LSTM_Training.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "machine_shape": "hm",
      "gpuClass": "premium"
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "gpuClass": "premium"
  },
  "cells": [
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "i6szVJsNf621"
      },
      "outputs": [],
      "source": [
        "# import core libraries\n",
        "import sys\n",
        "import os\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import json\n",
        "from sklearn.model_selection import train_test_split\n",
        "import random as rn\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "\n",
        "%tensorflow_version 2.x\n",
        "import tensorflow as tf\n",
        "print(\"TensorFlow version:\", tf.__version__)"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# load Keras libraries\n",
        "from tensorflow.keras.models import Model\n",
        "from tensorflow.keras.layers import Input, LSTM, Dense, Dropout, Concatenate\n",
        "from tensorflow.keras.initializers import RandomNormal\n",
        "from tensorflow.keras import regularizers\n",
        "from tensorflow.keras.optimizers import Adam\n",
        "from tensorflow.keras.callbacks import ReduceLROnPlateau, ModelCheckpoint, TensorBoard\n",
        "from tensorflow.keras.utils import Sequence\n",
        "import pickle"
      ],
      "metadata": {
        "id": "1SL1s6rkhjY7"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# enable GPU processing\n",
        "device_name = tf.test.gpu_device_name()\n",
        "if device_name != '/device:GPU:0':\n",
        "  raise SystemError('GPU device not found')\n",
        "print('Found GPU at: {}'.format(device_name))\n",
        "gpu_info = !nvidia-smi\n",
        "gpu_info = '\\n'.join(gpu_info)\n",
        "if gpu_info.find('failed') >= 0:\n",
        "  print('Select the Runtime → \"Change runtime type\" menu to enable a GPU accelerator, ')\n",
        "  print('and then re-execute this cell.')\n",
        "else:\n",
        "  print(gpu_info)"
      ],
      "metadata": {
        "id": "ovkVHP5Nhle6"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# set random seed\n",
        "seed = 777\n",
        "np.random.seed(seed)\n",
        "rn.seed(seed)"
      ],
      "metadata": {
        "id": "KNpkJLCxhnhN"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# create data load functions\n",
        "def load_data(data):\n",
        "    with open(data, 'rb') as fp:\n",
        "        smiles = pickle.load(fp)\n",
        "\n",
        "    return smiles\n",
        "\n",
        "def load_dictionaries(input_dict):\n",
        "    with open(input_dict, 'r') as fp:\n",
        "        new_dict = json.load(fp)\n",
        "    return new_dict"
      ],
      "metadata": {
        "id": "CIenMgbUho5E"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# load dataset & break into train/test sets\n",
        "smifile = '/content/valid_smiles_data'\n",
        "data = load_data(smifile)\n",
        "data = [mol for mol in data if len(mol) < 99]\n",
        "data = np.array(data)\n",
        "full_train, test = train_test_split(data, test_size=0.2, random_state=seed)\n",
        "print(\"Training smiles shape:\", full_train.shape)\n",
        "print(\"Test smiles shape:\", test.shape)\n",
        "print(\"Sample training smile: \\n\", full_train[0])"
      ],
      "metadata": {
        "id": "nr_YvmH9hrRE"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# # save test file for smile generation\n",
        "# np.save('/drive/My Drive/chem_data/test.npy', test)\n",
        "# print('Test file saved to disk')"
      ],
      "metadata": {
        "id": "lAu3LbTchtAx"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# create our validation dataset\n",
        "val_split = 0.10\n",
        "train, val_set = train_test_split(full_train, test_size=val_split, random_state=seed)\n",
        "print(\"Training dataset shape:\", train.shape)\n",
        "print(\"Validation dataset shape:\", val_set.shape)"
      ],
      "metadata": {
        "id": "8wpTeiffhxQh"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "#find unique chars\n",
        "chars = []\n",
        "for mol in data:\n",
        "  chars.append(set(mol))\n",
        "\n",
        "chars = list(set([i for mol in chars for i in mol]))\n",
        "chars.append('E')\n",
        "chars.append('!')\n",
        "print(chars)"
      ],
      "metadata": {
        "id": "GX_0KnzXQ_s1"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "#Create array from characters in the dataset\n",
        "from numpy import argmax\n",
        "from sklearn.preprocessing import LabelEncoder\n",
        "from numpy import array\n",
        "\n",
        "values = array(chars)\n",
        "# print(\"Array of unique characters:\")\n",
        "# print(values)\n",
        "\n",
        "#Create unique, numerical labels for each character between 0 and n-1, where n is the number of unique characters\n",
        "label_encoder = LabelEncoder()\n",
        "integer_encoded = label_encoder.fit_transform(values)\n",
        "#print(\"Array of labels for each character:\")\n",
        "#print(integer_encoded)\n",
        "\n",
        "values = np.reshape(values, (np.shape(values)[0], 1))\n",
        "integer_encoded = integer_encoded.reshape(len(integer_encoded), 1)\n",
        "#print(values.shape)\n",
        "vocab = np.concatenate((values, integer_encoded.astype(object)), axis = 1)\n",
        "vocab = vocab[vocab[:,1].argsort()]\n",
        "#print(vocab)\n",
        "int_to_char = {}\n",
        "char_to_int = {}\n",
        "for row in vocab:\n",
        "  int_to_char[str(row[1])]=row[0]\n",
        "  char_to_int[row[0]] = row[1]\n",
        "print(int_to_char)\n",
