n",
      " - 4s - loss: 0.6171 - acc: 0.8359 - val_loss: 2.1553 - val_acc: 0.4286\n",
      "Epoch 245/1000\n",
      " - 4s - loss: 0.5788 - acc: 0.8046 - val_loss: 2.3143 - val_acc: 0.5000\n",
      "Epoch 246/1000\n",
      " - 4s - loss: 0.5775 - acc: 0.8124 - val_loss: 2.1322 - val_acc: 0.5357\n",
      "Epoch 247/1000\n",
      " - 4s - loss: 0.5116 - acc: 0.8126 - val_loss: 2.5857 - val_acc: 0.5357\n",
      "Epoch 248/1000\n",
      " - 4s - loss: 0.6366 - acc: 0.8124 - val_loss: 1.9382 - val_acc: 0.4286\n",
      "Epoch 249/1000\n",
      " - 4s - loss: 0.6808 - acc: 0.7579 - val_loss: 2.8848 - val_acc: 0.5000\n",
      "Epoch 250/1000\n",
      " - 4s - loss: 0.6492 - acc: 0.7970 - val_loss: 2.1741 - val_acc: 0.5000\n",
      "Epoch 251/1000\n",
      " - 4s - loss: 0.5830 - acc: 0.8202 - val_loss: 2.8062 - val_acc: 0.3571\n",
      "Epoch 252/1000\n",
      " - 4s - loss: 0.6122 - acc: 0.8048 - val_loss: 2.1681 - val_acc: 0.5000\n",
      "Epoch 253/1000\n",
      " - 4s - loss: 0.7151 - acc: 0.7500 - val_loss: 2.6314 - val_acc: 0.5000\n",
      "Epoch 254/1000\n",
      " - 4s - loss: 0.5452 - acc: 0.8361 - val_loss: 2.6550 - val_acc: 0.4286\n",
      "Epoch 255/1000\n",
      " - 4s - loss: 0.5393 - acc: 0.8283 - val_loss: 2.5861 - val_acc: 0.5714\n",
      "Epoch 256/1000\n",
      " - 4s - loss: 0.5377 - acc: 0.8046 - val_loss: 2.4275 - val_acc: 0.5000\n",
      "Epoch 257/1000\n",
      " - 4s - loss: 0.6676 - acc: 0.7656 - val_loss: 2.0626 - val_acc: 0.5000\n",
      "Epoch 258/1000\n",
      " - 4s - loss: 0.4118 - acc: 0.8359 - val_loss: 2.1864 - val_acc: 0.4643\n",
      "Epoch 259/1000\n",
      " - 4s - loss: 0.5071 - acc: 0.8671 - val_loss: 2.7130 - val_acc: 0.3929\n",
      "Epoch 260/1000\n",
      " - 4s - loss: 0.5888 - acc: 0.7736 - val_loss: 2.3770 - val_acc: 0.3214\n",
      "Epoch 261/1000\n",
      " - 4s - loss: 0.5612 - acc: 0.8046 - val_loss: 2.4945 - val_acc: 0.4286\n",
      "Epoch 262/1000\n",
      " - 4s - loss: 0.6681 - acc: 0.7814 - val_loss: 2.7160 - val_acc: 0.3929\n",
      "Epoch 263/1000\n",
      " - 4s - loss: 0.5778 - acc: 0.8124 - val_loss: 2.3321 - val_acc: 0.5000\n",
      "Epoch 264/1000\n",
      " - 4s - loss: 0.8177 - acc: 0.7733 - val_loss: 2.5849 - val_acc: 0.3571\n",
      "Epoch 265/1000\n",
      " - 3s - loss: 0.6973 - acc: 0.7345 - val_loss: 2.2981 - val_acc: 0.4643\n",
      "Epoch 266/1000\n",
      " - 4s - loss: 0.6834 - acc: 0.7890 - val_loss: 2.4543 - val_acc: 0.2857\n",
      "Epoch 267/1000\n",
      " - 4s - loss: 0.6443 - acc: 0.7814 - val_loss: 1.5797 - val_acc: 0.5357\n",
      "Epoch 268/1000\n",
      " - 4s - loss: 0.6636 - acc: 0.7894 - val_loss: 2.5565 - val_acc: 0.5357\n",
      "Epoch 269/1000\n",
      " - 4s - loss: 0.8010 - acc: 0.7423 - val_loss: 3.0582 - val_acc: 0.4286\n",
      "Epoch 270/1000\n",
      " - 4s - loss: 0.4467 - acc: 0.8749 - val_loss: 2.4222 - val_acc: 0.4643\n",
      "Epoch 271/1000\n",
      " - 4s - loss: 0.6451 - acc: 0.8361 - val_loss: 2.1998 - val_acc: 0.4643\n",
      "Epoch 272/1000\n",
      " - 4s - loss: 0.4718 - acc: 0.8283 - val_loss: 1.7807 - val_acc: 0.5357\n",
      "Epoch 273/1000\n",
      " - 4s - loss: 0.6385 - acc: 0.7891 - val_loss: 2.6360 - val_acc: 0.5357\n",
      "Epoch 274/1000\n",
      " - 4s - loss: 0.5710 - acc: 0.8046 - val_loss: 2.5000 - val_acc: 0.5000\n",
      "Epoch 275/1000\n",
      " - 4s - loss: 0.5877 - acc: 0.8204 - val_loss: 2.1347 - val_acc: 0.4643\n",
      "Epoch 276/1000\n",
      " - 4s - loss: 0.4594 - acc: 0.8516 - val_loss: 2.1928 - val_acc: 0.4286\n",
      "Epoch 277/1000\n",
      " - 4s - loss: 0.6151 - acc: 0.7969 - val_loss: 2.4816 - val_acc: 0.5000\n",
      "Epoch 278/1000\n",
      " - 4s - loss: 0.5740 - acc: 0.8127 - val_loss: 2.9045 - val_acc: 0.4643\n",
      "Epoch 279/1000\n",
      " - 4s - loss: 0.5868 - acc: 0.7814 - val_loss: 2.1598 - val_acc: 0.4286\n",
      "Epoch 280/1000\n",
      " - 4s - loss: 0.5288 - acc: 0.8361 - val_loss: 2.2272 - val_acc: 0.5000\n",
      "Epoch 281/1000\n",
      " - 4s - loss: 0.5669 - acc: 0.8205 - val_loss: 2.3993 - val_acc: 0.5357\n",
      "Epoch 282/1000\n",
      " - 4s - loss: 0.5070 - acc: 0.8280 - val_loss: 2.5195 - val_acc: 0.5000\n",
      "Epoch 283/1000\n",
      " - 4s - loss: 0.4843 - acc: 0.8359 - val_loss: 2.9196 - val_acc: 0.3571\n",
      "Epoch 284/1000\n",
      " - 4s - loss: 0.5269 - acc: 0.7813 - val_loss: 2.2984 - val_acc: 0.4286\n",
      "Epoch 285/1000\n",
      " - 4s - loss: 0.4671 - acc: 0.8359 - val_loss: 2.2709 - val_acc: 0.4643\n",
      "Epoch 286/1000\n",
      " - 4s - loss: 0.5424 - acc: 0.8280 - val_loss: 2.6067 - val_acc: 0.4286\n",
      "Epoch 287/1000\n",
      " - 4s - loss: 0.4850 - acc: 0.8124 - val_loss: 2.7176 - val_acc: 0.5714\n",
      "Epoch 288/1000\n",
      " - 4s - loss: 0.4702 - acc: 0.8280 - val_loss: 2.5675 - val_acc: 0.4286\n",
      "Epoch 289/1000\n",
      " - 4s - loss: 0.6022 - acc: 0.8202 - val_loss: 2.5769 - val_acc: 0.4643\n",
      "Epoch 290/1000\n",
      " - 4s - loss: 0.5273 - acc: 0.8202 - val_loss: 2.4231 - val_acc: 0.4286\n",
      "Epoch 291/1000\n",
      " - 4s - loss: 0.4926 - acc: 0.8280 - val_loss: 2.2091 - val_acc: 0.4643\n",
      "Epoch 292/1000\n",
      " - 4s - loss: 0.3698 - acc: 0.8515 - val_loss: 2.4915 - val_acc: 0.5357\n",
      "Epoch 293/1000\n",
      " - 4s - loss: 0.4431 - acc: 0.8593 - val_loss: 2.4248 - val_acc: 0.5000\n",
      "Epoch 294/1000\n",
      " - 4s - loss: 0.5753 - acc: 0.8359 - val_loss: 2.3368 - val_acc: 0.4643\n",
      "Epoch 295/1000\n",
      " - 4s - loss: 0.3908 - acc: 0.8359 - val_loss: 2.7314 - val_acc: 0.4286\n",
      "Epoch 296/1000\n",
      " - 4s - loss: 0.6179 - acc: 0.8280 - val_loss: 2.2284 - val_acc: 0.5714\n",
      "Epoch 297/1000\n",
      " - 4s - loss: 0.4674 - acc: 0.8280 - val_loss: 2.3072 - val_acc: 0.3929\n",
      "Epoch 298/1000\n",
      " - 4s - loss: 0.4514 - acc: 0.8515 - val_loss: 2.4936 - val_acc: 0.5357\n",
      "Epoch 299/1000\n",
      " - 4s - loss: 0.6335 - acc: 0.8048 - val_loss: 2.2256 - val_acc: 0.5357\n",
      "Epoch 300/1000\n",
      " - 4s - loss: 0.5010 - acc: 0.8359 - val_loss: 2.5296 - val_acc: 0.5000\n",
      "Epoch 301/1000\n",
      " - 4s - loss: 0.3756 - acc: 0.8593 - val_loss: 2.5888 - val_acc: 0.3929\n",
      "Epoch 302/1000\n",
      " - 4s - loss: 0.5744 - acc: 0.7970 - val_loss: 2.5813 - val_acc: 0.5357\n",
      "Epoch 303/1000\n",
      " - 4s - loss: 0.4685 - acc: 0.8671 - val_loss: 3.0534 - val_acc: 0.4286\n",
      "Epoch 304/1000\n",
      " - 4s - loss: 0.5444 - acc: 0.8124 - val_loss: 2.4310 - val_acc: 0.5000\n",
      "Epoch 305/1000\n",
      " - 4s - loss: 0.5711 - acc: 0.8124 - val_loss: 2.2550 - val_acc: 0.4643\n",
      "Epoch 306/1000\n",
      " - 4s - loss: 0.3986 - acc: 0.8437 - val_loss: 2.0656 - val_acc: 0.5357\n",
      "Epoch 307/1000\n",
      " - 4s - loss: 0.4365 - acc: 0.8595 - val_loss: 1.8414 - val_acc: 0.5000\n",
      "Epoch 308/1000\n",
      " - 4s - loss: 0.5185 - acc: 0.8438 - val_loss: 2.2819 - val_acc: 0.5000\n",
      "Epoch 309/1000\n",
      " - 4s - loss: 0.4517 - acc: 0.8437 - val_loss: 2.6523 - val_acc: 0.5000\n",
      "Epoch 310/1000\n",
      " - 4s - loss: 0.4082 - acc: 0.8595 - val_loss: 2.0774 - val_acc: 0.5000\n",
      "Epoch 311/1000\n",
      " - 4s - loss: 0.3956 - acc: 0.8437 - val_loss: 2.4260 - val_acc: 0.5000\n",
      "Epoch 312/1000\n",
      " - 4s - loss: 0.2930 - acc: 0.8749 - val_loss: 3.2473 - val_acc: 0.5000\n",
      "Epoch 313/1000\n",
      " - 4s - loss: 0.4123 - acc: 0.8672 - val_loss: 2.3381 - val_acc: 0.3214\n",
      "Epoch 314/1000\n",
      " - 4s - loss: 0.6907 - acc: 0.7811 - val_loss: 2.3732 - val_acc: 0.3214\n",
      "Epoch 315/1000\n",
      " - 4s - loss: 0.5926 - acc: 0.8126 - val_loss: 2.3937 - val_acc: 0.4643\n",
      "Epoch 316/1000\n",
      " - 4s - loss: 0.5503 - acc: 0.7812 - val_loss: 3.0326 - val_acc: 0.4643\n",
      "Epoch 317/1000\n",
      " - 4s - loss: 0.4843 - acc: 0.8358 - val_loss: 2.0851 - val_acc: 0.5357\n",
      "Epoch 318/1000\n",
      " - 4s - loss: 0.5236 - acc: 0.7969 - val_loss: 2.3726 - val_acc: 0.5357\n",
      "Epoch 319/1000\n",
      " - 4s - loss: 0.2234 - acc: 0.8986 - val_loss: 2.0468 - val_acc: 0.5000\n",
      "Epoch 320/1000\n",
      " - 4s - loss: 0.4354 - acc: 0.8906 - val_loss: 2.1279 - val_acc: 0.4643\n",
      "Epoch 321/1000\n",
      " - 4s - loss: 0.4752 - acc: 0.8515 - val_loss: 2.6415 - val_acc: 0.4286\n",
      "Epoch 322/1000\n",
      " - 4s - loss: 0.5998 - acc: 0.8283 - val_loss: 2.8602 - val_acc: 0.4286\n",
      "Epoch 323/1000\n",
      " - 4s - loss: 0.5906 - acc: 0.8201 - val_loss: 2.4546 - val_acc: 0.5000\n",
      "Epoch 324/1000\n",
      " - 4s - loss: 0.6138 - acc: 0.8047 - val_loss: 2.3568 - val_acc: 0.4286\n",
      "Epoch 325/1000\n",
      " - 4s - loss: 0.3817 - acc: 0.8515 - val_loss: 2.5908 - val_acc: 0.5000\n",
      "Epoch 326/1000\n",
      " - 4s - loss: 0.6054 - acc: 0.8048 - val_loss: 1.7720 - val_acc: 0.5357\n",
      "Epoch 327/1000\n",
      " - 4s - loss: 0.4636 - acc: 0.8280 - val_loss: 2.3659 - val_acc: 0.3571\n",
      "Epoch 328/1000\n",
      " - 4s - loss: 0.4482 - acc: 0.8752 - val_loss: 2.6795 - val_acc: 0.4286\n",
      "Epoch 329/1000\n",
      " - 4s - loss: 0.6638 - acc: 0.7892 - val_loss: 2.5506 - val_acc: 0.4286\n",
      "Epoch 330/1000\n",
      " - 4s - loss: 0.4721 - acc: 0.8127 - val_loss: 2.9469 - val_acc: 0.3929\n",
      "Epoch 331/1000\n",
      " - 3s - loss: 0.6241 - acc: 0.7892 - val_loss: 2.2918 - val_acc: 0.5357\n",
      "Epoch 332/1000\n",
      " - 4s - loss: 0.6287 - acc: 0.8283 - val_loss: 2.1786 - val_acc: 0.4643\n",
      "Epoch 333/1000\n",
      " - 4s - loss: 0.4702 - acc: 0.8437 - val_loss: 2.6109 - val_acc: 0.5000\n",
      "Epoch 334/1000\n",
      " - 4s - loss: 0.3975 - acc: 0.8517 - val_loss: 2.6624 - val_acc: 0.3929\n",
      "Epoch 335/1000\n",
      " - 4s - loss: 0.4908 - acc: 0.8359 - val_loss: 2.5356 - val_acc: 0.4643\n",
      "Epoch 336/1000\n",
      " - 4s - loss: 0.4663 - acc: 0.7970 - val_loss: 2.7358 - val_acc: 0.5357\n",
      "Epoch 337/1000\n",
      " - 4s - loss: 0.4640 - acc: 0.8517 - val_loss: 2.1922 - val_acc: 0.5000\n",
      "Epoch 338/1000\n",
      " - 4s - loss: 0.4412 - acc: 0.8359 - val_loss: 2.5268 - val_acc: 0.4643\n",
      "Epoch 339/1000\n",
      " - 4s - loss: 0.2952 - acc: 0.8984 - val_loss: 1.5362 - val_acc: 0.4643\n",
      "Epoch 340/1000\n",