        "print(char_to_int)"
      ],
      "metadata": {
        "id": "0JYZ1A8OS2R2"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# load the dictionaries that map characters-to-integers and intergers-to-characters\n",
        "n_vocab = len(char_to_int)\n",
        "print(\"Character set vocabulary length:\", n_vocab)\n",
        "print(\"Dictionary mapping characters-to-integers:\\n\", char_to_int)\n",
        "print(\"Dictionary mapping integers-to-characters:\\n\", int_to_char)"
      ],
      "metadata": {
        "id": "4EdfQP6rhzVt"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# create a function to turn the dataset into a supervised problem, add the beginning and ending character markers,\n",
        "# add padding for constant sequence length, and turn the sequence into a sequence of one-hot vectors\n",
        "def vectorize(smiles, embed, n_vocab):\n",
        "    one_hot = np.zeros((smiles.shape[0], embed, n_vocab), dtype=np.int8)\n",
        "    for i, smile in enumerate(smiles):\n",
        "        # encode the start\n",
        "        one_hot[i,0,char_to_int[\"!\"]] = 1\n",
        "        #print(one_hot)\n",
        "        #encode the smiles characters\n",
        "        for j, c in enumerate(smile):\n",
        "            one_hot[i,j+1,char_to_int[c]] = 1\n",
        "        # encode the end of the smiles string\n",
        "        one_hot[i,len(smile)+1:,char_to_int[\"E\"]] = 1\n",
        "    # return two items, one for input and one for output\n",
        "    return one_hot[:,0:-1,:], one_hot[:,1:,:]"
      ],
      "metadata": {
        "id": "_u3YL54Vh0YN"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# create our X & y datasets\n",
        "embed = 101\n",
        "X_train, y_train = vectorize(train, embed, n_vocab)\n",
        "X_val, y_val = vectorize(val_set, embed, n_vocab)\n",
        "X_test, y_test = vectorize(test, embed, n_vocab)\n",
        "print(\"Training input shape:\", X_train.shape)\n",
        "print(\"Training output shape:\", y_train.shape)\n",
        "print(\"Validation input shape:\", X_val.shape)\n",
        "print(\"Validation output shape:\", y_val.shape)\n",
        "print(\"Test input shape:\", X_test.shape)\n",
        "print(\"Test output shape:\", y_test.shape)"
      ],
      "metadata": {
        "id": "s-0JFprah1wh"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "\"\".join([int_to_char[str(idx)] for idx in np.argmax(X_train[0,:,:], axis=1)])"
      ],
      "metadata": {
        "id": "WunmW1NTh_XJ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# define the LSTM Chem model\n",
        "def lstm_model(X, y):\n",
        "    # define the encoder layers\n",
        "    enc_input = Input(shape=(X.shape[1:]))\n",
        "    _, state_h, state_c = LSTM(256, return_state=True)(enc_input)\n",
        "    states = Concatenate(axis=-1)([state_h, state_c])\n",
        "    bottle_neck = Dense(128, activation='relu')(states)\n",
        "\n",
        "    # define the decoder layers\n",
        "    state_h_decoded = Dense(256, activation='relu')(bottle_neck)\n",
        "    state_c_decoded = Dense(256, activation='relu')(bottle_neck)\n",
        "    encoder_states = [state_h_decoded, state_c_decoded]\n",
        "    dec_input = Input(shape=(X.shape[1:]))\n",
        "    dec1 = LSTM(256, return_sequences=True)(dec_input, initial_state=encoder_states)\n",
        "    output = Dense(y.shape[2], activation='softmax')(dec1)\n",
        "\n",
        "    model = Model(inputs=[enc_input, dec_input], outputs=output)\n",
        "    return model"
      ],
      "metadata": {
        "id": "EEigzsfDiAj5"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# instantiate our neural network model\n",
        "model = lstm_model(X_train, y_train)\n",
        "model.summary()"
      ],
      "metadata": {
        "id": "XFhGR-1FiCA1"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# compile the model\n",
        "batch_size = 256\n",
        "steps_per_epoch = len(X_train) // batch_size\n",
        "lr_schedule = tf.keras.optimizers.schedules.InverseTimeDecay(0.001, decay_steps=steps_per_epoch*50, decay_rate=1.0, staircase=False)\n",
        "opt = Adam()\n",
        "model.compile(optimizer=opt, loss='categorical_crossentropy', metrics=['acc'])"
      ],
      "metadata": {
        "id": "bVqUyuTMiDgw"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# define our data sequence generator class\n",
        "class Data_Generator(Sequence):\n",
        "    def __init__(self, input_data, labels, batch_size):\n",
        "        self.input_data, self.labels = input_data, labels\n",
        "        self.batch_size = batch_size\n",
        "\n",
        "    def __len__(self):\n",
        "        return int(np.ceil(len(self.input_data) / float(self.batch_size)))\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        x = self.input_data[idx * self.batch_size:(idx + 1) * self.batch_size]\n",
        "        y = self.labels[idx * self.batch_size:(idx + 1) * self.batch_size]\n",
        "\n",
        "        batch_x, batch_y = np.array(x), np.array(y)\n",
        "\n",
        "        return [batch_x, batch_x], batch_y"
      ],
      "metadata": {
        "id": "qViPtchRiE_6"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "\n",
        "# create object instances of our sequence generator\n",
        "training_generator = Data_Generator(X_train, y_train, batch_size)\n",
        "validation_generator = Data_Generator(X_val, y_val, batch_size)"
      ],
      "metadata": {
        "id": "A5dg0N4PiGTt"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# create our callbacks\n",
        "file=\"LSTM_Chem_weights-{epoch:02d}-{acc:.4f}-{val_loss:.4f}.hdf5\"\n",