      " - 4s - loss: 0.4675 - acc: 0.8124 - val_loss: 2.3037 - val_acc: 0.4643\n",
      "Epoch 341/1000\n",
      " - 4s - loss: 0.3550 - acc: 0.8749 - val_loss: 2.4557 - val_acc: 0.4286\n",
      "Epoch 342/1000\n",
      " - 4s - loss: 0.3167 - acc: 0.8830 - val_loss: 2.2567 - val_acc: 0.5000\n",
      "Epoch 343/1000\n",
      " - 4s - loss: 0.4190 - acc: 0.8750 - val_loss: 2.3129 - val_acc: 0.5000\n",
      "Epoch 344/1000\n",
      " - 4s - loss: 0.7698 - acc: 0.7732 - val_loss: 2.0904 - val_acc: 0.4286\n",
      "Epoch 345/1000\n",
      " - 4s - loss: 0.4200 - acc: 0.8359 - val_loss: 2.0610 - val_acc: 0.4286\n",
      "Epoch 346/1000\n",
      " - 4s - loss: 0.5072 - acc: 0.8437 - val_loss: 2.3088 - val_acc: 0.5000\n",
      "Epoch 347/1000\n",
      " - 4s - loss: 0.5668 - acc: 0.8281 - val_loss: 2.3238 - val_acc: 0.3929\n",
      "Epoch 348/1000\n",
      " - 4s - loss: 0.3879 - acc: 0.8595 - val_loss: 2.0830 - val_acc: 0.5714\n",
      "Epoch 349/1000\n",
      " - 4s - loss: 0.4491 - acc: 0.8595 - val_loss: 2.5843 - val_acc: 0.5714\n",
      "Epoch 350/1000\n",
      " - 4s - loss: 0.5509 - acc: 0.8437 - val_loss: 2.1452 - val_acc: 0.5000\n",
      "Epoch 351/1000\n",
      " - 4s - loss: 0.3747 - acc: 0.8515 - val_loss: 2.4063 - val_acc: 0.4286\n",
      "Epoch 352/1000\n",
      " - 4s - loss: 0.5322 - acc: 0.8359 - val_loss: 2.9560 - val_acc: 0.3214\n",
      "Epoch 353/1000\n",
      " - 4s - loss: 0.4204 - acc: 0.8592 - val_loss: 2.8479 - val_acc: 0.5000\n",
      "Epoch 354/1000\n",
      " - 4s - loss: 0.3526 - acc: 0.8750 - val_loss: 3.0699 - val_acc: 0.3571\n",
      "Epoch 355/1000\n",
      " - 4s - loss: 0.4909 - acc: 0.8517 - val_loss: 2.1360 - val_acc: 0.5357\n",
      "Epoch 356/1000\n",
      " - 4s - loss: 0.4500 - acc: 0.8830 - val_loss: 1.7840 - val_acc: 0.5714\n",
      "Epoch 357/1000\n",
      " - 4s - loss: 0.5826 - acc: 0.8359 - val_loss: 2.5059 - val_acc: 0.4286\n",
      "Epoch 358/1000\n",
      " - 4s - loss: 0.4020 - acc: 0.8671 - val_loss: 2.3816 - val_acc: 0.5000\n",
      "Epoch 359/1000\n",
      " - 4s - loss: 0.6923 - acc: 0.8124 - val_loss: 2.0213 - val_acc: 0.4643\n",
      "Epoch 360/1000\n",
      " - 4s - loss: 0.5285 - acc: 0.8515 - val_loss: 2.4503 - val_acc: 0.4643\n",
      "Epoch 361/1000\n",
      " - 4s - loss: 0.5191 - acc: 0.8515 - val_loss: 1.6302 - val_acc: 0.5357\n",
      "Epoch 362/1000\n",
      " - 4s - loss: 0.5147 - acc: 0.8124 - val_loss: 3.1425 - val_acc: 0.4286\n",
      "Epoch 363/1000\n",
      " - 4s - loss: 0.4222 - acc: 0.8749 - val_loss: 2.8835 - val_acc: 0.3571\n",
      "Epoch 364/1000\n",
      " - 3s - loss: 0.4477 - acc: 0.8515 - val_loss: 3.2684 - val_acc: 0.4643\n",
      "Epoch 365/1000\n",
      " - 4s - loss: 0.3628 - acc: 0.8671 - val_loss: 2.7695 - val_acc: 0.3929\n",
      "Epoch 366/1000\n",
      " - 4s - loss: 0.2417 - acc: 0.9218 - val_loss: 2.6174 - val_acc: 0.3571\n",
      "Epoch 367/1000\n",
      " - 4s - loss: 0.4566 - acc: 0.8439 - val_loss: 1.6509 - val_acc: 0.5714\n",
      "Epoch 368/1000\n",
      " - 4s - loss: 0.4497 - acc: 0.8205 - val_loss: 2.4692 - val_acc: 0.5000\n",
      "Epoch 369/1000\n",
      " - 4s - loss: 0.6222 - acc: 0.8205 - val_loss: 2.6044 - val_acc: 0.5357\n",
      "Epoch 370/1000\n",
      " - 4s - loss: 0.4829 - acc: 0.8437 - val_loss: 2.6116 - val_acc: 0.4286\n",
      "Epoch 371/1000\n",
      " - 4s - loss: 0.4302 - acc: 0.8516 - val_loss: 2.4591 - val_acc: 0.4643\n",
      "Epoch 372/1000\n",
      " - 4s - loss: 0.3979 - acc: 0.8595 - val_loss: 2.2925 - val_acc: 0.4643\n",
      "Epoch 373/1000\n",
      " - 4s - loss: 0.2650 - acc: 0.8671 - val_loss: 2.2233 - val_acc: 0.4643\n",
      "Epoch 374/1000\n",
      " - 4s - loss: 0.3490 - acc: 0.8906 - val_loss: 2.6551 - val_acc: 0.3929\n",
      "Epoch 375/1000\n",
      " - 4s - loss: 0.4931 - acc: 0.8203 - val_loss: 2.5052 - val_acc: 0.4286\n",
      "Epoch 376/1000\n",
      " - 4s - loss: 0.4122 - acc: 0.8592 - val_loss: 3.0810 - val_acc: 0.4286\n",
      "Epoch 377/1000\n",
      " - 4s - loss: 0.3526 - acc: 0.8984 - val_loss: 2.2391 - val_acc: 0.3929\n",
      "Epoch 378/1000\n",
      " - 4s - loss: 0.4301 - acc: 0.8204 - val_loss: 2.8385 - val_acc: 0.4643\n",
      "Epoch 379/1000\n",
      " - 4s - loss: 0.3638 - acc: 0.8752 - val_loss: 3.0538 - val_acc: 0.3929\n",
      "Epoch 380/1000\n",
      " - 4s - loss: 0.5046 - acc: 0.8749 - val_loss: 2.2537 - val_acc: 0.4286\n",
      "Epoch 381/1000\n",
      " - 4s - loss: 0.4244 - acc: 0.8515 - val_loss: 2.3092 - val_acc: 0.4643\n",
      "Epoch 382/1000\n",
      " - 4s - loss: 0.4572 - acc: 0.8828 - val_loss: 2.3591 - val_acc: 0.5714\n",
      "Epoch 383/1000\n",
      " - 4s - loss: 0.3140 - acc: 0.8906 - val_loss: 2.0273 - val_acc: 0.5000\n",
      "Epoch 384/1000\n",
      " - 4s - loss: 0.3120 - acc: 0.8984 - val_loss: 2.8413 - val_acc: 0.5000\n",
      "Epoch 385/1000\n",
      " - 4s - loss: 0.6232 - acc: 0.8437 - val_loss: 2.8642 - val_acc: 0.4286\n",
      "Epoch 386/1000\n",
      " - 4s - loss: 0.4849 - acc: 0.8438 - val_loss: 2.0623 - val_acc: 0.3929\n",
      "Epoch 387/1000\n",
      " - 4s - loss: 0.5198 - acc: 0.8438 - val_loss: 2.6957 - val_acc: 0.5357\n",
      "Epoch 388/1000\n",
      " - 4s - loss: 0.3501 - acc: 0.8593 - val_loss: 2.7851 - val_acc: 0.2857\n",
      "Epoch 389/1000\n",
      " - 4s - loss: 0.4098 - acc: 0.8360 - val_loss: 2.2696 - val_acc: 0.4643\n",
      "Epoch 390/1000\n",
      " - 4s - loss: 0.6958 - acc: 0.8125 - val_loss: 2.2159 - val_acc: 0.5000\n",
      "Epoch 391/1000\n",
      " - 4s - loss: 0.2832 - acc: 0.8828 - val_loss: 2.2483 - val_acc: 0.4643\n",
      "Epoch 392/1000\n",
      " - 4s - loss: 0.3416 - acc: 0.8595 - val_loss: 2.5595 - val_acc: 0.5000\n",
      "Epoch 393/1000\n",
      " - 4s - loss: 0.5281 - acc: 0.8047 - val_loss: 2.4902 - val_acc: 0.5357\n",
      "Epoch 394/1000\n",
      " - 4s - loss: 0.4108 - acc: 0.8517 - val_loss: 2.3917 - val_acc: 0.5714\n",
      "Epoch 395/1000\n",
      " - 4s - loss: 0.6255 - acc: 0.8517 - val_loss: 2.6305 - val_acc: 0.5714\n",
      "Epoch 396/1000\n",
      " - 4s - loss: 0.3818 - acc: 0.8828 - val_loss: 2.8068 - val_acc: 0.4643\n",
      "Epoch 397/1000\n",
      " - 4s - loss: 0.4759 - acc: 0.8437 - val_loss: 1.9631 - val_acc: 0.5000\n",
      "Epoch 398/1000\n",
      " - 5s - loss: 0.3973 - acc: 0.8749 - val_loss: 2.2770 - val_acc: 0.4286\n",
      "Epoch 399/1000\n",
      " - 4s - loss: 0.4262 - acc: 0.8906 - val_loss: 2.0015 - val_acc: 0.5000\n",
      "Epoch 400/1000\n",
      " - 4s - loss: 0.3088 - acc: 0.8828 - val_loss: 2.2988 - val_acc: 0.5000\n",
      "Epoch 401/1000\n",
      " - 4s - loss: 0.3639 - acc: 0.8828 - val_loss: 2.5362 - val_acc: 0.5357\n",
      "Epoch 402/1000\n",
      " - 4s - loss: 0.4961 - acc: 0.8438 - val_loss: 2.9314 - val_acc: 0.5000\n",
      "Epoch 403/1000\n",
      " - 4s - loss: 0.4249 - acc: 0.8359 - val_loss: 2.3382 - val_acc: 0.4643\n",
      "Epoch 404/1000\n",
      " - 4s - loss: 0.5472 - acc: 0.8046 - val_loss: 2.3157 - val_acc: 0.4643\n",
      "Epoch 405/1000\n",
      " - 4s - loss: 0.4701 - acc: 0.8359 - val_loss: 2.5467 - val_acc: 0.4286\n",
      "Epoch 406/1000\n",
      " - 4s - loss: 0.2761 - acc: 0.8827 - val_loss: 2.7759 - val_acc: 0.5000\n",
      "Epoch 407/1000\n",
      " - 4s - loss: 0.3443 - acc: 0.8671 - val_loss: 2.3206 - val_acc: 0.3571\n",
      "Epoch 408/1000\n",
      " - 4s - loss: 0.3505 - acc: 0.8828 - val_loss: 2.7263 - val_acc: 0.4643\n",
      "Epoch 409/1000\n",
      " - 4s - loss: 0.3762 - acc: 0.8595 - val_loss: 2.9271 - val_acc: 0.4286\n",
      "Epoch 410/1000\n",
      " - 4s - loss: 0.2466 - acc: 0.9062 - val_loss: 1.7089 - val_acc: 0.5714\n",
      "Epoch 411/1000\n",
      " - 4s - loss: 0.4766 - acc: 0.8517 - val_loss: 2.6320 - val_acc: 0.4643\n",
      "Epoch 412/1000\n",
      " - 4s - loss: 0.3394 - acc: 0.9062 - val_loss: 2.2712 - val_acc: 0.5000\n",
      "Epoch 413/1000\n",
      " - 4s - loss: 0.4140 - acc: 0.8515 - val_loss: 3.5249 - val_acc: 0.4643\n",
      "Epoch 414/1000\n",
      " - 4s - loss: 0.3209 - acc: 0.8828 - val_loss: 2.9497 - val_acc: 0.4286\n",
      "Epoch 415/1000\n",
      " - 4s - loss: 0.3227 - acc: 0.8516 - val_loss: 1.8906 - val_acc: 0.5357\n",
      "Epoch 416/1000\n",
      " - 4s - loss: 0.3518 - acc: 0.8905 - val_loss: 2.5738 - val_acc: 0.4643\n",
      "Epoch 417/1000\n",
      " - 4s - loss: 0.2091 - acc: 0.9297 - val_loss: 2.9752 - val_acc: 0.5000\n",
      "Epoch 418/1000\n",
      " - 4s - loss: 0.3613 - acc: 0.9062 - val_loss: 2.1357 - val_acc: 0.5714\n",
      "Epoch 419/1000\n",
      " - 5s - loss: 0.3864 - acc: 0.8671 - val_loss: 2.4683 - val_acc: 0.5357\n",
      "Epoch 420/1000\n",
      " - 4s - loss: 0.3144 - acc: 0.8983 - val_loss: 2.6298 - val_acc: 0.4643\n",
      "Epoch 421/1000\n",
      " - 4s - loss: 0.2772 - acc: 0.8906 - val_loss: 2.3925 - val_acc: 0.4643\n",
      "Epoch 422/1000\n",
      " - 4s - loss: 0.3728 - acc: 0.8984 - val_loss: 2.0312 - val_acc: 0.4643\n",
      "Epoch 423/1000\n",
      " - 4s - loss: 0.3447 - acc: 0.8984 - val_loss: 2.4739 - val_acc: 0.4643\n",
      "Epoch 424/1000\n",
      " - 4s - loss: 0.4347 - acc: 0.8280 - val_loss: 3.1420 - val_acc: 0.4286\n",
      "Epoch 425/1000\n",
      " - 4s - loss: 0.3587 - acc: 0.8515 - val_loss: 2.3441 - val_acc: 0.3214\n",
      "Epoch 426/1000\n",
      " - 4s - loss: 0.3639 - acc: 0.8671 - val_loss: 2.6375 - val_acc: 0.3571\n",
      "Epoch 427/1000\n",
      " - 4s - loss: 0.3137 - acc: 0.9062 - val_loss: 2.9148 - val_acc: 0.4286\n",
      "Epoch 428/1000\n",
      " - 4s - loss: 0.5549 - acc: 0.8360 - val_loss: 1.8641 - val_acc: 0.4643\n",
      "Epoch 429/1000\n",
      " - 4s - loss: 0.3722 - acc: 0.8594 - val_loss: 2.4186 - val_acc: 0.5714\n",
      "Epoch 430/1000\n",
      " - 4s - loss: 0.2416 - acc: 0.9140 - val_loss: 2.3850 - val_acc: 0.3929\n",
      "Epoch 431/1000\n",
      " - 5s - loss: 0.4391 - acc: 0.8593 - val_loss: 2.5208 - val_acc: 0.5714\n",
      "Epoch 432/1000\n",
      " - 4s - loss: 0.3974 - acc: 0.8906 - val_loss: 1.8247 - val_acc: 0.4643\n",
      "Epoch 433/1000\n",
      " - 4s - loss: 0.4030 - acc: 0.8749 - val_loss: 3.0643 - val_acc: 0.3571\n",
      "Epoch 434/1000\n",
      " - 4s - loss: 0.3310 - acc: 0.8828 - val_loss: 2.6060 - val_acc: 0.4643\n",
      "Epoch 435/1000\n",
      " - 4s - loss: 0.4036 - acc: 0.8515 - val_loss: 2.3283 - val_acc: 0.5357\n",
      "Epoch 436/1000\n",
      " - 4s - loss: 0.5143 - acc: 0.8517 - val_loss: 2.2260 - val_acc: 0.5714\n",
      "Epoch 437/1000\n",
      " - 4s - loss: 0.3559 - acc: 0.8908 - val_loss: 2.2584 - val_acc: 0.5357\n",
      "Epoch 438/1000\n",
      " - 4s - loss: 0.5383 - acc: 0.8125 - val_loss: 3.0118 - val_acc: 0.4286\n",
      "Epoch 439/1000\n",
      " - 4s - loss: 0.2207 - acc: 0.9218 - val_loss: 2.9283 - val_acc: 0.4286\n",
      "Epoch 440/1000\n",
      " - 4s - loss: 0.2466 - acc: 0.8984 - val_loss: 1.9689 - val_acc: 0.5000\n",
      "Epoch 441/1000\n",
      " - 4s - loss: 0.3066 - acc: 0.8983 - val_loss: 2.8719 - val_acc: 0.4643\n",
      "Epoch 442/1000\n",
      " - 4s - loss: 0.5152 - acc: 0.8359 - val_loss: 2.5820 - val_acc: 0.3929\n",
      "Epoch 443/1000\n",
      " - 4s - loss: 0.4087 - acc: 0.8436 - val_loss: 2.0974 - val_acc: 0.5000\n",
      "Epoch 444/1000\n",
      " - 4s - loss: 0.3587 - acc: 0.8750 - val_loss: 1.9594 - val_acc: 0.5000\n",
      "Epoch 445/1000\n",
      " - 4s - loss: 0.3742 - acc: 0.8905 - val_loss: 2.2123 - val_acc: 0.4286\n",
      "Epoch 446/1000\n",
      " - 4s - loss: 0.4431 - acc: 0.8437 - val_loss: 2.6489 - val_acc: 0.4643\n",
      "Epoch 447/1000\n",
      " - 4s - loss: 0.4094 - acc: 0.8828 - val_loss: 2.5319 - val_acc: 0.4286\n",
      "Epoch 448/1000\n",
      " - 4s - loss: 0.2820 - acc: 0.8984 - val_loss: 2.1056 - val_acc: 0.3571\n",
      "Epoch 449/1000\n",
      " - 4s - loss: 0.4306 - acc: 0.8515 - val_loss: 2.6235 - val_acc: 0.5357\n",
      "Epoch 450/1000\n",
      " - 4s - loss: 0.5060 - acc: 0.8441 - val_loss: 2.9624 - val_acc: 0.4643\n",
      "Epoch 451/1000\n",
      " - 4s - loss: 0.3499 - acc: 0.8828 - val_loss: 2.4048 - val_acc: 0.4643\n",
      "Epoch 452/1000\n",
      " - 5s - loss: 0.2154 - acc: 0.9062 - val_loss: 2.1213 - val_acc: 0.5000\n",
      "Epoch 453/1000\n",
      " - 4s - loss: 0.2439 - acc: 0.9062 - val_loss: 2.7979 - val_acc: 0.5357\n",
      "Epoch 454/1000\n",
      " - 4s - loss: 0.3227 - acc: 0.8671 - val_loss: 2.6946 - val_acc: 0.5000\n",
      "Epoch 455/1000\n",
      " - 4s - loss: 0.4529 - acc: 0.8437 - val_loss: 2.3363 - val_acc: 0.4286\n",
      "Epoch 456/1000\n",
      " - 4s - loss: 0.2600 - acc: 0.8828 - val_loss: 3.5065 - val_acc: 0.3929\n",
      "Epoch 457/1000\n",
      " - 4s - loss: 0.2589 - acc: 0.8984 - val_loss: 2.7591 - val_acc: 0.4643\n",
      "Epoch 458/1000\n",
      " - 4s - loss: 0.4252 - acc: 0.8283 - val_loss: 2.1224 - val_acc: 0.5357\n",
      "Epoch 459/1000\n",
      " - 4s - loss: 0.3266 - acc: 0.8828 - val_loss: 2.3139 - val_acc: 0.5000\n",
      "Epoch 460/1000\n",
      " - 4s - loss: 0.4389 - acc: 0.8749 - val_loss: 2.4989 - val_acc: 0.4643\n",
      "Epoch 461/1000\n",
      " - 4s - loss: 0.4599 - acc: 0.8437 - val_loss: 2.3918 - val_acc: 0.5000\n",
      "Epoch 462/1000\n",
      " - 4s - loss: 0.4050 - acc: 0.8671 - val_loss: 2.1820 - val_acc: 0.6071\n",
      "Epoch 463/1000\n",
      " - 4s - loss: 0.5279 - acc: 0.8048 - val_loss: 2.9629 - val_acc: 0.5000\n",
      "Epoch 464/1000\n",
      " - 4s - loss: 0.3699 - acc: 0.8828 - val_loss: 2.3131 - val_acc: 0.6071\n",
      "Epoch 465/1000\n",
      " - 4s - loss: 0.2661 - acc: 0.9062 - val_loss: 2.8784 - val_acc: 0.5000\n",
      "Epoch 466/1000\n",
      " - 4s - loss: 0.3193 - acc: 0.8751 - val_loss: 2.5240 - val_acc: 0.6071\n",
      "Epoch 467/1000\n",
      " - 4s - loss: 0.3705 - acc: 0.8671 - val_loss: 2.7751 - val_acc: 0.5000\n",
      "Epoch 468/1000\n",
      " - 4s - loss: 0.4729 - acc: 0.8671 - val_loss: 2.5291 - val_acc: 0.5357\n",
      "Epoch 469/1000\n",
      " - 4s - loss: 0.2662 - acc: 0.8828 - val_loss: 1.9689 - val_acc: 0.5714\n",
      "Epoch 470/1000\n",
      " - 4s - loss: 0.3942 - acc: 0.8515 - val_loss: 3.3297 - val_acc: 0.4643\n",
      "Epoch 471/1000\n",
      " - 4s - loss: 0.3233 - acc: 0.9141 - val_loss: 2.6828 - val_acc: 0.4286\n",
      "Epoch 472/1000\n",
      " - 4s - loss: 0.4345 - acc: 0.8671 - val_loss: 2.4560 - val_acc: 0.4286\n",
      "Epoch 473/1000\n",
      " - 4s - loss: 0.3052 - acc: 0.9140 - val_loss: 2.5017 - val_acc: 0.4286\n",
      "Epoch 474/1000\n",
      " - 4s - loss: 0.3365 - acc: 0.8906 - val_loss: 2.8621 - val_acc: 0.4286\n",
      "Epoch 475/1000\n",
      " - 4s - loss: 0.4318 - acc: 0.8359 - val_loss: 2.2955 - val_acc: 0.5000\n",
      "Epoch 476/1000\n",
      " - 4s - loss: 0.2565 - acc: 0.9062 - val_loss: 2.6575 - val_acc: 0.4643\n",
      "Epoch 477/1000\n",
      " - 4s - loss: 0.4847 - acc: 0.8830 - val_loss: 2.8682 - val_acc: 0.4286\n",
      "Epoch 478/1000\n",
      " - 4s - loss: 0.5318 - acc: 0.8202 - val_loss: 2.6866 - val_acc: 0.5000\n",
      "Epoch 479/1000\n",
      " - 4s - loss: 0.4745 - acc: 0.8906 - val_loss: 2.6567 - val_acc: 0.5714\n",
      "Epoch 480/1000\n",
      " - 4s - loss: 0.2358 - acc: 0.9140 - val_loss: 2.9292 - val_acc: 0.5357\n",
      "Epoch 481/1000\n",
      " - 4s - loss: 0.2739 - acc: 0.8984 - val_loss: 2.4750 - val_acc: 0.5000\n",
      "Epoch 482/1000\n",
      " - 4s - loss: 0.4003 - acc: 0.8908 - val_loss: 2.2506 - val_acc: 0.5000\n",
      "Epoch 483/1000\n",
      " - 4s - loss: 0.4328 - acc: 0.8360 - val_loss: 2.8392 - val_acc: 0.4286\n",
      "Epoch 484/1000\n",
      " - 4s - loss: 0.3036 - acc: 0.8906 - val_loss: 2.2950 - val_acc: 0.5000\n",
      "Epoch 485/1000\n",
      " - 4s - loss: 0.2983 - acc: 0.8828 - val_loss: 2.4625 - val_acc: 0.4643\n",
      "Epoch 486/1000\n",
      " - 4s - loss: 0.4753 - acc: 0.8437 - val_loss: 3.1687 - val_acc: 0.4643\n",
      "Epoch 487/1000\n",
      " - 4s - loss: 0.3398 - acc: 0.8830 - val_loss: 2.6950 - val_acc: 0.5357\n",
      "Epoch 488/1000\n",
      " - 4s - loss: 0.2950 - acc: 0.8984 - val_loss: 1.7311 - val_acc: 0.6429\n",
      "Epoch 489/1000\n",
      " - 4s - loss: 0.3418 - acc: 0.9140 - val_loss: 2.5271 - val_acc: 0.5714\n",
      "Epoch 490/1000\n",
      " - 4s - loss: 0.3010 - acc: 0.8906 - val_loss: 2.4413 - val_acc: 0.5714\n",
      "Epoch 491/1000\n",
      " - 4s - loss: 0.2939 - acc: 0.8752 - val_loss: 2.3683 - val_acc: 0.5714\n",
      "Epoch 492/1000\n",
      " - 4s - loss: 0.2757 - acc: 0.9297 - val_loss: 2.1733 - val_acc: 0.5357\n",
      "Epoch 493/1000\n",
      " - 4s - loss: 0.3282 - acc: 0.8986 - val_loss: 2.9937 - val_acc: 0.5000\n",
      "Epoch 494/1000\n",
      " - 4s - loss: 0.3059 - acc: 0.8749 - val_loss: 3.0081 - val_acc: 0.6429\n",
      "Epoch 495/1000\n",
      " - 4s - loss: 0.2295 - acc: 0.9140 - val_loss: 3.7827 - val_acc: 0.3571\n",
      "Epoch 496/1000\n",
      " - 3s - loss: 0.4568 - acc: 0.8205 - val_loss: 2.9161 - val_acc: 0.3929\n",
      "Epoch 497/1000\n",
      " - 4s - loss: 0.3137 - acc: 0.9140 - val_loss: 2.7068 - val_acc: 0.5357\n",
      "Epoch 498/1000\n",
      " - 4s - loss: 0.3354 - acc: 0.8984 - val_loss: 3.1266 - val_acc: 0.5000\n",
      "Epoch 499/1000\n",
      " - 4s - loss: 0.3233 - acc: 0.8750 - val_loss: 2.4032 - val_acc: 0.4286\n",
      "Epoch 500/1000\n",
      " - 4s - loss: 0.3518 - acc: 0.8829 - val_loss: 3.0824 - val_acc: 0.5000\n",
      "Epoch 501/1000\n",
      " - 4s - loss: 0.3243 - acc: 0.8828 - val_loss: 2.1119 - val_acc: 0.5357\n",
      "Epoch 502/1000\n",
      " - 4s - loss: 0.5224 - acc: 0.8046 - val_loss: 2.7369 - val_acc: 0.4643\n",
      "Epoch 503/1000\n",
      " - 4s - loss: 0.4237 - acc: 0.8984 - val_loss: 2.3966 - val_acc: 0.4643\n",
      "Epoch 504/1000\n",
      " - 4s - loss: 0.3159 - acc: 0.9218 - val_loss: 1.7174 - val_acc: 0.5000\n",
      "Epoch 505/1000\n",
      " - 4s - loss: 0.2308 - acc: 0.9141 - val_loss: 2.2028 - val_acc: 0.4643\n",
      "Epoch 506/1000\n",
      " - 4s - loss: 0.4064 - acc: 0.8592 - val_loss: 3.0181 - val_acc: 0.4643\n",
      "Epoch 507/1000\n",
      " - 4s - loss: 0.4143 - acc: 0.8749 - val_loss: 2.5148 - val_acc: 0.5357\n",
      "Epoch 508/1000\n",
      " - 4s - loss: 0.3317 - acc: 0.8984 - val_loss: 2.9885 - val_acc: 0.3929\n",
      "Epoch 509/1000\n",
      " - 4s - loss: 0.3316 - acc: 0.8905 - val_loss: 3.2044 - val_acc: 0.4643\n",
      "Epoch 510/1000\n",
      " - 4s - loss: 0.3608 - acc: 0.8828 - val_loss: 2.6692 - val_acc: 0.5714\n",
      "Epoch 511/1000\n",
      " - 4s - loss: 0.1772 - acc: 0.9299 - val_loss: 2.9971 - val_acc: 0.5000\n",
      "Epoch 512/1000\n",
      " - 4s - loss: 0.5229 - acc: 0.8829 - val_loss: 1.8970 - val_acc: 0.5714\n",
      "Epoch 513/1000\n",
      " - 4s - loss: 0.3635 - acc: 0.8594 - val_loss: 2.6484 - val_acc: 0.5714\n",
      "Epoch 514/1000\n",
      " - 5s - loss: 0.3147 - acc: 0.8751 - val_loss: 3.1693 - val_acc: 0.4643\n",
      "Epoch 515/1000\n",
      " - 4s - loss: 0.4322 - acc: 0.8749 - val_loss: 2.8905 - val_acc: 0.4643\n",
      "Epoch 516/1000\n",
      " - 4s - loss: 0.2706 - acc: 0.8906 - val_loss: 3.3639 - val_acc: 0.3571\n",
      "Epoch 517/1000\n",
      " - 4s - loss: 0.3996 - acc: 0.8671 - val_loss: 2.5750 - val_acc: 0.4643\n",
      "Epoch 518/1000\n",
      " - 4s - loss: 0.3024 - acc: 0.8516 - val_loss: 2.0453 - val_acc: 0.6071\n",
      "Epoch 519/1000\n",
      " - 4s - loss: 0.3976 - acc: 0.8827 - val_loss: 2.8885 - val_acc: 0.5000\n",
      "Epoch 520/1000\n",
      " - 4s - loss: 0.4801 - acc: 0.8672 - val_loss: 2.4394 - val_acc: 0.4643\n",
      "Epoch 521/1000\n",
      " - 4s - loss: 0.2929 - acc: 0.9375 - val_loss: 3.2138 - val_acc: 0.5357\n",
      "Epoch 522/1000\n",
      " - 4s - loss: 0.3171 - acc: 0.8906 - val_loss: 2.7420 - val_acc: 0.4286\n",
      "Epoch 523/1000\n",
      " - 4s - loss: 0.2695 - acc: 0.9140 - val_loss: 3.1767 - val_acc: 0.4286\n",
      "Epoch 524/1000\n",
      " - 4s - loss: 0.5238 - acc: 0.8281 - val_loss: 2.6389 - val_acc: 0.4643\n",
      "Epoch 525/1000\n",
      " - 4s - loss: 0.3641 - acc: 0.8906 - val_loss: 3.2223 - val_acc: 0.3929\n",
      "Epoch 526/1000\n",
      " - 4s - loss: 0.3113 - acc: 0.8593 - val_loss: 2.4609 - val_acc: 0.3929\n",
      "Epoch 527/1000\n",
      " - 4s - loss: 0.4260 - acc: 0.8671 - val_loss: 2.6833 - val_acc: 0.5357\n",
      "Epoch 528/1000\n",
      " - 4s - loss: 0.5382 - acc: 0.8749 - val_loss: 2.5773 - val_acc: 0.4643\n",
      "Epoch 529/1000\n",
      " - 4s - loss: 0.4448 - acc: 0.8984 - val_loss: 2.9455 - val_acc: 0.3929\n",
      "Epoch 530/1000\n",
      " - 4s - loss: 0.4278 - acc: 0.9062 - val_loss: 3.2036 - val_acc: 0.3929\n",
      "Epoch 531/1000\n",
      " - 4s - loss: 0.3654 - acc: 0.8671 - val_loss: 2.5681 - val_acc: 0.4286\n",
      "Epoch 532/1000\n",