        "checkpoint = ModelCheckpoint(file, monitor='val_loss', verbose=1, save_best_only=True, save_weights_only=True, mode='min')\n",
        "rlrp = ReduceLROnPlateau(monitor='val_loss', factor=0.5, patience=10, min_lr=1e-5)\n",
        "#tb = TensorBoard(log_dir=\"/drive/My Drive/logs/chem/\", histogram_freq=1, write_graph=True, write_images=True)"
      ],
      "metadata": {
        "id": "aCk_5nmDiHUJ"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# fit the model\n",
        "nb_epochs = 225\n",
        "validation_steps = len(X_val) // batch_size\n",
        "history = model.fit(training_generator, steps_per_epoch=steps_per_epoch, epochs=nb_epochs, verbose=1,\n",
        "                              validation_data=validation_generator, validation_steps=validation_steps,\n",
        "                             use_multiprocessing=False, shuffle=True, callbacks=[])"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lX66vXdTiIkN",
        "outputId": "5a898a98-748c-44a7-882e-96f631390bf0"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/225\n",
            "1355/1355 [==============================] - 58s 41ms/step - loss: 0.5786 - acc: 0.8237 - val_loss: 0.3831 - val_acc: 0.8749\n",
            "Epoch 2/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.3468 - acc: 0.8839 - val_loss: 0.3176 - val_acc: 0.8918\n",
            "Epoch 3/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.3038 - acc: 0.8953 - val_loss: 0.2894 - val_acc: 0.8993\n",
            "Epoch 4/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.2831 - acc: 0.9010 - val_loss: 0.2760 - val_acc: 0.9030\n",
            "Epoch 5/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.2686 - acc: 0.9054 - val_loss: 0.2614 - val_acc: 0.9079\n",
            "Epoch 6/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.2593 - acc: 0.9083 - val_loss: 0.2538 - val_acc: 0.9101\n",
            "Epoch 7/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.2513 - acc: 0.9107 - val_loss: 0.2465 - val_acc: 0.9125\n",
            "Epoch 8/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.2390 - acc: 0.9145 - val_loss: 0.2247 - val_acc: 0.9193\n",
            "Epoch 9/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.2229 - acc: 0.9205 - val_loss: 0.2142 - val_acc: 0.9233\n",
            "Epoch 10/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.2034 - acc: 0.9275 - val_loss: 0.1943 - val_acc: 0.9306\n",
            "Epoch 11/225\n",
            "1355/1355 [==============================] - 55s 40ms/step - loss: 0.1870 - acc: 0.9333 - val_loss: 0.1775 - val_acc: 0.9367\n",
            "Epoch 12/225\n",
            "1355/1355 [==============================] - 55s 40ms/step - loss: 0.1752 - acc: 0.9373 - val_loss: 0.1683 - val_acc: 0.9399\n",
            "Epoch 13/225\n",
            "1355/1355 [==============================] - 55s 40ms/step - loss: 0.1681 - acc: 0.9399 - val_loss: 0.1638 - val_acc: 0.9415\n",
            "Epoch 14/225\n",
            "1355/1355 [==============================] - 55s 40ms/step - loss: 0.1618 - acc: 0.9422 - val_loss: 0.1824 - val_acc: 0.9346\n",
            "Epoch 15/225\n",
            "1355/1355 [==============================] - 55s 40ms/step - loss: 0.1569 - acc: 0.9439 - val_loss: 0.1569 - val_acc: 0.9439\n",
            "Epoch 16/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1535 - acc: 0.9452 - val_loss: 0.1554 - val_acc: 0.9446\n",
            "Epoch 17/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1495 - acc: 0.9467 - val_loss: 0.1465 - val_acc: 0.9480\n",
            "Epoch 18/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1465 - acc: 0.9478 - val_loss: 0.1461 - val_acc: 0.9482\n",
            "Epoch 19/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1438 - acc: 0.9488 - val_loss: 0.1462 - val_acc: 0.9477\n",
            "Epoch 20/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1388 - acc: 0.9507 - val_loss: 0.1359 - val_acc: 0.9519\n",
            "Epoch 21/225\n",
            "1355/1355 [==============================] - 55s 40ms/step - loss: 0.1327 - acc: 0.9529 - val_loss: 0.1315 - val_acc: 0.9534\n",
            "Epoch 22/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1734 - acc: 0.9391 - val_loss: 0.1679 - val_acc: 0.9411\n",
            "Epoch 23/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1578 - acc: 0.9444 - val_loss: 0.1553 - val_acc: 0.9448\n",
            "Epoch 24/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1416 - acc: 0.9498 - val_loss: 0.1371 - val_acc: 0.9515\n",
            "Epoch 25/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1329 - acc: 0.9529 - val_loss: 0.1339 - val_acc: 0.9526\n",
            "Epoch 26/225\n",
            "1355/1355 [==============================] - 55s 40ms/step - loss: 0.1258 - acc: 0.9555 - val_loss: 0.1232 - val_acc: 0.9566\n",
            "Epoch 27/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1193 - acc: 0.9579 - val_loss: 0.1167 - val_acc: 0.9589\n",
            "Epoch 28/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1133 - acc: 0.9600 - val_loss: 0.1113 - val_acc: 0.9609\n",
            "Epoch 29/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1081 - acc: 0.9619 - val_loss: 0.1175 - val_acc: 0.9587\n",
            "Epoch 30/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.1037 - acc: 0.9636 - val_loss: 0.1031 - val_acc: 0.9639\n",
            "Epoch 31/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0995 - acc: 0.9651 - val_loss: 0.1028 - val_acc: 0.9638\n",
            "Epoch 32/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0961 - acc: 0.9663 - val_loss: 0.0972 - val_acc: 0.9660\n",
            "Epoch 33/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0921 - acc: 0.9678 - val_loss: 0.0902 - val_acc: 0.9687\n",