      " - 4s - loss: 0.2981 - acc: 0.8906 - val_loss: 2.6321 - val_acc: 0.5357\n",
      "Epoch 533/1000\n",
      " - 4s - loss: 0.3586 - acc: 0.8828 - val_loss: 2.0020 - val_acc: 0.5000\n",
      "Epoch 534/1000\n",
      " - 4s - loss: 0.2399 - acc: 0.9297 - val_loss: 1.6746 - val_acc: 0.5714\n",
      "Epoch 535/1000\n",
      " - 4s - loss: 0.2059 - acc: 0.9218 - val_loss: 3.1108 - val_acc: 0.5000\n",
      "Epoch 536/1000\n",
      " - 4s - loss: 0.2916 - acc: 0.9140 - val_loss: 2.4425 - val_acc: 0.5714\n",
      "Epoch 537/1000\n",
      " - 4s - loss: 0.4249 - acc: 0.8906 - val_loss: 3.0210 - val_acc: 0.4286\n",
      "Epoch 538/1000\n",
      " - 4s - loss: 0.5792 - acc: 0.8284 - val_loss: 3.0266 - val_acc: 0.5000\n",
      "Epoch 539/1000\n",
      " - 4s - loss: 0.3920 - acc: 0.8361 - val_loss: 2.7237 - val_acc: 0.5357\n",
      "Epoch 540/1000\n",
      " - 4s - loss: 0.3220 - acc: 0.8828 - val_loss: 3.5907 - val_acc: 0.4286\n",
      "Epoch 541/1000\n",
      " - 4s - loss: 0.2745 - acc: 0.8827 - val_loss: 2.7609 - val_acc: 0.4643\n",
      "Epoch 542/1000\n",
      " - 4s - loss: 0.3710 - acc: 0.8828 - val_loss: 2.4805 - val_acc: 0.5000\n",
      "Epoch 543/1000\n",
      " - 4s - loss: 0.2531 - acc: 0.9375 - val_loss: 2.5461 - val_acc: 0.5714\n",
      "Epoch 544/1000\n",
      " - 4s - loss: 0.3167 - acc: 0.8671 - val_loss: 2.2950 - val_acc: 0.4643\n",
      "Epoch 545/1000\n",
      " - 4s - loss: 0.3158 - acc: 0.8828 - val_loss: 2.0912 - val_acc: 0.5714\n",
      "Epoch 546/1000\n",
      " - 4s - loss: 0.4054 - acc: 0.8905 - val_loss: 1.9091 - val_acc: 0.5000\n",
      "Epoch 547/1000\n",
      " - 4s - loss: 0.2898 - acc: 0.8830 - val_loss: 3.1412 - val_acc: 0.4286\n",
      "Epoch 548/1000\n",
      " - 4s - loss: 0.3582 - acc: 0.8750 - val_loss: 2.2766 - val_acc: 0.5714\n",
      "Epoch 549/1000\n",
      " - 4s - loss: 0.4306 - acc: 0.8671 - val_loss: 2.2707 - val_acc: 0.5357\n",
      "Epoch 550/1000\n",
      " - 4s - loss: 0.2985 - acc: 0.9062 - val_loss: 2.4677 - val_acc: 0.4643\n",
      "Epoch 551/1000\n",
      " - 4s - loss: 0.2223 - acc: 0.8984 - val_loss: 2.5507 - val_acc: 0.5357\n",
      "Epoch 552/1000\n",
      " - 4s - loss: 0.2815 - acc: 0.9062 - val_loss: 2.5717 - val_acc: 0.5357\n",
      "Epoch 553/1000\n",
      " - 4s - loss: 0.2034 - acc: 0.9297 - val_loss: 2.6473 - val_acc: 0.5714\n",
      "Epoch 554/1000\n",
      " - 4s - loss: 0.3521 - acc: 0.8910 - val_loss: 2.9101 - val_acc: 0.5000\n",
      "Epoch 555/1000\n",
      " - 4s - loss: 0.3545 - acc: 0.8438 - val_loss: 2.5557 - val_acc: 0.4643\n",
      "Epoch 556/1000\n",
      " - 4s - loss: 0.3412 - acc: 0.8906 - val_loss: 2.4317 - val_acc: 0.5714\n",
      "Epoch 557/1000\n",
      " - 4s - loss: 0.2914 - acc: 0.9064 - val_loss: 3.4566 - val_acc: 0.3929\n",
      "Epoch 558/1000\n",
      " - 4s - loss: 0.2387 - acc: 0.9297 - val_loss: 2.3921 - val_acc: 0.5714\n",
      "Epoch 559/1000\n",
      " - 4s - loss: 0.3850 - acc: 0.8828 - val_loss: 2.0440 - val_acc: 0.5714\n",
      "Epoch 560/1000\n",
      " - 4s - loss: 0.1942 - acc: 0.9297 - val_loss: 3.1202 - val_acc: 0.4643\n",
      "Epoch 561/1000\n",
      " - 4s - loss: 0.4162 - acc: 0.8515 - val_loss: 3.0166 - val_acc: 0.5000\n",
      "Epoch 562/1000\n",
      " - 4s - loss: 0.3919 - acc: 0.8908 - val_loss: 2.2831 - val_acc: 0.3929\n",
      "Epoch 563/1000\n",
      " - 4s - loss: 0.3263 - acc: 0.9062 - val_loss: 2.2270 - val_acc: 0.4286\n",
      "Epoch 564/1000\n",
      " - 4s - loss: 0.2368 - acc: 0.9062 - val_loss: 2.9702 - val_acc: 0.5000\n",
      "Epoch 565/1000\n",
      " - 4s - loss: 0.2880 - acc: 0.8828 - val_loss: 2.8245 - val_acc: 0.4643\n",
      "Epoch 566/1000\n",
      " - 4s - loss: 0.2636 - acc: 0.8984 - val_loss: 3.2620 - val_acc: 0.4286\n",
      "Epoch 567/1000\n",
      " - 4s - loss: 0.2791 - acc: 0.9218 - val_loss: 2.7574 - val_acc: 0.6071\n",
      "Epoch 568/1000\n",
      " - 4s - loss: 0.4001 - acc: 0.8906 - val_loss: 3.0318 - val_acc: 0.4643\n",
      "Epoch 569/1000\n",
      " - 4s - loss: 0.4166 - acc: 0.8750 - val_loss: 3.0185 - val_acc: 0.4643\n",
      "Epoch 570/1000\n",
      " - 4s - loss: 0.4562 - acc: 0.8908 - val_loss: 3.1574 - val_acc: 0.5714\n",
      "Epoch 571/1000\n",
      " - 4s - loss: 0.3059 - acc: 0.9062 - val_loss: 2.8008 - val_acc: 0.5357\n",
      "Epoch 572/1000\n",
      " - 4s - loss: 0.4488 - acc: 0.8283 - val_loss: 3.5229 - val_acc: 0.4643\n",
      "Epoch 573/1000\n",
      " - 4s - loss: 0.3911 - acc: 0.8515 - val_loss: 2.9030 - val_acc: 0.5357\n",
      "Epoch 574/1000\n",
      " - 4s - loss: 0.2217 - acc: 0.9067 - val_loss: 3.5439 - val_acc: 0.5000\n",
      "Epoch 575/1000\n",
      " - 4s - loss: 0.3391 - acc: 0.8986 - val_loss: 2.7116 - val_acc: 0.5000\n",
      "Epoch 576/1000\n",
      " - 4s - loss: 0.2657 - acc: 0.8984 - val_loss: 2.7915 - val_acc: 0.5357\n",
      "Epoch 577/1000\n",
      " - 4s - loss: 0.2203 - acc: 0.9064 - val_loss: 2.5818 - val_acc: 0.5357\n",
      "Epoch 578/1000\n",
      " - 4s - loss: 0.5106 - acc: 0.8752 - val_loss: 3.2794 - val_acc: 0.6071\n",
      "Epoch 579/1000\n",
      " - 4s - loss: 0.2494 - acc: 0.9141 - val_loss: 3.0704 - val_acc: 0.5000\n",
      "Epoch 580/1000\n",
      " - 5s - loss: 0.2922 - acc: 0.8984 - val_loss: 2.7949 - val_acc: 0.5714\n",
      "Epoch 581/1000\n",
      " - 5s - loss: 0.3001 - acc: 0.8828 - val_loss: 3.5317 - val_acc: 0.4286\n",
      "Epoch 582/1000\n",
      " - 4s - loss: 0.1590 - acc: 0.9455 - val_loss: 2.5793 - val_acc: 0.5714\n",
      "Epoch 583/1000\n",
      " - 4s - loss: 0.2553 - acc: 0.9219 - val_loss: 3.4941 - val_acc: 0.3571\n",
      "Epoch 584/1000\n",
      " - 4s - loss: 0.3013 - acc: 0.8828 - val_loss: 2.5314 - val_acc: 0.5357\n",
      "Epoch 585/1000\n",
      " - 4s - loss: 0.1987 - acc: 0.9375 - val_loss: 3.4434 - val_acc: 0.5000\n",
      "Epoch 586/1000\n",
      " - 4s - loss: 0.3205 - acc: 0.8906 - val_loss: 2.3696 - val_acc: 0.5357\n",
      "Epoch 587/1000\n",
      " - 4s - loss: 0.1412 - acc: 0.9609 - val_loss: 2.6748 - val_acc: 0.5000\n",
      "Epoch 588/1000\n",
      " - 4s - loss: 0.2136 - acc: 0.9297 - val_loss: 3.1356 - val_acc: 0.4643\n",
      "Epoch 589/1000\n",
      " - 4s - loss: 0.4490 - acc: 0.8595 - val_loss: 2.6905 - val_acc: 0.5000\n",
      "Epoch 590/1000\n",
      " - 4s - loss: 0.2907 - acc: 0.9218 - val_loss: 2.8749 - val_acc: 0.5714\n",
      "Epoch 591/1000\n",
      " - 4s - loss: 0.2960 - acc: 0.9062 - val_loss: 2.9514 - val_acc: 0.3929\n",
      "Epoch 592/1000\n",
      " - 4s - loss: 0.2300 - acc: 0.9062 - val_loss: 2.5912 - val_acc: 0.5714\n",
      "Epoch 593/1000\n",
      " - 4s - loss: 0.3549 - acc: 0.8749 - val_loss: 2.4737 - val_acc: 0.5714\n",
      "Epoch 594/1000\n",
      " - 4s - loss: 0.4657 - acc: 0.8984 - val_loss: 2.7406 - val_acc: 0.4286\n",
      "Epoch 595/1000\n",
      " - 4s - loss: 0.2829 - acc: 0.8906 - val_loss: 2.5962 - val_acc: 0.4286\n",
      "Epoch 596/1000\n",
      " - 4s - loss: 0.2926 - acc: 0.8986 - val_loss: 3.4241 - val_acc: 0.4286\n",
      "Epoch 597/1000\n",
      " - 4s - loss: 0.3781 - acc: 0.8593 - val_loss: 3.0043 - val_acc: 0.4643\n",
      "Epoch 598/1000\n",
      " - 4s - loss: 0.2825 - acc: 0.8908 - val_loss: 2.9761 - val_acc: 0.5357\n",
      "Epoch 599/1000\n",
      " - 4s - loss: 0.3709 - acc: 0.8984 - val_loss: 2.4103 - val_acc: 0.5000\n",
      "Epoch 600/1000\n",
      " - 4s - loss: 0.3166 - acc: 0.8906 - val_loss: 2.8032 - val_acc: 0.5357\n",
      "Epoch 601/1000\n",
      " - 4s - loss: 0.3009 - acc: 0.8984 - val_loss: 2.3485 - val_acc: 0.6071\n",
      "Epoch 602/1000\n",
      " - 4s - loss: 0.2622 - acc: 0.9297 - val_loss: 2.3971 - val_acc: 0.5714\n",
      "Epoch 603/1000\n",
      " - 4s - loss: 0.2910 - acc: 0.9143 - val_loss: 3.1900 - val_acc: 0.4643\n",
      "Epoch 604/1000\n",
      " - 4s - loss: 0.2784 - acc: 0.8908 - val_loss: 2.9127 - val_acc: 0.5000\n",
      "Epoch 605/1000\n",
      " - 4s - loss: 0.1761 - acc: 0.9297 - val_loss: 2.4917 - val_acc: 0.5000\n",
      "Epoch 606/1000\n",
      " - 4s - loss: 0.2941 - acc: 0.8905 - val_loss: 3.1499 - val_acc: 0.4286\n",
      "Epoch 607/1000\n",
      " - 4s - loss: 0.3242 - acc: 0.8984 - val_loss: 2.2884 - val_acc: 0.5357\n",
      "Epoch 608/1000\n",
      " - 4s - loss: 0.2745 - acc: 0.9377 - val_loss: 1.8722 - val_acc: 0.5000\n",
      "Epoch 609/1000\n",
      " - 4s - loss: 0.3600 - acc: 0.8752 - val_loss: 2.4466 - val_acc: 0.5000\n",
      "Epoch 610/1000\n",
      " - 4s - loss: 0.2794 - acc: 0.8906 - val_loss: 2.2185 - val_acc: 0.6429\n",
      "Epoch 611/1000\n",
      " - 4s - loss: 0.3706 - acc: 0.8749 - val_loss: 2.7610 - val_acc: 0.4286\n",
      "Epoch 612/1000\n",
      " - 4s - loss: 0.2742 - acc: 0.8986 - val_loss: 2.9550 - val_acc: 0.5357\n",
      "Epoch 613/1000\n",
      " - 4s - loss: 0.1653 - acc: 0.9453 - val_loss: 2.9407 - val_acc: 0.4643\n",
      "Epoch 614/1000\n",
      " - 4s - loss: 0.1669 - acc: 0.9453 - val_loss: 2.5875 - val_acc: 0.4286\n",
      "Epoch 615/1000\n",
      " - 4s - loss: 0.2375 - acc: 0.9062 - val_loss: 2.8584 - val_acc: 0.5000\n",
      "Epoch 616/1000\n",
      " - 4s - loss: 0.1970 - acc: 0.9221 - val_loss: 3.4558 - val_acc: 0.5000\n",
      "Epoch 617/1000\n",
      " - 4s - loss: 0.2243 - acc: 0.9297 - val_loss: 3.0370 - val_acc: 0.4286\n",
      "Epoch 618/1000\n",
      " - 4s - loss: 0.3338 - acc: 0.9062 - val_loss: 2.3752 - val_acc: 0.5000\n",
      "Epoch 619/1000\n",
      " - 4s - loss: 0.1243 - acc: 0.9531 - val_loss: 2.4141 - val_acc: 0.5000\n",
      "Epoch 620/1000\n",
      " - 4s - loss: 0.3266 - acc: 0.8906 - val_loss: 2.4744 - val_acc: 0.5714\n",
      "Epoch 621/1000\n",
      " - 4s - loss: 0.1828 - acc: 0.9218 - val_loss: 2.7283 - val_acc: 0.5714\n",
      "Epoch 622/1000\n",
      " - 4s - loss: 0.3904 - acc: 0.9062 - val_loss: 3.3291 - val_acc: 0.5000\n",
      "Epoch 623/1000\n",
      " - 4s - loss: 0.3610 - acc: 0.8827 - val_loss: 2.6420 - val_acc: 0.5000\n",
      "Epoch 624/1000\n",
      " - 4s - loss: 0.1441 - acc: 0.9453 - val_loss: 2.4036 - val_acc: 0.5000\n",
      "Epoch 625/1000\n",
      " - 4s - loss: 0.3036 - acc: 0.9140 - val_loss: 3.1594 - val_acc: 0.5357\n",
      "Epoch 626/1000\n",
      " - 4s - loss: 0.3036 - acc: 0.8986 - val_loss: 2.8961 - val_acc: 0.4643\n",
      "Epoch 627/1000\n",
      " - 4s - loss: 0.1850 - acc: 0.9453 - val_loss: 2.6049 - val_acc: 0.5357\n",
      "Epoch 628/1000\n",