            "Epoch 34/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0885 - acc: 0.9691 - val_loss: 0.0854 - val_acc: 0.9705\n",
            "Epoch 35/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0842 - acc: 0.9706 - val_loss: 0.0858 - val_acc: 0.9701\n",
            "Epoch 36/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0805 - acc: 0.9720 - val_loss: 0.0770 - val_acc: 0.9735\n",
            "Epoch 37/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0761 - acc: 0.9736 - val_loss: 0.0779 - val_acc: 0.9730\n",
            "Epoch 38/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0724 - acc: 0.9749 - val_loss: 0.0831 - val_acc: 0.9709\n",
            "Epoch 39/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0682 - acc: 0.9764 - val_loss: 0.0657 - val_acc: 0.9776\n",
            "Epoch 40/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0650 - acc: 0.9776 - val_loss: 0.0632 - val_acc: 0.9784\n",
            "Epoch 41/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0609 - acc: 0.9791 - val_loss: 0.0647 - val_acc: 0.9777\n",
            "Epoch 42/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0570 - acc: 0.9805 - val_loss: 0.0566 - val_acc: 0.9807\n",
            "Epoch 43/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0533 - acc: 0.9817 - val_loss: 0.0549 - val_acc: 0.9812\n",
            "Epoch 44/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0510 - acc: 0.9827 - val_loss: 0.0478 - val_acc: 0.9839\n",
            "Epoch 45/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0474 - acc: 0.9839 - val_loss: 0.0502 - val_acc: 0.9829\n",
            "Epoch 46/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0438 - acc: 0.9851 - val_loss: 0.0457 - val_acc: 0.9845\n",
            "Epoch 47/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0415 - acc: 0.9860 - val_loss: 0.0442 - val_acc: 0.9850\n",
            "Epoch 48/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0384 - acc: 0.9870 - val_loss: 0.0429 - val_acc: 0.9854\n",
            "Epoch 49/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0359 - acc: 0.9879 - val_loss: 0.0354 - val_acc: 0.9883\n",
            "Epoch 50/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0336 - acc: 0.9887 - val_loss: 0.0331 - val_acc: 0.9890\n",
            "Epoch 51/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0319 - acc: 0.9893 - val_loss: 0.0330 - val_acc: 0.9890\n",
            "Epoch 52/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0300 - acc: 0.9899 - val_loss: 0.0299 - val_acc: 0.9902\n",
            "Epoch 53/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0283 - acc: 0.9905 - val_loss: 0.0309 - val_acc: 0.9897\n",
            "Epoch 54/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0267 - acc: 0.9910 - val_loss: 0.0292 - val_acc: 0.9903\n",
            "Epoch 55/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0255 - acc: 0.9915 - val_loss: 0.0256 - val_acc: 0.9916\n",
            "Epoch 56/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0242 - acc: 0.9919 - val_loss: 0.0264 - val_acc: 0.9912\n",
            "Epoch 57/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0256 - acc: 0.9916 - val_loss: 0.0267 - val_acc: 0.9911\n",
            "Epoch 58/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0217 - acc: 0.9928 - val_loss: 0.0239 - val_acc: 0.9921\n",
            "Epoch 59/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0207 - acc: 0.9931 - val_loss: 0.0240 - val_acc: 0.9921\n",
            "Epoch 60/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0208 - acc: 0.9931 - val_loss: 0.0221 - val_acc: 0.9928\n",
            "Epoch 61/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0189 - acc: 0.9937 - val_loss: 0.0195 - val_acc: 0.9937\n",
            "Epoch 62/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0186 - acc: 0.9938 - val_loss: 0.0207 - val_acc: 0.9932\n",
            "Epoch 63/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0181 - acc: 0.9941 - val_loss: 0.0189 - val_acc: 0.9938\n",
            "Epoch 64/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0170 - acc: 0.9944 - val_loss: 0.0181 - val_acc: 0.9941\n",
            "Epoch 65/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0162 - acc: 0.9946 - val_loss: 0.0445 - val_acc: 0.9860\n",
            "Epoch 66/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0151 - acc: 0.9950 - val_loss: 0.0156 - val_acc: 0.9950\n",
            "Epoch 67/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0192 - acc: 0.9941 - val_loss: 0.1132 - val_acc: 0.9698\n",
            "Epoch 68/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0155 - acc: 0.9950 - val_loss: 0.0152 - val_acc: 0.9951\n",
            "Epoch 69/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0139 - acc: 0.9954 - val_loss: 0.0155 - val_acc: 0.9950\n",
            "Epoch 70/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0128 - acc: 0.9958 - val_loss: 0.0141 - val_acc: 0.9955\n",
            "Epoch 71/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0125 - acc: 0.9959 - val_loss: 0.0132 - val_acc: 0.9958\n",
            "Epoch 72/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0128 - acc: 0.9958 - val_loss: 0.0131 - val_acc: 0.9958\n",
            "Epoch 73/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0122 - acc: 0.9960 - val_loss: 0.0126 - val_acc: 0.9960\n",
            "Epoch 74/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0106 - acc: 0.9965 - val_loss: 0.0141 - val_acc: 0.9954\n",
            "Epoch 75/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0107 - acc: 0.9965 - val_loss: 0.0130 - val_acc: 0.9958\n",
            "Epoch 76/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0102 - acc: 0.9966 - val_loss: 0.0117 - val_acc: 0.9963\n",
            "Epoch 77/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0118 - acc: 0.9964 - val_loss: 0.0110 - val_acc: 0.9965\n",
            "Epoch 78/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0109 - acc: 0.9966 - val_loss: 0.0115 - val_acc: 0.9964\n",
            "Epoch 79/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0090 - acc: 0.9971 - val_loss: 0.0134 - val_acc: 0.9956\n",
            "Epoch 80/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0102 - acc: 0.9968 - val_loss: 0.0107 - val_acc: 0.9966\n",
            "Epoch 81/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0082 - acc: 0.9973 - val_loss: 0.0100 - val_acc: 0.9968\n",
            "Epoch 82/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0090 - acc: 0.9971 - val_loss: 0.0103 - val_acc: 0.9968\n",
            "Epoch 83/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0103 - acc: 0.9967 - val_loss: 0.0098 - val_acc: 0.9970\n",
            "Epoch 84/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0070 - acc: 0.9978 - val_loss: 0.0095 - val_acc: 0.9970\n",
            "Epoch 85/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0073 - acc: 0.9976 - val_loss: 0.0116 - val_acc: 0.9963\n",
            "Epoch 86/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0078 - acc: 0.9975 - val_loss: 0.0209 - val_acc: 0.9932\n",
            "Epoch 87/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0078 - acc: 0.9975 - val_loss: 0.0089 - val_acc: 0.9972\n",
            "Epoch 88/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0068 - acc: 0.9978 - val_loss: 0.0110 - val_acc: 0.9965\n",
            "Epoch 89/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0070 - acc: 0.9977 - val_loss: 0.0092 - val_acc: 0.9972\n",
            "Epoch 90/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0069 - acc: 0.9977 - val_loss: 0.0094 - val_acc: 0.9971\n",
            "Epoch 91/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0074 - acc: 0.9977 - val_loss: 0.0092 - val_acc: 0.9971\n",
            "Epoch 92/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0065 - acc: 0.9979 - val_loss: 0.0094 - val_acc: 0.9971\n",
            "Epoch 93/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0073 - acc: 0.9978 - val_loss: 0.0077 - val_acc: 0.9976\n",
            "Epoch 94/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0059 - acc: 0.9981 - val_loss: 0.0077 - val_acc: 0.9976\n",
            "Epoch 95/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0067 - acc: 0.9978 - val_loss: 0.0084 - val_acc: 0.9974\n",
            "Epoch 96/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0051 - acc: 0.9984 - val_loss: 0.0088 - val_acc: 0.9972\n",
            "Epoch 97/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0055 - acc: 0.9982 - val_loss: 0.0086 - val_acc: 0.9973\n",
            "Epoch 98/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0058 - acc: 0.9981 - val_loss: 0.0071 - val_acc: 0.9978\n",
            "Epoch 99/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0065 - acc: 0.9980 - val_loss: 0.0064 - val_acc: 0.9981\n",
            "Epoch 100/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0041 - acc: 0.9987 - val_loss: 0.0081 - val_acc: 0.9975\n",
            "Epoch 101/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0050 - acc: 0.9984 - val_loss: 0.0092 - val_acc: 0.9971\n",
            "Epoch 102/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0054 - acc: 0.9983 - val_loss: 0.0075 - val_acc: 0.9977\n",
            "Epoch 103/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0058 - acc: 0.9982 - val_loss: 0.0063 - val_acc: 0.9981\n",
            "Epoch 104/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0039 - acc: 0.9987 - val_loss: 0.0077 - val_acc: 0.9976\n",
            "Epoch 105/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0057 - acc: 0.9982 - val_loss: 0.0103 - val_acc: 0.9968\n",
            "Epoch 106/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0038 - acc: 0.9988 - val_loss: 0.0084 - val_acc: 0.9975\n",
            "Epoch 107/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0043 - acc: 0.9986 - val_loss: 0.0079 - val_acc: 0.9976\n",
            "Epoch 108/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0051 - acc: 0.9984 - val_loss: 0.0058 - val_acc: 0.9983\n",
            "Epoch 109/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0042 - acc: 0.9987 - val_loss: 0.0059 - val_acc: 0.9983\n",
            "Epoch 110/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0038 - acc: 0.9988 - val_loss: 0.0073 - val_acc: 0.9978\n",
            "Epoch 111/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0043 - acc: 0.9986 - val_loss: 0.0070 - val_acc: 0.9979\n",
            "Epoch 112/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0038 - acc: 0.9987 - val_loss: 0.0076 - val_acc: 0.9977\n",
            "Epoch 113/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0051 - acc: 0.9985 - val_loss: 0.0054 - val_acc: 0.9984\n",
            "Epoch 114/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0031 - acc: 0.9990 - val_loss: 0.0071 - val_acc: 0.9979\n",
            "Epoch 115/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0043 - acc: 0.9986 - val_loss: 0.0058 - val_acc: 0.9983\n",
            "Epoch 116/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0035 - acc: 0.9988 - val_loss: 0.0189 - val_acc: 0.9944\n",
            "Epoch 117/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0039 - acc: 0.9987 - val_loss: 0.0074 - val_acc: 0.9977\n",
            "Epoch 118/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0033 - acc: 0.9989 - val_loss: 0.0075 - val_acc: 0.9977\n",