      " - 4s - loss: 0.3155 - acc: 0.9140 - val_loss: 2.4294 - val_acc: 0.5000\n",
      "Epoch 629/1000\n",
      " - 4s - loss: 0.2491 - acc: 0.8828 - val_loss: 1.8440 - val_acc: 0.5714\n",
      "Epoch 630/1000\n",
      " - 4s - loss: 0.2242 - acc: 0.9297 - val_loss: 2.5178 - val_acc: 0.4286\n",
      "Epoch 631/1000\n",
      " - 4s - loss: 0.2714 - acc: 0.9140 - val_loss: 2.8253 - val_acc: 0.3929\n",
      "Epoch 632/1000\n",
      " - 4s - loss: 0.2412 - acc: 0.9375 - val_loss: 2.6440 - val_acc: 0.5000\n",
      "Epoch 633/1000\n",
      " - 4s - loss: 0.2192 - acc: 0.9062 - val_loss: 2.5053 - val_acc: 0.5714\n",
      "Epoch 634/1000\n",
      " - 4s - loss: 0.3177 - acc: 0.8984 - val_loss: 2.3341 - val_acc: 0.5000\n",
      "Epoch 635/1000\n",
      " - 4s - loss: 0.2435 - acc: 0.9297 - val_loss: 3.7202 - val_acc: 0.3571\n",
      "Epoch 636/1000\n",
      " - 4s - loss: 0.3572 - acc: 0.8517 - val_loss: 3.1532 - val_acc: 0.4286\n",
      "Epoch 637/1000\n",
      " - 4s - loss: 0.3281 - acc: 0.9140 - val_loss: 3.0312 - val_acc: 0.4286\n",
      "Epoch 638/1000\n",
      " - 4s - loss: 0.2058 - acc: 0.9297 - val_loss: 3.5459 - val_acc: 0.4643\n",
      "Epoch 639/1000\n",
      " - 4s - loss: 0.4839 - acc: 0.8984 - val_loss: 2.6420 - val_acc: 0.5000\n",
      "Epoch 640/1000\n",
      " - 4s - loss: 0.2078 - acc: 0.9140 - val_loss: 2.8250 - val_acc: 0.4286\n",
      "Epoch 641/1000\n",
      " - 4s - loss: 0.1755 - acc: 0.9375 - val_loss: 2.8782 - val_acc: 0.4643\n",
      "Epoch 642/1000\n",
      " - 4s - loss: 0.3467 - acc: 0.9064 - val_loss: 2.8637 - val_acc: 0.5000\n",
      "Epoch 643/1000\n",
      " - 4s - loss: 0.3348 - acc: 0.9140 - val_loss: 3.4573 - val_acc: 0.3571\n",
      "Epoch 644/1000\n",
      " - 4s - loss: 0.3334 - acc: 0.8984 - val_loss: 2.4770 - val_acc: 0.5357\n",
      "Epoch 645/1000\n",
      " - 4s - loss: 0.2529 - acc: 0.9062 - val_loss: 3.8227 - val_acc: 0.3929\n",
      "Epoch 646/1000\n",
      " - 4s - loss: 0.4444 - acc: 0.8672 - val_loss: 2.3435 - val_acc: 0.6071\n",
      "Epoch 647/1000\n",
      " - 4s - loss: 0.2296 - acc: 0.9140 - val_loss: 2.6714 - val_acc: 0.5357\n",
      "Epoch 648/1000\n",
      " - 4s - loss: 0.3748 - acc: 0.8749 - val_loss: 2.3738 - val_acc: 0.6429\n",
      "Epoch 649/1000\n",
      " - 4s - loss: 0.3606 - acc: 0.8749 - val_loss: 2.7928 - val_acc: 0.4643\n",
      "Epoch 650/1000\n",
      " - 4s - loss: 0.1947 - acc: 0.9218 - val_loss: 3.2363 - val_acc: 0.3929\n",
      "Epoch 651/1000\n",
      " - 4s - loss: 0.2842 - acc: 0.9062 - val_loss: 3.2568 - val_acc: 0.5714\n",
      "Epoch 652/1000\n",
      " - 4s - loss: 0.3176 - acc: 0.9140 - val_loss: 3.2043 - val_acc: 0.5000\n",
      "Epoch 653/1000\n",
      " - 4s - loss: 0.2825 - acc: 0.9218 - val_loss: 2.8713 - val_acc: 0.5000\n",
      "Epoch 654/1000\n",
      " - 4s - loss: 0.3131 - acc: 0.9064 - val_loss: 3.1356 - val_acc: 0.4286\n",
      "Epoch 655/1000\n",
      " - 4s - loss: 0.4043 - acc: 0.8749 - val_loss: 3.0735 - val_acc: 0.5357\n",
      "Epoch 656/1000\n",
      " - 4s - loss: 0.2948 - acc: 0.8749 - val_loss: 3.3377 - val_acc: 0.5000\n",
      "Epoch 657/1000\n",
      " - 4s - loss: 0.2420 - acc: 0.9218 - val_loss: 2.2819 - val_acc: 0.5714\n",
      "Epoch 658/1000\n",
      " - 4s - loss: 0.2267 - acc: 0.8828 - val_loss: 2.4663 - val_acc: 0.5357\n",
      "Epoch 659/1000\n",
      " - 4s - loss: 0.3737 - acc: 0.8832 - val_loss: 2.7641 - val_acc: 0.5000\n",
      "Epoch 660/1000\n",
      " - 4s - loss: 0.3855 - acc: 0.8749 - val_loss: 2.9897 - val_acc: 0.4643\n",
      "Epoch 661/1000\n",
      " - 4s - loss: 0.3020 - acc: 0.9219 - val_loss: 2.8056 - val_acc: 0.5000\n",
      "Epoch 662/1000\n",
      " - 4s - loss: 0.2229 - acc: 0.9140 - val_loss: 2.5941 - val_acc: 0.5000\n",
      "Epoch 663/1000\n",
      " - 4s - loss: 0.4197 - acc: 0.8749 - val_loss: 2.3961 - val_acc: 0.5714\n",
      "Epoch 664/1000\n",
      " - 4s - loss: 0.2340 - acc: 0.9140 - val_loss: 2.9760 - val_acc: 0.4643\n",
      "Epoch 665/1000\n",
      " - 4s - loss: 0.2863 - acc: 0.8984 - val_loss: 2.9954 - val_acc: 0.4643\n",
      "Epoch 666/1000\n",
      " - 4s - loss: 0.3633 - acc: 0.8984 - val_loss: 2.9227 - val_acc: 0.5714\n",
      "Epoch 667/1000\n",
      " - 4s - loss: 0.2100 - acc: 0.9143 - val_loss: 2.5859 - val_acc: 0.5000\n",
      "Epoch 668/1000\n",
      " - 4s - loss: 0.2297 - acc: 0.9141 - val_loss: 3.4870 - val_acc: 0.4643\n",
      "Epoch 669/1000\n",
      " - 4s - loss: 0.1746 - acc: 0.9375 - val_loss: 3.1142 - val_acc: 0.4286\n",
      "Epoch 670/1000\n",
      " - 4s - loss: 0.2341 - acc: 0.9297 - val_loss: 2.7337 - val_acc: 0.5714\n",
      "Epoch 671/1000\n",
      " - 4s - loss: 0.3284 - acc: 0.9140 - val_loss: 3.3823 - val_acc: 0.5714\n",
      "Epoch 672/1000\n",
      " - 4s - loss: 0.3090 - acc: 0.8984 - val_loss: 2.9522 - val_acc: 0.5000\n",
      "Epoch 673/1000\n",
      " - 4s - loss: 0.3607 - acc: 0.8908 - val_loss: 3.2386 - val_acc: 0.5714\n",
      "Epoch 674/1000\n",
      " - 4s - loss: 0.1592 - acc: 0.9531 - val_loss: 3.0516 - val_acc: 0.5000\n",
      "Epoch 675/1000\n",
      " - 4s - loss: 0.2253 - acc: 0.9218 - val_loss: 2.6385 - val_acc: 0.5714\n",
      "Epoch 676/1000\n",
      " - 4s - loss: 0.1878 - acc: 0.9297 - val_loss: 3.2693 - val_acc: 0.4643\n",
      "Epoch 677/1000\n",
      " - 4s - loss: 0.4433 - acc: 0.8672 - val_loss: 3.2100 - val_acc: 0.5357\n",
      "Epoch 678/1000\n",
      " - 4s - loss: 0.3369 - acc: 0.9062 - val_loss: 3.3989 - val_acc: 0.4286\n",
      "Epoch 679/1000\n",
      " - 4s - loss: 0.2936 - acc: 0.8827 - val_loss: 2.9403 - val_acc: 0.5000\n",
      "Epoch 680/1000\n",
      " - 4s - loss: 0.3062 - acc: 0.9375 - val_loss: 2.3951 - val_acc: 0.6071\n",
      "Epoch 681/1000\n",
      " - 4s - loss: 0.3971 - acc: 0.8517 - val_loss: 2.8895 - val_acc: 0.5000\n",
      "Epoch 682/1000\n",
      " - 4s - loss: 0.1976 - acc: 0.9455 - val_loss: 3.2379 - val_acc: 0.4643\n",
      "Epoch 683/1000\n",
      " - 4s - loss: 0.2280 - acc: 0.9297 - val_loss: 2.6328 - val_acc: 0.4643\n",
      "Epoch 684/1000\n",
      " - 4s - loss: 0.1123 - acc: 0.9531 - val_loss: 2.7027 - val_acc: 0.5000\n",
      "Epoch 685/1000\n",
      " - 4s - loss: 0.2585 - acc: 0.9062 - val_loss: 2.2342 - val_acc: 0.4643\n",
      "Epoch 686/1000\n",
      " - 4s - loss: 0.2451 - acc: 0.9218 - val_loss: 3.3334 - val_acc: 0.5000\n",
      "Epoch 687/1000\n",
      " - 4s - loss: 0.4515 - acc: 0.8828 - val_loss: 2.6244 - val_acc: 0.5000\n",
      "Epoch 688/1000\n",
      " - 4s - loss: 0.5202 - acc: 0.8908 - val_loss: 3.4262 - val_acc: 0.3571\n",
      "Epoch 689/1000\n",
      " - 4s - loss: 0.2432 - acc: 0.8988 - val_loss: 3.3391 - val_acc: 0.4643\n",
      "Epoch 690/1000\n",
      " - 4s - loss: 0.3871 - acc: 0.8828 - val_loss: 2.1165 - val_acc: 0.5357\n",
      "Epoch 691/1000\n",
      " - 4s - loss: 0.1641 - acc: 0.9453 - val_loss: 2.5175 - val_acc: 0.4643\n",
      "Epoch 692/1000\n",
      " - 4s - loss: 0.2669 - acc: 0.9141 - val_loss: 2.0355 - val_acc: 0.5000\n",
      "Epoch 693/1000\n",
      " - 4s - loss: 0.2143 - acc: 0.8984 - val_loss: 3.2234 - val_acc: 0.5357\n",
      "Epoch 694/1000\n",
      " - 4s - loss: 0.3436 - acc: 0.8520 - val_loss: 1.8400 - val_acc: 0.5714\n",
      "Epoch 695/1000\n",
      " - 4s - loss: 0.2937 - acc: 0.9140 - val_loss: 2.7806 - val_acc: 0.5000\n",
      "Epoch 696/1000\n",
      " - 4s - loss: 0.1853 - acc: 0.9453 - val_loss: 3.7624 - val_acc: 0.3929\n",
      "Epoch 697/1000\n",
      " - 4s - loss: 0.4386 - acc: 0.8671 - val_loss: 2.4023 - val_acc: 0.5357\n",
      "Epoch 698/1000\n",
      " - 4s - loss: 0.3683 - acc: 0.8984 - val_loss: 2.7503 - val_acc: 0.4643\n",
      "Epoch 699/1000\n",
      " - 4s - loss: 0.1322 - acc: 0.9453 - val_loss: 2.8882 - val_acc: 0.5357\n",
      "Epoch 700/1000\n",
      " - 4s - loss: 0.1376 - acc: 0.9299 - val_loss: 2.8349 - val_acc: 0.5000\n",
      "Epoch 701/1000\n",
      " - 4s - loss: 0.2875 - acc: 0.8986 - val_loss: 2.2487 - val_acc: 0.5357\n",
      "Epoch 702/1000\n",
      " - 4s - loss: 0.2202 - acc: 0.9062 - val_loss: 2.4793 - val_acc: 0.3571\n",
      "Epoch 703/1000\n",
      " - 4s - loss: 0.1799 - acc: 0.9297 - val_loss: 2.3110 - val_acc: 0.5357\n",
      "Epoch 704/1000\n",
      " - 4s - loss: 0.2920 - acc: 0.9064 - val_loss: 2.9432 - val_acc: 0.4286\n",
      "Epoch 705/1000\n",
      " - 4s - loss: 0.3372 - acc: 0.8751 - val_loss: 2.4471 - val_acc: 0.5714\n",
      "Epoch 706/1000\n",
      " - 4s - loss: 0.4571 - acc: 0.8671 - val_loss: 2.7383 - val_acc: 0.4286\n",
      "Epoch 707/1000\n",
      " - 4s - loss: 0.2052 - acc: 0.9375 - val_loss: 3.9005 - val_acc: 0.3929\n",
      "Epoch 708/1000\n",
      " - 4s - loss: 0.4180 - acc: 0.8828 - val_loss: 3.3956 - val_acc: 0.4286\n",
      "Epoch 709/1000\n",
      " - 4s - loss: 0.2481 - acc: 0.9062 - val_loss: 2.1370 - val_acc: 0.6786\n",
      "Epoch 710/1000\n",
      " - 4s - loss: 0.2196 - acc: 0.9062 - val_loss: 2.8524 - val_acc: 0.4286\n",
      "Epoch 711/1000\n",
      " - 4s - loss: 0.2256 - acc: 0.9140 - val_loss: 2.5298 - val_acc: 0.3929\n",
      "Epoch 712/1000\n",
      " - 4s - loss: 0.1624 - acc: 0.9297 - val_loss: 2.7376 - val_acc: 0.5000\n",
      "Epoch 713/1000\n",
      " - 4s - loss: 0.2577 - acc: 0.9221 - val_loss: 2.8249 - val_acc: 0.4286\n",
      "Epoch 714/1000\n",
      " - 4s - loss: 0.2359 - acc: 0.9219 - val_loss: 2.6989 - val_acc: 0.5000\n",
      "Epoch 715/1000\n",
      " - 4s - loss: 0.5024 - acc: 0.8438 - val_loss: 3.1562 - val_acc: 0.5000\n",
      "Epoch 716/1000\n",
      " - 4s - loss: 0.2017 - acc: 0.9375 - val_loss: 3.8479 - val_acc: 0.3571\n",
      "Epoch 717/1000\n",
      " - 4s - loss: 0.1321 - acc: 0.9455 - val_loss: 2.3081 - val_acc: 0.4643\n",
      "Epoch 718/1000\n",
      " - 4s - loss: 0.3010 - acc: 0.8905 - val_loss: 3.0694 - val_acc: 0.4643\n",
      "Epoch 719/1000\n",
      " - 4s - loss: 0.3658 - acc: 0.8984 - val_loss: 2.9275 - val_acc: 0.5357\n",
      "Epoch 720/1000\n",
      " - 4s - loss: 0.5833 - acc: 0.8515 - val_loss: 2.6743 - val_acc: 0.3929\n",
      "Epoch 721/1000\n",
      " - 4s - loss: 0.2798 - acc: 0.9140 - val_loss: 2.7322 - val_acc: 0.5714\n",
      "Epoch 722/1000\n",
      " - 4s - loss: 0.2717 - acc: 0.9297 - val_loss: 3.2445 - val_acc: 0.4286\n",
      "Epoch 723/1000\n",
      " - 4s - loss: 0.2768 - acc: 0.9220 - val_loss: 2.4491 - val_acc: 0.5714\n",
      "Epoch 724/1000\n",
      " - 4s - loss: 0.3248 - acc: 0.9062 - val_loss: 2.6481 - val_acc: 0.5000\n",
      "Epoch 725/1000\n",
      " - 4s - loss: 0.3390 - acc: 0.8906 - val_loss: 2.9974 - val_acc: 0.4643\n",
      "Epoch 726/1000\n",
      " - 4s - loss: 0.1207 - acc: 0.9531 - val_loss: 2.8713 - val_acc: 0.5000\n",
      "Epoch 727/1000\n",
      " - 4s - loss: 0.2542 - acc: 0.9140 - val_loss: 3.1038 - val_acc: 0.5357\n",
      "Epoch 728/1000\n",
      " - 4s - loss: 0.2266 - acc: 0.9143 - val_loss: 2.4780 - val_acc: 0.4643\n",
      "Epoch 729/1000\n",
      " - 4s - loss: 0.1399 - acc: 0.9377 - val_loss: 3.4442 - val_acc: 0.4286\n",
      "Epoch 730/1000\n",
      " - 4s - loss: 0.2181 - acc: 0.9140 - val_loss: 3.1598 - val_acc: 0.5000\n",
      "Epoch 731/1000\n",
      " - 4s - loss: 0.1691 - acc: 0.9375 - val_loss: 2.9605 - val_acc: 0.6071\n",
      "Epoch 732/1000\n",
      " - 4s - loss: 0.3493 - acc: 0.8905 - val_loss: 3.1515 - val_acc: 0.5000\n",
      "Epoch 733/1000\n",
      " - 4s - loss: 0.2504 - acc: 0.9062 - val_loss: 2.1888 - val_acc: 0.5714\n",
      "Epoch 734/1000\n",
      " - 4s - loss: 0.1451 - acc: 0.9299 - val_loss: 2.9574 - val_acc: 0.4643\n",
      "Epoch 735/1000\n",
      " - 4s - loss: 0.2222 - acc: 0.9453 - val_loss: 4.0157 - val_acc: 0.3929\n",
      "Epoch 736/1000\n",
      " - 4s - loss: 0.2405 - acc: 0.9062 - val_loss: 2.9150 - val_acc: 0.5357\n",
      "Epoch 737/1000\n",
      " - 4s - loss: 0.2767 - acc: 0.8986 - val_loss: 3.0203 - val_acc: 0.6071\n",
      "Epoch 738/1000\n",
      " - 4s - loss: 0.3144 - acc: 0.9140 - val_loss: 2.8068 - val_acc: 0.4643\n",
      "Epoch 739/1000\n",
      " - 4s - loss: 0.1797 - acc: 0.9375 - val_loss: 3.0293 - val_acc: 0.4286\n",
      "Epoch 740/1000\n",
      " - 4s - loss: 0.1826 - acc: 0.9143 - val_loss: 2.6097 - val_acc: 0.5357\n",
      "Epoch 741/1000\n",
      " - 4s - loss: 0.4098 - acc: 0.9062 - val_loss: 2.9655 - val_acc: 0.5357\n",
      "Epoch 742/1000\n",
      " - 4s - loss: 0.1929 - acc: 0.9296 - val_loss: 2.4672 - val_acc: 0.5714\n",
      "Epoch 743/1000\n",
      " - 4s - loss: 0.3179 - acc: 0.8674 - val_loss: 3.0536 - val_acc: 0.4643\n",
      "Epoch 744/1000\n",
      " - 4s - loss: 0.3274 - acc: 0.9062 - val_loss: 2.5633 - val_acc: 0.5357\n",
      "Epoch 745/1000\n",
      " - 4s - loss: 0.2439 - acc: 0.9297 - val_loss: 2.7033 - val_acc: 0.5357\n",
      "Epoch 746/1000\n",
      " - 4s - loss: 0.3643 - acc: 0.9218 - val_loss: 2.6127 - val_acc: 0.4643\n",
      "Epoch 747/1000\n",
      " - 4s - loss: 0.2718 - acc: 0.9064 - val_loss: 3.3605 - val_acc: 0.5000\n",
      "Epoch 748/1000\n",
      " - 4s - loss: 0.1446 - acc: 0.9297 - val_loss: 3.4313 - val_acc: 0.4286\n",
      "Epoch 749/1000\n",
      " - 4s - loss: 0.1584 - acc: 0.9375 - val_loss: 2.3765 - val_acc: 0.6071\n",
      "Epoch 750/1000\n",
      " - 4s - loss: 0.3229 - acc: 0.9062 - val_loss: 2.4743 - val_acc: 0.5714\n",
      "Epoch 751/1000\n",
      " - 4s - loss: 0.2202 - acc: 0.9145 - val_loss: 2.4344 - val_acc: 0.5000\n",
      "Epoch 752/1000\n",
      " - 4s - loss: 0.2839 - acc: 0.8750 - val_loss: 3.1749 - val_acc: 0.3929\n",
      "Epoch 753/1000\n",
      " - 4s - loss: 0.1512 - acc: 0.9765 - val_loss: 2.7425 - val_acc: 0.4643\n",
      "Epoch 754/1000\n",
      " - 4s - loss: 0.1270 - acc: 0.9453 - val_loss: 3.1983 - val_acc: 0.3571\n",
      "Epoch 755/1000\n",
      " - 4s - loss: 0.2613 - acc: 0.9140 - val_loss: 2.7681 - val_acc: 0.4643\n",
      "Epoch 756/1000\n",
      " - 4s - loss: 0.2467 - acc: 0.9297 - val_loss: 3.3721 - val_acc: 0.4643\n",
      "Epoch 757/1000\n",
      " - 4s - loss: 0.2625 - acc: 0.8828 - val_loss: 2.7602 - val_acc: 0.4643\n",
      "Epoch 758/1000\n",
      " - 4s - loss: 0.1738 - acc: 0.9531 - val_loss: 2.8034 - val_acc: 0.5357\n",
      "Epoch 759/1000\n",
      " - 4s - loss: 0.1416 - acc: 0.9375 - val_loss: 2.7493 - val_acc: 0.5000\n",
      "Epoch 760/1000\n",
      " - 4s - loss: 0.1965 - acc: 0.9218 - val_loss: 2.7097 - val_acc: 0.4643\n",
      "Epoch 761/1000\n",
      " - 4s - loss: 0.2461 - acc: 0.9218 - val_loss: 3.0371 - val_acc: 0.4286\n",
      "Epoch 762/1000\n",
      " - 4s - loss: 0.3144 - acc: 0.8986 - val_loss: 3.3645 - val_acc: 0.5000\n",
      "Epoch 763/1000\n",
      " - 4s - loss: 0.4232 - acc: 0.8828 - val_loss: 2.5342 - val_acc: 0.4643\n",
      "Epoch 764/1000\n",
      " - 4s - loss: 0.2612 - acc: 0.9140 - val_loss: 3.3474 - val_acc: 0.4286\n",
      "Epoch 765/1000\n",
      " - 4s - loss: 0.2533 - acc: 0.9140 - val_loss: 2.6569 - val_acc: 0.5000\n",
      "Epoch 766/1000\n",
      " - 4s - loss: 0.3251 - acc: 0.9062 - val_loss: 2.6065 - val_acc: 0.5000\n",
      "Epoch 767/1000\n",
      " - 4s - loss: 0.2887 - acc: 0.9296 - val_loss: 2.1667 - val_acc: 0.5357\n",
      "Epoch 768/1000\n",
      " - 4s - loss: 0.2463 - acc: 0.9218 - val_loss: 3.1784 - val_acc: 0.5000\n",
      "Epoch 769/1000\n",
      " - 4s - loss: 0.1796 - acc: 0.9453 - val_loss: 2.9248 - val_acc: 0.5714\n",
      "Epoch 770/1000\n",
      " - 4s - loss: 0.3390 - acc: 0.8984 - val_loss: 3.3162 - val_acc: 0.4643\n",
      "Epoch 771/1000\n",
      " - 4s - loss: 0.2109 - acc: 0.9377 - val_loss: 3.5985 - val_acc: 0.3929\n",
      "Epoch 772/1000\n",
      " - 4s - loss: 0.1904 - acc: 0.9141 - val_loss: 3.2829 - val_acc: 0.4643\n",
      "Epoch 773/1000\n",
      " - 4s - loss: 0.1716 - acc: 0.9299 - val_loss: 3.1633 - val_acc: 0.4286\n",
      "Epoch 774/1000\n",
      " - 4s - loss: 0.2073 - acc: 0.9297 - val_loss: 3.2821 - val_acc: 0.3571\n",
      "Epoch 775/1000\n",
      " - 4s - loss: 0.3527 - acc: 0.8671 - val_loss: 2.9075 - val_acc: 0.5714\n",
      "Epoch 776/1000\n",
      " - 4s - loss: 0.3279 - acc: 0.8908 - val_loss: 2.8939 - val_acc: 0.4286\n",
      "Epoch 777/1000\n",
      " - 4s - loss: 0.4577 - acc: 0.8749 - val_loss: 2.6828 - val_acc: 0.5000\n",
      "Epoch 778/1000\n",
      " - 4s - loss: 0.3386 - acc: 0.8984 - val_loss: 2.2415 - val_acc: 0.4643\n",
      "Epoch 779/1000\n",
      " - 4s - loss: 0.0921 - acc: 0.9609 - val_loss: 3.4920 - val_acc: 0.4643\n",
      "Epoch 780/1000\n",
      " - 4s - loss: 0.3175 - acc: 0.8517 - val_loss: 3.1153 - val_acc: 0.5000\n",
      "Epoch 781/1000\n",
      " - 4s - loss: 0.2731 - acc: 0.9219 - val_loss: 2.1130 - val_acc: 0.5714\n",
      "Epoch 782/1000\n",
      " - 4s - loss: 0.2756 - acc: 0.9218 - val_loss: 3.0941 - val_acc: 0.4286\n",
      "Epoch 783/1000\n",
      " - 4s - loss: 0.2611 - acc: 0.9064 - val_loss: 3.5871 - val_acc: 0.5357\n",
      "Epoch 784/1000\n",
      " - 4s - loss: 0.4120 - acc: 0.8828 - val_loss: 2.5566 - val_acc: 0.5357\n",
      "Epoch 785/1000\n",
      " - 4s - loss: 0.2891 - acc: 0.8828 - val_loss: 2.6831 - val_acc: 0.5357\n",
      "Epoch 786/1000\n",
      " - 4s - loss: 0.0647 - acc: 0.9612 - val_loss: 3.2643 - val_acc: 0.4643\n",
      "Epoch 787/1000\n",
      " - 4s - loss: 0.2670 - acc: 0.9531 - val_loss: 2.3380 - val_acc: 0.4286\n",
      "Epoch 788/1000\n",
      " - 4s - loss: 0.2916 - acc: 0.9143 - val_loss: 3.3170 - val_acc: 0.5000\n",
      "Epoch 789/1000\n",
      " - 4s - loss: 0.2274 - acc: 0.9375 - val_loss: 2.0137 - val_acc: 0.5357\n",
      "Epoch 790/1000\n",
      " - 4s - loss: 0.2776 - acc: 0.8908 - val_loss: 3.5672 - val_acc: 0.5357\n",
      "Epoch 791/1000\n",
      " - 4s - loss: 0.3035 - acc: 0.9297 - val_loss: 2.8744 - val_acc: 0.4643\n",
      "Epoch 792/1000\n",
      " - 4s - loss: 0.1619 - acc: 0.9218 - val_loss: 2.8922 - val_acc: 0.5714\n",
      "Epoch 793/1000\n",
      " - 4s - loss: 0.1315 - acc: 0.9375 - val_loss: 3.1790 - val_acc: 0.4643\n",
      "Epoch 794/1000\n",
      " - 4s - loss: 0.0998 - acc: 0.9612 - val_loss: 2.6398 - val_acc: 0.5000\n",
      "Epoch 795/1000\n",
      " - 4s - loss: 0.1618 - acc: 0.9375 - val_loss: 3.2539 - val_acc: 0.4286\n",
      "Epoch 796/1000\n",
      " - 4s - loss: 0.2518 - acc: 0.9297 - val_loss: 3.0161 - val_acc: 0.5714\n",
      "Epoch 797/1000\n",
      " - 4s - loss: 0.3097 - acc: 0.8983 - val_loss: 1.9150 - val_acc: 0.6071\n",
      "Epoch 798/1000\n",
      " - 4s - loss: 0.2666 - acc: 0.9140 - val_loss: 3.1986 - val_acc: 0.5357\n",
      "Epoch 799/1000\n",
      " - 4s - loss: 0.2948 - acc: 0.9062 - val_loss: 2.4080 - val_acc: 0.5714\n",
      "Epoch 800/1000\n",
      " - 4s - loss: 0.1703 - acc: 0.9297 - val_loss: 3.7103 - val_acc: 0.4286\n",
      "Epoch 801/1000\n",
      " - 4s - loss: 0.3267 - acc: 0.9064 - val_loss: 3.0865 - val_acc: 0.3929\n",
      "Epoch 802/1000\n",
      " - 4s - loss: 0.2469 - acc: 0.8906 - val_loss: 3.2345 - val_acc: 0.5000\n",
      "Epoch 803/1000\n",
      " - 4s - loss: 0.1446 - acc: 0.9375 - val_loss: 3.0814 - val_acc: 0.5000\n",
      "Epoch 804/1000\n",
      " - 4s - loss: 0.1979 - acc: 0.9375 - val_loss: 2.8763 - val_acc: 0.5000\n",
      "Epoch 805/1000\n",
      " - 4s - loss: 0.2361 - acc: 0.9375 - val_loss: 2.6912 - val_acc: 0.5357\n",
      "Epoch 806/1000\n",
      " - 4s - loss: 0.2628 - acc: 0.9140 - val_loss: 3.1326 - val_acc: 0.6071\n",
      "Epoch 807/1000\n",
      " - 4s - loss: 0.1839 - acc: 0.9453 - val_loss: 2.9077 - val_acc: 0.5000\n",
      "Epoch 808/1000\n",
      " - 4s - loss: 0.2114 - acc: 0.9140 - val_loss: 2.3873 - val_acc: 0.5000\n",
      "Epoch 809/1000\n",
      " - 4s - loss: 0.2215 - acc: 0.9453 - val_loss: 2.7270 - val_acc: 0.4286\n",
      "Epoch 810/1000\n",
      " - 4s - loss: 0.0954 - acc: 0.9609 - val_loss: 2.9869 - val_acc: 0.4643\n",
      "Epoch 811/1000\n",
      " - 4s - loss: 0.3181 - acc: 0.9141 - val_loss: 2.3444 - val_acc: 0.4643\n",
      "Epoch 812/1000\n",
      " - 4s - loss: 0.1730 - acc: 0.9297 - val_loss: 2.4339 - val_acc: 0.5357\n",
      "Epoch 813/1000\n",
      " - 4s - loss: 0.2274 - acc: 0.8988 - val_loss: 2.3643 - val_acc: 0.4286\n",
      "Epoch 814/1000\n",
      " - 4s - loss: 0.2574 - acc: 0.9140 - val_loss: 3.1180 - val_acc: 0.5714\n",
      "Epoch 815/1000\n",
      " - 4s - loss: 0.1925 - acc: 0.9297 - val_loss: 3.5452 - val_acc: 0.5357\n",
      "Epoch 816/1000\n",
      " - 4s - loss: 0.2037 - acc: 0.9297 - val_loss: 2.9107 - val_acc: 0.5357\n",
      "Epoch 817/1000\n",
      " - 4s - loss: 0.3322 - acc: 0.8828 - val_loss: 2.5916 - val_acc: 0.4286\n",
      "Epoch 818/1000\n",
      " - 4s - loss: 0.4220 - acc: 0.8911 - val_loss: 3.2893 - val_acc: 0.5000\n",
      "Epoch 819/1000\n",
      " - 4s - loss: 0.1120 - acc: 0.9609 - val_loss: 3.1595 - val_acc: 0.5000\n",
      "Epoch 820/1000\n",