            "Epoch 119/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0037 - acc: 0.9987 - val_loss: 0.0062 - val_acc: 0.9982\n",
            "Epoch 120/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0043 - acc: 0.9987 - val_loss: 0.0051 - val_acc: 0.9985\n",
            "Epoch 121/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0025 - acc: 0.9992 - val_loss: 0.0071 - val_acc: 0.9979\n",
            "Epoch 122/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0035 - acc: 0.9988 - val_loss: 0.0059 - val_acc: 0.9983\n",
            "Epoch 123/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0043 - acc: 0.9987 - val_loss: 0.0050 - val_acc: 0.9986\n",
            "Epoch 124/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0035 - acc: 0.9989 - val_loss: 0.0118 - val_acc: 0.9963\n",
            "Epoch 125/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0024 - acc: 0.9992 - val_loss: 0.0083 - val_acc: 0.9976\n",
            "Epoch 126/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0035 - acc: 0.9989 - val_loss: 0.0053 - val_acc: 0.9985\n",
            "Epoch 127/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0029 - acc: 0.9990 - val_loss: 0.0083 - val_acc: 0.9975\n",
            "Epoch 128/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0036 - acc: 0.9989 - val_loss: 0.0052 - val_acc: 0.9985\n",
            "Epoch 129/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0027 - acc: 0.9991 - val_loss: 0.0066 - val_acc: 0.9981\n",
            "Epoch 130/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0031 - acc: 0.9990 - val_loss: 0.0121 - val_acc: 0.9963\n",
            "Epoch 131/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0028 - acc: 0.9991 - val_loss: 0.0060 - val_acc: 0.9983\n",
            "Epoch 132/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0035 - acc: 0.9990 - val_loss: 0.0051 - val_acc: 0.9985\n",
            "Epoch 133/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0025 - acc: 0.9992 - val_loss: 0.0066 - val_acc: 0.9981\n",
            "Epoch 134/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0030 - acc: 0.9990 - val_loss: 0.0057 - val_acc: 0.9984\n",
            "Epoch 135/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0027 - acc: 0.9991 - val_loss: 0.0060 - val_acc: 0.9983\n",
            "Epoch 136/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0029 - acc: 0.9990 - val_loss: 0.0063 - val_acc: 0.9982\n",
            "Epoch 137/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0049 - acc: 0.9987 - val_loss: 0.0048 - val_acc: 0.9986\n",
            "Epoch 138/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0019 - acc: 0.9994 - val_loss: 0.0063 - val_acc: 0.9982\n",
            "Epoch 139/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0029 - acc: 0.9990 - val_loss: 0.0057 - val_acc: 0.9984\n",
            "Epoch 140/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0029 - acc: 0.9991 - val_loss: 0.0056 - val_acc: 0.9984\n",
            "Epoch 141/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0025 - acc: 0.9992 - val_loss: 0.0054 - val_acc: 0.9985\n",
            "Epoch 142/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0026 - acc: 0.9991 - val_loss: 0.0057 - val_acc: 0.9984\n",
            "Epoch 143/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0027 - acc: 0.9992 - val_loss: 0.0061 - val_acc: 0.9982\n",
            "Epoch 144/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0025 - acc: 0.9992 - val_loss: 0.0058 - val_acc: 0.9983\n",
            "Epoch 145/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0033 - acc: 0.9990 - val_loss: 0.0046 - val_acc: 0.9987\n",
            "Epoch 146/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0019 - acc: 0.9994 - val_loss: 0.0061 - val_acc: 0.9982\n",
            "Epoch 147/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0023 - acc: 0.9992 - val_loss: 0.0061 - val_acc: 0.9983\n",
            "Epoch 148/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0023 - acc: 0.9993 - val_loss: 0.0058 - val_acc: 0.9984\n",
            "Epoch 149/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0023 - acc: 0.9992 - val_loss: 0.0063 - val_acc: 0.9982\n",
            "Epoch 150/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0024 - acc: 0.9992 - val_loss: 0.0058 - val_acc: 0.9983\n",
            "Epoch 151/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0024 - acc: 0.9992 - val_loss: 0.0063 - val_acc: 0.9982\n",
            "Epoch 152/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0025 - acc: 0.9993 - val_loss: 0.0051 - val_acc: 0.9986\n",
            "Epoch 153/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0020 - acc: 0.9993 - val_loss: 0.0064 - val_acc: 0.9982\n",
            "Epoch 154/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0028 - acc: 0.9991 - val_loss: 0.0046 - val_acc: 0.9987\n",
            "Epoch 155/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0017 - acc: 0.9995 - val_loss: 0.0063 - val_acc: 0.9982\n",
            "Epoch 156/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0021 - acc: 0.9993 - val_loss: 0.0061 - val_acc: 0.9983\n",
            "Epoch 157/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0026 - acc: 0.9992 - val_loss: 0.0047 - val_acc: 0.9987\n",
            "Epoch 158/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0028 - acc: 0.9992 - val_loss: 0.0142 - val_acc: 0.9956\n",
            "Epoch 159/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0014 - acc: 0.9996 - val_loss: 0.0048 - val_acc: 0.9987\n",
            "Epoch 160/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0019 - acc: 0.9994 - val_loss: 0.0056 - val_acc: 0.9984\n",