      " - 4s - loss: 0.0924 - acc: 0.9766 - val_loss: 3.1745 - val_acc: 0.4286\n",
      "Epoch 821/1000\n",
      " - 5s - loss: 0.1313 - acc: 0.9453 - val_loss: 2.7150 - val_acc: 0.5000\n",
      "Epoch 822/1000\n",
      " - 5s - loss: 0.3819 - acc: 0.8674 - val_loss: 3.4264 - val_acc: 0.5000\n",
      "Epoch 823/1000\n",
      " - 4s - loss: 0.0869 - acc: 0.9609 - val_loss: 2.2896 - val_acc: 0.6071\n",
      "Epoch 824/1000\n",
      " - 4s - loss: 0.1808 - acc: 0.9531 - val_loss: 2.7972 - val_acc: 0.5000\n",
      "Epoch 825/1000\n",
      " - 5s - loss: 0.1638 - acc: 0.9453 - val_loss: 3.0602 - val_acc: 0.5000\n",
      "Epoch 826/1000\n",
      " - 4s - loss: 0.2102 - acc: 0.9218 - val_loss: 3.8008 - val_acc: 0.4643\n",
      "Epoch 827/1000\n",
      " - 4s - loss: 0.2300 - acc: 0.9218 - val_loss: 2.4783 - val_acc: 0.5000\n",
      "Epoch 828/1000\n",
      " - 4s - loss: 0.1418 - acc: 0.9297 - val_loss: 2.8255 - val_acc: 0.5714\n",
      "Epoch 829/1000\n",
      " - 4s - loss: 0.1831 - acc: 0.9531 - val_loss: 2.5185 - val_acc: 0.6071\n",
      "Epoch 830/1000\n",
      " - 4s - loss: 0.2579 - acc: 0.9297 - val_loss: 2.7086 - val_acc: 0.5714\n",
      "Epoch 831/1000\n",
      " - 4s - loss: 0.3124 - acc: 0.8908 - val_loss: 3.3680 - val_acc: 0.4643\n",
      "Epoch 832/1000\n",
      " - 4s - loss: 0.1591 - acc: 0.9453 - val_loss: 2.9728 - val_acc: 0.4643\n",
      "Epoch 833/1000\n",
      " - 4s - loss: 0.3443 - acc: 0.8984 - val_loss: 2.6711 - val_acc: 0.5357\n",
      "Epoch 834/1000\n",
      " - 4s - loss: 0.1341 - acc: 0.9531 - val_loss: 2.6493 - val_acc: 0.4643\n",
      "Epoch 835/1000\n",
      " - 4s - loss: 0.1765 - acc: 0.9531 - val_loss: 2.1754 - val_acc: 0.5714\n",
      "Epoch 836/1000\n",
      " - 4s - loss: 0.1352 - acc: 0.9453 - val_loss: 3.9706 - val_acc: 0.3571\n",
      "Epoch 837/1000\n",
      " - 4s - loss: 0.1316 - acc: 0.9609 - val_loss: 2.8094 - val_acc: 0.6071\n",
      "Epoch 838/1000\n",
      " - 4s - loss: 0.2194 - acc: 0.9374 - val_loss: 2.6635 - val_acc: 0.5357\n",
      "Epoch 839/1000\n",
      " - 4s - loss: 0.1238 - acc: 0.9609 - val_loss: 3.1734 - val_acc: 0.5357\n",
      "Epoch 840/1000\n",
      " - 4s - loss: 0.2041 - acc: 0.9218 - val_loss: 3.2407 - val_acc: 0.5357\n",
      "Epoch 841/1000\n",
      " - 4s - loss: 0.1474 - acc: 0.9218 - val_loss: 2.7400 - val_acc: 0.5357\n",
      "Epoch 842/1000\n",
      " - 4s - loss: 0.1257 - acc: 0.9609 - val_loss: 3.2551 - val_acc: 0.5357\n",
      "Epoch 843/1000\n",
      " - 4s - loss: 0.2652 - acc: 0.8986 - val_loss: 2.3248 - val_acc: 0.6071\n",
      "Epoch 844/1000\n",
      " - 4s - loss: 0.2700 - acc: 0.9375 - val_loss: 2.4246 - val_acc: 0.6071\n",
      "Epoch 845/1000\n",
      " - 4s - loss: 0.1783 - acc: 0.9218 - val_loss: 3.5421 - val_acc: 0.6071\n",
      "Epoch 846/1000\n",
      " - 4s - loss: 0.1388 - acc: 0.9609 - val_loss: 3.4910 - val_acc: 0.5000\n",
      "Epoch 847/1000\n",
      " - 4s - loss: 0.1037 - acc: 0.9609 - val_loss: 2.2482 - val_acc: 0.5357\n",
      "Epoch 848/1000\n",
      " - 4s - loss: 0.1904 - acc: 0.9218 - val_loss: 3.3782 - val_acc: 0.4643\n",
      "Epoch 849/1000\n",
      " - 4s - loss: 0.2160 - acc: 0.9299 - val_loss: 3.7359 - val_acc: 0.3929\n",
      "Epoch 850/1000\n",
      " - 4s - loss: 0.2153 - acc: 0.9218 - val_loss: 2.6979 - val_acc: 0.6429\n",
      "Epoch 851/1000\n",
      " - 4s - loss: 0.0744 - acc: 0.9844 - val_loss: 2.1343 - val_acc: 0.5357\n",
      "Epoch 852/1000\n",
      " - 4s - loss: 0.3274 - acc: 0.8984 - val_loss: 2.8628 - val_acc: 0.5357\n",
      "Epoch 853/1000\n",
      " - 4s - loss: 0.1929 - acc: 0.9609 - val_loss: 2.6523 - val_acc: 0.5000\n",
      "Epoch 854/1000\n",
      " - 4s - loss: 0.1896 - acc: 0.9297 - val_loss: 2.5411 - val_acc: 0.5357\n",
      "Epoch 855/1000\n",
      " - 4s - loss: 0.3387 - acc: 0.9145 - val_loss: 2.7761 - val_acc: 0.4643\n",
      "Epoch 856/1000\n",
      " - 4s - loss: 0.2558 - acc: 0.9062 - val_loss: 2.6566 - val_acc: 0.5357\n",
      "Epoch 857/1000\n",
      " - 4s - loss: 0.2465 - acc: 0.9297 - val_loss: 2.6940 - val_acc: 0.5000\n",
      "Epoch 858/1000\n",
      " - 4s - loss: 0.1497 - acc: 0.9375 - val_loss: 3.2053 - val_acc: 0.4643\n",
      "Epoch 859/1000\n",
      " - 3s - loss: 0.2534 - acc: 0.9375 - val_loss: 2.0833 - val_acc: 0.4643\n",
      "Epoch 860/1000\n",
      " - 4s - loss: 0.2595 - acc: 0.9140 - val_loss: 2.5839 - val_acc: 0.5000\n",
      "Epoch 861/1000\n",
      " - 4s - loss: 0.2110 - acc: 0.9221 - val_loss: 1.9917 - val_acc: 0.6429\n",
      "Epoch 862/1000\n",
      " - 4s - loss: 0.2312 - acc: 0.8984 - val_loss: 2.4526 - val_acc: 0.6429\n",
      "Epoch 863/1000\n",
      " - 4s - loss: 0.2540 - acc: 0.9218 - val_loss: 2.7442 - val_acc: 0.5714\n",
      "Epoch 864/1000\n",
      " - 4s - loss: 0.0993 - acc: 0.9531 - val_loss: 3.1150 - val_acc: 0.5000\n",
      "Epoch 865/1000\n",
      " - 4s - loss: 0.4194 - acc: 0.8593 - val_loss: 3.1470 - val_acc: 0.5000\n",
      "Epoch 866/1000\n",
      " - 4s - loss: 0.2543 - acc: 0.9297 - val_loss: 3.0670 - val_acc: 0.5357\n",
      "Epoch 867/1000\n",
      " - 4s - loss: 0.2757 - acc: 0.9062 - val_loss: 2.3382 - val_acc: 0.6071\n",
      "Epoch 868/1000\n",
      " - 4s - loss: 0.2003 - acc: 0.9223 - val_loss: 2.8011 - val_acc: 0.4643\n",
      "Epoch 869/1000\n",
      " - 4s - loss: 0.1871 - acc: 0.9221 - val_loss: 2.6891 - val_acc: 0.6429\n",
      "Epoch 870/1000\n",
      " - 4s - loss: 0.1850 - acc: 0.9143 - val_loss: 2.3075 - val_acc: 0.5714\n",
      "Epoch 871/1000\n",
      " - 4s - loss: 0.1742 - acc: 0.9218 - val_loss: 3.1332 - val_acc: 0.4643\n",
      "Epoch 872/1000\n",
      " - 4s - loss: 0.2887 - acc: 0.8984 - val_loss: 2.6866 - val_acc: 0.5000\n",
      "Epoch 873/1000\n",
      " - 4s - loss: 0.2240 - acc: 0.9062 - val_loss: 2.9200 - val_acc: 0.6429\n",
      "Epoch 874/1000\n",
      " - 4s - loss: 0.1336 - acc: 0.9687 - val_loss: 3.7355 - val_acc: 0.5000\n",
      "Epoch 875/1000\n",
      " - 4s - loss: 0.1950 - acc: 0.9221 - val_loss: 3.3256 - val_acc: 0.4286\n",
      "Epoch 876/1000\n",
      " - 4s - loss: 0.0729 - acc: 0.9687 - val_loss: 2.8946 - val_acc: 0.5357\n",
      "Epoch 877/1000\n",
      " - 4s - loss: 0.2728 - acc: 0.9219 - val_loss: 3.4696 - val_acc: 0.4286\n",
      "Epoch 878/1000\n",
      " - 4s - loss: 0.1375 - acc: 0.9531 - val_loss: 3.1066 - val_acc: 0.5000\n",
      "Epoch 879/1000\n",
      " - 4s - loss: 0.1454 - acc: 0.9453 - val_loss: 3.2106 - val_acc: 0.5357\n",
      "Epoch 880/1000\n",
      " - 4s - loss: 0.0570 - acc: 0.9766 - val_loss: 2.7535 - val_acc: 0.4643\n",
      "Epoch 881/1000\n",
      " - 4s - loss: 0.1873 - acc: 0.9374 - val_loss: 3.2462 - val_acc: 0.4643\n",
      "Epoch 882/1000\n",
      " - 4s - loss: 0.3167 - acc: 0.9141 - val_loss: 3.9766 - val_acc: 0.4286\n",
      "Epoch 883/1000\n",
      " - 4s - loss: 0.1991 - acc: 0.9375 - val_loss: 2.4173 - val_acc: 0.5714\n",
      "Epoch 884/1000\n",
      " - 4s - loss: 0.0661 - acc: 0.9922 - val_loss: 3.3864 - val_acc: 0.3929\n",
      "Epoch 885/1000\n",
      " - 4s - loss: 0.1945 - acc: 0.9218 - val_loss: 3.0300 - val_acc: 0.4643\n",
      "Epoch 886/1000\n",
      " - 4s - loss: 0.3217 - acc: 0.8984 - val_loss: 2.7854 - val_acc: 0.5357\n",
      "Epoch 887/1000\n",
      " - 4s - loss: 0.2092 - acc: 0.9453 - val_loss: 2.6713 - val_acc: 0.4643\n",
      "Epoch 888/1000\n",
      " - 4s - loss: 0.3505 - acc: 0.9218 - val_loss: 2.6691 - val_acc: 0.5714\n",
      "Epoch 889/1000\n",
      " - 4s - loss: 0.2298 - acc: 0.9377 - val_loss: 2.6765 - val_acc: 0.6071\n",
      "Epoch 890/1000\n",
      " - 4s - loss: 0.2198 - acc: 0.9219 - val_loss: 2.9088 - val_acc: 0.5000\n",
      "Epoch 891/1000\n",
      " - 4s - loss: 0.1866 - acc: 0.9297 - val_loss: 2.7453 - val_acc: 0.5714\n",
      "Epoch 892/1000\n",
      " - 4s - loss: 0.1367 - acc: 0.9297 - val_loss: 3.1290 - val_acc: 0.5357\n",
      "Epoch 893/1000\n",
      " - 4s - loss: 0.2390 - acc: 0.9297 - val_loss: 3.0040 - val_acc: 0.3571\n",
      "Epoch 894/1000\n",
      " - 4s - loss: 0.2606 - acc: 0.9062 - val_loss: 1.8733 - val_acc: 0.6786\n",
      "Epoch 895/1000\n",
      " - 4s - loss: 0.1102 - acc: 0.9531 - val_loss: 2.1519 - val_acc: 0.6071\n",
      "Epoch 896/1000\n",
      " - 4s - loss: 0.2214 - acc: 0.9297 - val_loss: 3.0433 - val_acc: 0.6071\n",
      "Epoch 897/1000\n",
      " - 4s - loss: 0.2639 - acc: 0.9140 - val_loss: 3.2644 - val_acc: 0.4286\n",
      "Epoch 898/1000\n",
      " - 4s - loss: 0.1866 - acc: 0.9375 - val_loss: 2.4835 - val_acc: 0.4643\n",
      "Epoch 899/1000\n",
      " - 4s - loss: 0.1730 - acc: 0.9375 - val_loss: 2.5033 - val_acc: 0.5000\n",
      "Epoch 900/1000\n",
      " - 4s - loss: 0.1487 - acc: 0.9609 - val_loss: 3.0773 - val_acc: 0.4286\n",
      "Epoch 901/1000\n",
      " - 4s - loss: 0.3103 - acc: 0.9140 - val_loss: 2.4263 - val_acc: 0.5714\n",
      "Epoch 902/1000\n",
      " - 4s - loss: 0.2121 - acc: 0.9140 - val_loss: 2.5387 - val_acc: 0.5714\n",
      "Epoch 903/1000\n",
      " - 4s - loss: 0.0944 - acc: 0.9609 - val_loss: 2.4370 - val_acc: 0.6429\n",
      "Epoch 904/1000\n",
      " - 4s - loss: 0.1882 - acc: 0.9375 - val_loss: 2.5897 - val_acc: 0.5000\n",
      "Epoch 905/1000\n",
      " - 4s - loss: 0.1815 - acc: 0.9453 - val_loss: 2.7288 - val_acc: 0.6071\n",
      "Epoch 906/1000\n",
      " - 4s - loss: 0.1553 - acc: 0.9609 - val_loss: 2.9031 - val_acc: 0.6071\n",
      "Epoch 907/1000\n",
      " - 4s - loss: 0.2334 - acc: 0.9220 - val_loss: 3.2548 - val_acc: 0.6071\n",
      "Epoch 908/1000\n",
      " - 4s - loss: 0.1792 - acc: 0.9375 - val_loss: 2.1579 - val_acc: 0.6071\n",
      "Epoch 909/1000\n",
      " - 4s - loss: 0.2411 - acc: 0.9297 - val_loss: 2.7037 - val_acc: 0.5000\n",
      "Epoch 910/1000\n",
      " - 4s - loss: 0.1965 - acc: 0.9143 - val_loss: 3.1231 - val_acc: 0.6071\n",
      "Epoch 911/1000\n",
      " - 4s - loss: 0.2965 - acc: 0.8986 - val_loss: 2.9690 - val_acc: 0.5714\n",
      "Epoch 912/1000\n",
      " - 4s - loss: 0.2908 - acc: 0.9062 - val_loss: 3.1011 - val_acc: 0.5357\n",
      "Epoch 913/1000\n",
      " - 4s - loss: 0.2117 - acc: 0.9377 - val_loss: 2.6400 - val_acc: 0.5357\n",
      "Epoch 914/1000\n",
      " - 4s - loss: 0.2688 - acc: 0.9064 - val_loss: 3.2284 - val_acc: 0.5357\n",
      "Epoch 915/1000\n",
      " - 4s - loss: 0.1316 - acc: 0.9609 - val_loss: 3.2157 - val_acc: 0.5000\n",
      "Epoch 916/1000\n",