            "Epoch 161/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0025 - acc: 0.9992 - val_loss: 0.0046 - val_acc: 0.9988\n",
            "Epoch 162/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0026 - acc: 0.9993 - val_loss: 0.0040 - val_acc: 0.9989\n",
            "Epoch 163/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0013 - acc: 0.9996 - val_loss: 0.0061 - val_acc: 0.9983\n",
            "Epoch 164/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0021 - acc: 0.9993 - val_loss: 0.0061 - val_acc: 0.9983\n",
            "Epoch 165/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0021 - acc: 0.9993 - val_loss: 0.0056 - val_acc: 0.9984\n",
            "Epoch 166/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0031 - acc: 0.9991 - val_loss: 0.0046 - val_acc: 0.9988\n",
            "Epoch 167/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 9.8121e-04 - acc: 0.9997 - val_loss: 0.0050 - val_acc: 0.9987\n",
            "Epoch 168/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0020 - acc: 0.9993 - val_loss: 0.0051 - val_acc: 0.9986\n",
            "Epoch 169/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0019 - acc: 0.9994 - val_loss: 0.0051 - val_acc: 0.9986\n",
            "Epoch 170/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0019 - acc: 0.9994 - val_loss: 0.0051 - val_acc: 0.9986\n",
            "Epoch 171/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0035 - acc: 0.9990 - val_loss: 0.0074 - val_acc: 0.9978\n",
            "Epoch 172/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 8.5968e-04 - acc: 0.9998 - val_loss: 0.0041 - val_acc: 0.9989\n",
            "Epoch 173/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0015 - acc: 0.9995 - val_loss: 0.0051 - val_acc: 0.9986\n",
            "Epoch 174/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0019 - acc: 0.9994 - val_loss: 0.0057 - val_acc: 0.9984\n",
            "Epoch 175/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0018 - acc: 0.9994 - val_loss: 0.0046 - val_acc: 0.9987\n",
            "Epoch 176/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0018 - acc: 0.9994 - val_loss: 0.0051 - val_acc: 0.9986\n",
            "Epoch 177/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0019 - acc: 0.9994 - val_loss: 0.0055 - val_acc: 0.9985\n",
            "Epoch 178/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0018 - acc: 0.9994 - val_loss: 0.0049 - val_acc: 0.9987\n",
            "Epoch 179/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0018 - acc: 0.9994 - val_loss: 0.0048 - val_acc: 0.9987\n",
            "Epoch 180/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0018 - acc: 0.9994 - val_loss: 0.0044 - val_acc: 0.9988\n",
            "Epoch 181/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0016 - acc: 0.9995 - val_loss: 0.0046 - val_acc: 0.9988\n",
            "Epoch 182/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0025 - acc: 0.9992 - val_loss: 0.0044 - val_acc: 0.9988\n",
            "Epoch 183/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 8.9169e-04 - acc: 0.9997 - val_loss: 0.0053 - val_acc: 0.9986\n",
            "Epoch 184/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0017 - acc: 0.9995 - val_loss: 0.0048 - val_acc: 0.9987\n",
            "Epoch 185/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0019 - acc: 0.9995 - val_loss: 0.0036 - val_acc: 0.9991\n",
            "Epoch 186/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0012 - acc: 0.9996 - val_loss: 0.0054 - val_acc: 0.9985\n",
            "Epoch 187/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0017 - acc: 0.9994 - val_loss: 0.0055 - val_acc: 0.9985\n",
            "Epoch 188/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0016 - acc: 0.9995 - val_loss: 0.0053 - val_acc: 0.9986\n",
            "Epoch 189/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0016 - acc: 0.9995 - val_loss: 0.0054 - val_acc: 0.9985\n",
            "Epoch 190/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0016 - acc: 0.9995 - val_loss: 0.0073 - val_acc: 0.9980\n",
            "Epoch 191/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0016 - acc: 0.9995 - val_loss: 0.0048 - val_acc: 0.9987\n",
            "Epoch 192/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0016 - acc: 0.9995 - val_loss: 0.0054 - val_acc: 0.9985\n",
            "Epoch 193/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0015 - acc: 0.9995 - val_loss: 0.0055 - val_acc: 0.9985\n",
            "Epoch 194/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0027 - acc: 0.9992 - val_loss: 0.0037 - val_acc: 0.9991\n",
            "Epoch 195/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 6.4662e-04 - acc: 0.9998 - val_loss: 0.0044 - val_acc: 0.9989\n",
            "Epoch 196/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0014 - acc: 0.9996 - val_loss: 0.0043 - val_acc: 0.9989\n",
            "Epoch 197/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0015 - acc: 0.9995 - val_loss: 0.0052 - val_acc: 0.9986\n",
            "Epoch 198/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0020 - acc: 0.9994 - val_loss: 0.0037 - val_acc: 0.9991\n",
            "Epoch 199/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 9.8449e-04 - acc: 0.9997 - val_loss: 0.0054 - val_acc: 0.9985\n",
            "Epoch 200/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0016 - acc: 0.9995 - val_loss: 0.0050 - val_acc: 0.9987\n",
            "Epoch 201/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0014 - acc: 0.9995 - val_loss: 0.0043 - val_acc: 0.9989\n",
            "Epoch 202/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0014 - acc: 0.9996 - val_loss: 0.0050 - val_acc: 0.9987\n",