      " - 4s - loss: 0.1812 - acc: 0.9297 - val_loss: 3.2125 - val_acc: 0.4286\n",
      "Epoch 917/1000\n",
      " - 4s - loss: 0.1745 - acc: 0.9375 - val_loss: 2.8627 - val_acc: 0.5357\n",
      "Epoch 918/1000\n",
      " - 4s - loss: 0.2232 - acc: 0.9375 - val_loss: 3.3543 - val_acc: 0.5000\n",
      "Epoch 919/1000\n",
      " - 4s - loss: 0.3121 - acc: 0.9062 - val_loss: 2.0923 - val_acc: 0.5000\n",
      "Epoch 920/1000\n",
      " - 4s - loss: 0.1984 - acc: 0.9219 - val_loss: 3.5742 - val_acc: 0.5714\n",
      "Epoch 921/1000\n",
      " - 4s - loss: 0.1584 - acc: 0.9297 - val_loss: 2.7351 - val_acc: 0.6071\n",
      "Epoch 922/1000\n",
      " - 4s - loss: 0.1643 - acc: 0.9531 - val_loss: 3.6828 - val_acc: 0.5357\n",
      "Epoch 923/1000\n",
      " - 4s - loss: 0.2623 - acc: 0.9062 - val_loss: 3.1109 - val_acc: 0.5000\n",
      "Epoch 924/1000\n",
      " - 4s - loss: 0.1621 - acc: 0.9531 - val_loss: 3.0320 - val_acc: 0.4643\n",
      "Epoch 925/1000\n",
      " - 4s - loss: 0.1011 - acc: 0.9609 - val_loss: 2.2589 - val_acc: 0.5357\n",
      "Epoch 926/1000\n",
      " - 4s - loss: 0.0992 - acc: 0.9453 - val_loss: 2.9798 - val_acc: 0.5357\n",
      "Epoch 927/1000\n",
      " - 4s - loss: 0.2076 - acc: 0.9297 - val_loss: 3.2565 - val_acc: 0.4643\n",
      "Epoch 928/1000\n",
      " - 4s - loss: 0.1173 - acc: 0.9453 - val_loss: 2.3242 - val_acc: 0.5714\n",
      "Epoch 929/1000\n",
      " - 4s - loss: 0.2253 - acc: 0.9453 - val_loss: 2.7237 - val_acc: 0.4643\n",
      "Epoch 930/1000\n",
      " - 4s - loss: 0.2880 - acc: 0.9140 - val_loss: 2.4147 - val_acc: 0.5357\n",
      "Epoch 931/1000\n",
      " - 4s - loss: 0.1946 - acc: 0.9531 - val_loss: 3.0403 - val_acc: 0.5000\n",
      "Epoch 932/1000\n",
      " - 4s - loss: 0.0731 - acc: 0.9766 - val_loss: 2.6372 - val_acc: 0.4643\n",
      "Epoch 933/1000\n",
      " - 4s - loss: 0.1688 - acc: 0.9296 - val_loss: 2.7882 - val_acc: 0.5000\n",
      "Epoch 934/1000\n",
      " - 4s - loss: 0.2336 - acc: 0.9218 - val_loss: 2.9801 - val_acc: 0.5714\n",
      "Epoch 935/1000\n",
      " - 4s - loss: 0.1243 - acc: 0.9531 - val_loss: 2.7704 - val_acc: 0.5357\n",
      "Epoch 936/1000\n",
      " - 4s - loss: 0.3321 - acc: 0.9062 - val_loss: 2.9474 - val_acc: 0.5714\n",
      "Epoch 937/1000\n",
      " - 4s - loss: 0.2121 - acc: 0.9531 - val_loss: 2.9165 - val_acc: 0.5000\n",
      "Epoch 938/1000\n",
      " - 4s - loss: 0.2804 - acc: 0.9299 - val_loss: 2.6236 - val_acc: 0.5000\n",
      "Epoch 939/1000\n",
      " - 4s - loss: 0.2797 - acc: 0.8906 - val_loss: 3.1633 - val_acc: 0.3571\n",
      "Epoch 940/1000\n",
      " - 4s - loss: 0.3506 - acc: 0.8984 - val_loss: 2.7295 - val_acc: 0.4286\n",
      "Epoch 941/1000\n",
      " - 4s - loss: 0.3984 - acc: 0.8906 - val_loss: 2.7677 - val_acc: 0.5714\n",
      "Epoch 942/1000\n",
      " - 4s - loss: 0.2046 - acc: 0.9375 - val_loss: 1.8122 - val_acc: 0.5714\n",
      "Epoch 943/1000\n",
      " - 4s - loss: 0.1949 - acc: 0.9375 - val_loss: 2.4032 - val_acc: 0.4643\n",
      "Epoch 944/1000\n",
      " - 4s - loss: 0.1437 - acc: 0.9531 - val_loss: 2.9231 - val_acc: 0.6071\n",
      "Epoch 945/1000\n",
      " - 4s - loss: 0.2734 - acc: 0.9062 - val_loss: 2.5085 - val_acc: 0.5000\n",
      "Epoch 946/1000\n",
      " - 4s - loss: 0.1363 - acc: 0.9453 - val_loss: 2.7185 - val_acc: 0.5714\n",
      "Epoch 947/1000\n",
      " - 4s - loss: 0.1515 - acc: 0.9687 - val_loss: 2.6828 - val_acc: 0.5714\n",
      "Epoch 948/1000\n",
      " - 4s - loss: 0.2391 - acc: 0.9375 - val_loss: 2.3573 - val_acc: 0.6071\n",
      "Epoch 949/1000\n",
      " - 4s - loss: 0.0845 - acc: 0.9687 - val_loss: 3.4338 - val_acc: 0.4643\n",
      "Epoch 950/1000\n",
      " - 4s - loss: 0.1886 - acc: 0.9453 - val_loss: 3.0719 - val_acc: 0.5714\n",
      "Epoch 951/1000\n",
      " - 4s - loss: 0.1974 - acc: 0.9375 - val_loss: 2.6020 - val_acc: 0.5000\n",
      "Epoch 952/1000\n",
      " - 4s - loss: 0.1809 - acc: 0.9375 - val_loss: 2.6796 - val_acc: 0.5000\n",
      "Epoch 953/1000\n",
      " - 4s - loss: 0.2542 - acc: 0.9062 - val_loss: 2.8209 - val_acc: 0.5357\n",
      "Epoch 954/1000\n",
      " - 4s - loss: 0.1593 - acc: 0.9453 - val_loss: 2.7876 - val_acc: 0.6071\n",
      "Epoch 955/1000\n",
      " - 4s - loss: 0.1459 - acc: 0.9218 - val_loss: 3.3966 - val_acc: 0.5000\n",
      "Epoch 956/1000\n",
      " - 4s - loss: 0.2124 - acc: 0.9531 - val_loss: 3.7879 - val_acc: 0.5000\n",
      "Epoch 957/1000\n",
      " - 4s - loss: 0.2956 - acc: 0.8828 - val_loss: 3.1539 - val_acc: 0.4643\n",
      "Epoch 958/1000\n",
      " - 4s - loss: 0.1745 - acc: 0.9531 - val_loss: 2.6231 - val_acc: 0.5357\n",
      "Epoch 959/1000\n",
      " - 4s - loss: 0.1196 - acc: 0.9531 - val_loss: 2.2316 - val_acc: 0.5357\n",
      "Epoch 960/1000\n",
      " - 4s - loss: 0.2920 - acc: 0.9297 - val_loss: 3.1413 - val_acc: 0.5000\n",
      "Epoch 961/1000\n",
      " - 4s - loss: 0.0825 - acc: 0.9688 - val_loss: 2.8958 - val_acc: 0.5357\n",
      "Epoch 962/1000\n",
      " - 4s - loss: 0.2815 - acc: 0.9533 - val_loss: 2.8295 - val_acc: 0.5357\n",
      "Epoch 963/1000\n",
      " - 4s - loss: 0.1901 - acc: 0.9453 - val_loss: 3.2310 - val_acc: 0.4286\n",
      "Epoch 964/1000\n",
      " - 4s - loss: 0.2294 - acc: 0.9140 - val_loss: 2.8465 - val_acc: 0.5357\n",
      "Epoch 965/1000\n",
      " - 4s - loss: 0.1870 - acc: 0.9221 - val_loss: 3.6503 - val_acc: 0.4643\n",
      "Epoch 966/1000\n",
      " - 4s - loss: 0.2610 - acc: 0.8984 - val_loss: 2.9122 - val_acc: 0.5357\n",
      "Epoch 967/1000\n",
      " - 4s - loss: 0.3418 - acc: 0.9064 - val_loss: 2.4355 - val_acc: 0.6071\n",
      "Epoch 968/1000\n",
      " - 4s - loss: 0.2267 - acc: 0.9218 - val_loss: 3.0088 - val_acc: 0.4643\n",
      "Epoch 969/1000\n",
      " - 4s - loss: 0.1230 - acc: 0.9766 - val_loss: 3.2741 - val_acc: 0.4643\n",
      "Epoch 970/1000\n",
      " - 4s - loss: 0.4480 - acc: 0.8984 - val_loss: 2.3558 - val_acc: 0.5714\n",
      "Epoch 971/1000\n",
      " - 4s - loss: 0.3460 - acc: 0.9140 - val_loss: 3.1658 - val_acc: 0.5357\n",
      "Epoch 972/1000\n",
      " - 4s - loss: 0.1610 - acc: 0.9609 - val_loss: 3.9226 - val_acc: 0.4286\n",
      "Epoch 973/1000\n",
      " - 4s - loss: 0.0931 - acc: 0.9609 - val_loss: 3.0811 - val_acc: 0.5000\n",
      "Epoch 974/1000\n",
      " - 4s - loss: 0.0592 - acc: 0.9688 - val_loss: 3.6018 - val_acc: 0.5357\n",
      "Epoch 975/1000\n",
      " - 4s - loss: 0.3350 - acc: 0.8984 - val_loss: 2.7652 - val_acc: 0.4643\n",
      "Epoch 976/1000\n",
      " - 4s - loss: 0.1679 - acc: 0.9609 - val_loss: 2.6188 - val_acc: 0.6429\n",
      "Epoch 977/1000\n",
      " - 4s - loss: 0.1035 - acc: 0.9766 - val_loss: 2.8679 - val_acc: 0.5357\n",
      "Epoch 978/1000\n",
      " - 4s - loss: 0.1504 - acc: 0.9687 - val_loss: 2.6913 - val_acc: 0.5714\n",
      "Epoch 979/1000\n",
      " - 4s - loss: 0.1740 - acc: 0.9531 - val_loss: 3.1010 - val_acc: 0.5000\n",
      "Epoch 980/1000\n",
      " - 4s - loss: 0.2572 - acc: 0.9453 - val_loss: 2.4026 - val_acc: 0.6429\n",
      "Epoch 981/1000\n",
      " - 4s - loss: 0.3347 - acc: 0.8906 - val_loss: 2.1177 - val_acc: 0.5000\n",
      "Epoch 982/1000\n",
      " - 4s - loss: 0.1312 - acc: 0.9533 - val_loss: 3.1799 - val_acc: 0.5000\n",
      "Epoch 983/1000\n",
      " - 4s - loss: 0.3010 - acc: 0.9140 - val_loss: 2.3829 - val_acc: 0.5714\n",
      "Epoch 984/1000\n",
      " - 4s - loss: 0.1319 - acc: 0.9609 - val_loss: 3.0750 - val_acc: 0.6429\n",
      "Epoch 985/1000\n",
      " - 4s - loss: 0.2013 - acc: 0.9221 - val_loss: 3.0692 - val_acc: 0.5714\n",
      "Epoch 986/1000\n",
      " - 4s - loss: 0.1885 - acc: 0.9062 - val_loss: 3.1405 - val_acc: 0.5357\n",
      "Epoch 987/1000\n",
      " - 4s - loss: 0.2641 - acc: 0.9221 - val_loss: 3.3970 - val_acc: 0.5357\n",
      "Epoch 988/1000\n",
      " - 4s - loss: 0.2976 - acc: 0.9218 - val_loss: 2.8570 - val_acc: 0.5357\n",
      "Epoch 989/1000\n",
      " - 4s - loss: 0.1180 - acc: 0.9688 - val_loss: 4.0021 - val_acc: 0.5000\n",
      "Epoch 990/1000\n",
      " - 4s - loss: 0.0825 - acc: 0.9609 - val_loss: 3.7900 - val_acc: 0.3571\n",
      "Epoch 991/1000\n",
      " - 4s - loss: 0.2111 - acc: 0.8984 - val_loss: 3.0984 - val_acc: 0.4643\n",
      "Epoch 992/1000\n",
      " - 4s - loss: 0.0714 - acc: 0.9687 - val_loss: 2.4760 - val_acc: 0.5714\n",
      "Epoch 993/1000\n",
      " - 4s - loss: 0.1687 - acc: 0.9375 - val_loss: 3.0009 - val_acc: 0.4643\n",
      "Epoch 994/1000\n",
      " - 4s - loss: 0.1765 - acc: 0.9531 - val_loss: 3.1991 - val_acc: 0.4643\n",
      "Epoch 995/1000\n",
      " - 4s - loss: 0.1838 - acc: 0.9531 - val_loss: 2.9890 - val_acc: 0.5357\n",
      "Epoch 996/1000\n",
      " - 4s - loss: 0.1978 - acc: 0.9140 - val_loss: 2.5003 - val_acc: 0.5714\n",
      "Epoch 997/1000\n",
      " - 4s - loss: 0.1489 - acc: 0.9609 - val_loss: 2.9091 - val_acc: 0.5714\n",
      "Epoch 998/1000\n",
      " - 4s - loss: 0.2649 - acc: 0.9218 - val_loss: 2.7952 - val_acc: 0.3929\n",
      "Epoch 999/1000\n",
      " - 4s - loss: 0.1717 - acc: 0.9531 - val_loss: 3.3237 - val_acc: 0.5714\n",
      "Epoch 1000/1000\n",
      " - 4s - loss: 0.1737 - acc: 0.9375 - val_loss: 3.6903 - val_acc: 0.4643\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f6d967dbb38>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit_generator(\n",
    "    train_generator,\n",
    "    steps_per_epoch=train_step // batch_size,\n",
    "    epochs=epochs,\n",
    "    validation_data=validation_generator,\n",
    "    validation_steps=valid_step // batch_size,\n",
    "    callbacks=[tensorboard],\n",
    "    verbose=2\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_path = './saved/svgg'+date_str+'_'+str(epochs)+'x.h5'\n",
    "model.save(model_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1\n",
      "1\n",
      "3\n",
      "3\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f6d27845da0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x, y = validation_generator.next()\n",
    "img = x[0] * 255\n",
    "img = img.astype(\"uint8\")\n",
    "plt.imshow(img, cmap = 'gray')\n",
    "\n",
    "for yy in y:\n",
    "    print(np.argmax(yy)+1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred = model.predict(x)\n",
    "for pr in pred:\n",
    "    print(np.argmax(pr)+1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:dlearn]",
   "language": "python",
   "name": "conda-env-dlearn-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}