            "Epoch 203/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0015 - acc: 0.9995 - val_loss: 0.0043 - val_acc: 0.9989\n",
            "Epoch 204/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0014 - acc: 0.9995 - val_loss: 0.0055 - val_acc: 0.9985\n",
            "Epoch 205/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0013 - acc: 0.9996 - val_loss: 0.0055 - val_acc: 0.9986\n",
            "Epoch 206/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0014 - acc: 0.9995 - val_loss: 0.0055 - val_acc: 0.9985\n",
            "Epoch 207/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0023 - acc: 0.9993 - val_loss: 0.0036 - val_acc: 0.9991\n",
            "Epoch 208/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 4.8808e-04 - acc: 0.9999 - val_loss: 0.0042 - val_acc: 0.9989\n",
            "Epoch 209/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0013 - acc: 0.9996 - val_loss: 0.0044 - val_acc: 0.9989\n",
            "Epoch 210/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0013 - acc: 0.9996 - val_loss: 0.0051 - val_acc: 0.9986\n",
            "Epoch 211/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0017 - acc: 0.9995 - val_loss: 0.0038 - val_acc: 0.9990\n",
            "Epoch 212/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0010 - acc: 0.9997 - val_loss: 0.0059 - val_acc: 0.9984\n",
            "Epoch 213/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0013 - acc: 0.9996 - val_loss: 0.0043 - val_acc: 0.9989\n",
            "Epoch 214/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0014 - acc: 0.9996 - val_loss: 0.0049 - val_acc: 0.9987\n",
            "Epoch 215/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0015 - acc: 0.9995 - val_loss: 0.0045 - val_acc: 0.9989\n",
            "Epoch 216/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0012 - acc: 0.9996 - val_loss: 0.0043 - val_acc: 0.9989\n",
            "Epoch 217/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0012 - acc: 0.9996 - val_loss: 0.0054 - val_acc: 0.9985\n",
            "Epoch 218/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0013 - acc: 0.9996 - val_loss: 0.0045 - val_acc: 0.9988\n",
            "Epoch 219/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0016 - acc: 0.9995 - val_loss: 0.0038 - val_acc: 0.9990\n",
            "Epoch 220/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0022 - acc: 0.9994 - val_loss: 0.0032 - val_acc: 0.9992\n",
            "Epoch 221/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 3.6714e-04 - acc: 0.9999 - val_loss: 0.0047 - val_acc: 0.9988\n",
            "Epoch 222/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0012 - acc: 0.9996 - val_loss: 0.0044 - val_acc: 0.9988\n",
            "Epoch 223/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0012 - acc: 0.9996 - val_loss: 0.0044 - val_acc: 0.9989\n",
            "Epoch 224/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0012 - acc: 0.9996 - val_loss: 0.0044 - val_acc: 0.9989\n",
            "Epoch 225/225\n",
            "1355/1355 [==============================] - 54s 40ms/step - loss: 0.0012 - acc: 0.9996 - val_loss: 0.0051 - val_acc: 0.9987\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# plot the model loss and accuracy\n",
        "fig, (axis1, axis2) = plt.subplots(nrows=1, ncols=2, figsize=(16,6))\n",
        "\n",
        "# summarize history for accuracy\n",
        "axis1.plot(history.history[\"acc\"], label='Train', linewidth=3)\n",
        "axis1.plot(history.history[\"val_acc\"], label='Validation', linewidth=3)\n",
        "axis1.set_title('Model accuracy', fontsize=16, color=\"white\")\n",
        "axis1.set_ylabel('accuracy')\n",
        "axis1.set_xlabel('epoch')\n",
        "axis1.legend(loc='lower right')\n",
        "\n",
        "# summarize history for loss\n",
        "axis2.plot(history.history[\"loss\"], label='Train', linewidth=3)\n",
        "axis2.plot(history.history[\"val_loss\"], label='Validation', linewidth=3)\n",
        "axis2.set_title('Model loss', fontsize=16, color=\"white\")\n",
        "axis2.set_ylabel('loss')\n",
        "axis2.set_xlabel('epoch')\n",
        "axis2.legend(loc='upper right')\n",
        "plt.savefig(\"TrainValAccuracyLoss.jpg\",\tformat=\"jpg\",\tdpi=1200,bbox_inches='tight')\n",
        "plt.show()"
      ],
      "metadata": {
        "id": "TF2wPYHbiKiN",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 366
        },
        "outputId": "3c7b6767-01df-4cc3-c3df-d5735708350a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1152x432 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "with open('trainHistoryDict', 'wb') as file_pi:\n",
        "  pickle.dump(history.history, file_pi)"
      ],
      "metadata": {
        "id": "ZRPaI1WJampR"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# evaluate our test set\n",
        "score, acc = model.evaluate([X_test, X_test], y_test, batch_size=batch_size, verbose=0)\n",
        "print(f'Test score: {score}')\n",
        "print(f'Test accuracy: {acc}')"
      ],
      "metadata": {
        "id": "akY0LcybiNPS"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# save our baseline training model & weights\n",
        "mod_file = 'LSTM_model.h5'\n",
        "model.save(mod_file)\n",
        "print(\"Model Saved\")"
      ],
      "metadata": {
        "id": "4ZVx9-OqiOXm"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import json\n",
        "\n",
        "with open(\"char_to_int.json\", \"w\") as outfile:\n",
        "    json.dump(char_to_int, outfile)\n",
        "with open(\"int_to_char.json\", \"w\") as outfile:\n",
        "    json.dump(int_to_char, outfile)"
      ],
      "metadata": {
        "id": "ZgPklb4FguVA"
      },
      "execution_count": null,
      "outputs": []
    }
  ]
}