7. Recurrent Neural Networks II.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "7. Recurrent Neural Networks II.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/DataScienceUB/DeepLearningMaster2019/blob/master/7.%20Recurrent%20Neural%20Networks%20II.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "metadata": {
        "id": "dyuShyc1I_Kn",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# ``seq2seq`` architecture\n",
        "\n",
        "Source: https://github.com/keras-team/keras/blob/master/examples/addition_rnn.py \n",
        "\n",
        "The objective of this code is to train a RNN for adding a sequence of integers. In this case we must process all the sequence in order to produce a result. We will use a ``keras`` implementation of the ``seq2seq`` architecure, originally proposed by Sutskever, Vinyals and Le in 2014. \n",
        "\n",
        "The architecture diagram from their paper is:\n",
        "\n",
        "<center>\n",
        "<img src=\"https://github.com/DataScienceUB/DeepLearningMaster2019/blob/master/images/seq2seq.png?raw=1\" alt=\"\" style=\"width: 800px;\"/> \n",
        "Source: https://arxiv.org/abs/1409.3215\n",
        "</center>\n",
        "\n",
        "Rectangles are recurrent layers. Encoder receives ``[A, B, C]`` sequence as inputs. **We don't care about encoder outputs, only about the hidden state it accumulates while reading the sequence**. \n",
        "\n",
        "After input sequence ends, encoder passes its final state to decoder, which receives ``[<EOS>, W, X, Y, Z]`` and is trained to output ``[W, X, Y, Z, <EOS>]``. "
      ]
    },
    {
      "metadata": {
        "id": "hdoFETSNI_Kp",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Sequence-to-sequence, or *Seq2Seq*, is a paradigm made up of **two recurrent networks**: an *encoder* which takes the model's input sequence as input and encodes it into a fixed-size *context vector*, and a *decoder*, which uses the context vector from above as a *seed* from which to generate an output sequence.\n",
        "\n",
        "> For this reason, Seq2Seq models are often referred to as *encoder-decoder* models.\n",
        "\n",
        "## Encoder\n",
        "\n",
        "The encoder network's job is to read the input sequence and generate a fixed-dimensional context vector $C$ for the sequence. \n",
        "\n",
        "To do so, the encoder will use a recurrent neural network cell -- usually an LSTM -- to read the input tokens one at a time. The final hidden state of the cell will then become $C$. \n",
        "\n",
        "However, because it's so difficult to compress an arbitrary-length sequence into a single fixed-size vector, the encoder will usually consist of *stacked* LSTMs. The *final* layer's LSTM hidden state will be used as $C$.\n",
        "\n",
        "## Decoder\n",
        "\n",
        "The decoder is also an LSTM network, but its usage is a little more complex than the encoder network. Essentially, we'd like to use it as a **language model that's \"aware\" of the words that it's generated so far *and* of the input**. \n",
        "\n",
        "To that end, we'll keep the \"stacked\" LSTM architecture from the encoder, but we'll initialize the hidden state of our first layer with the context vector from above; the decoder will literally use the context of the input to generate an output."
      ]
    },
    {
      "metadata": {
        "id": "Uku9q1HoI_Kq",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## One or two RNN?\n",
        "\n",
        "When both input sequences and output sequences have the same length, you can implement such models simply with one Keras LSTM or GRU layer (or stack thereof).\n",
        "\n",
        "One caveat of this approach is that it assumes that it is possible to generate ``target[...t]`` given ``input[...t]``. That works in some cases but does not work for most use cases. In the general case, information about the entire input sequence is necessary in order to start generating the target sequence.\n",
        "\n"
      ]
    },
    {
      "metadata": {
        "id": "mxKIZh8dYTfS",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Example"
      ]
    },
    {
      "metadata": {
        "id": "80uzbRKYI_Kr",
        "colab_type": "code",
        "outputId": "00b2cf65-44c4-407b-dc07-914e81d392ec",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 257
        }
      },
      "cell_type": "code",
      "source": [
        "'''An implementation of sequence to sequence learning for performing addition\n",
        "Input: \"535+61 \"\n",
        "Output:\"596 \"\n",
        "Padding is handled by using a repeated space character\n",
        "'''\n",
        "\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras import layers\n",
        "import numpy as np\n",
        "from six.moves import range\n",
        "\n",
        "\n",
        "class CharacterTable(object):\n",
        "  \n",
        "    \"\"\"Given a set of characters:\n",
        "    + Encode them to a one hot integer representation\n",
        "    + Decode the one hot integer representation to their character output\n",
        "    + Decode a vector of probabilities to their character output\n",
        "    \"\"\"\n",
        "    \n",
        "    def __init__(self, chars):\n",
        "        \"\"\"Initialize character table.\n",
        "        # Arguments\n",
        "            chars: Characters that can appear in the input.\n",
        "        \"\"\"\n",
        "        self.chars        = sorted(set(chars))\n",
        "        self.char_indices = dict((c, i) for i, c in enumerate(self.chars))\n",
        "        self.indices_char = dict((i, c) for i, c in enumerate(self.chars))\n",
        "\n",
        "    def encode(self, C, num_rows):\n",
        "        \"\"\"One hot encode given string C.\n",
        "        # Arguments\n",
        "            num_rows: Number of rows in the returned one hot encoding. This is\n",
        "                used to keep the # of rows for each data the same.\n",
        "        \"\"\"\n",
        "        x = np.zeros((num_rows, len(self.chars)))\n",
        "        for i, c in enumerate(C):\n",
        "            x[i, self.char_indices[c]] = 1\n",
        "        return x\n",
        "\n",
        "    def decode(self, x, calc_argmax=True):\n",
        "        if calc_argmax:\n",
        "            x = x.argmax(axis=-1)\n",
        "        return ''.join(self.indices_char[x] for x in x)\n",
        "\n",
        "\n",
        "class colors:\n",
        "    ok = '\\033[92m'\n",
        "    fail = '\\033[91m'\n",
        "    close = '\\033[0m'\n",
        "\n",
        "# Parameters for the model and dataset.\n",
        "TRAINING_SIZE = 50000\n",
        "DIGITS = 3\n",
        "REVERSE = True\n",
        "\n",
        "\n",
        "# Maximum length of input is 'int + int' (e.g., '345+678'). \n",
        "# Maximum length of int is DIGITS.\n",
        "MAXLEN = DIGITS + 1 + DIGITS\n",
        "\n",
        "# All the numbers, plus sign and space for padding.\n",
        "chars = '0123456789+ '\n",
        "ctable = CharacterTable(chars)\n",
        "\n",
        "questions = []\n",
        "expected = []\n",
        "seen = set()\n",
        "print('Generating data...','\\n')\n",
        "\n",
        "while len(questions) < TRAINING_SIZE:\n",
        "    f = lambda: int(''.join(np.random.choice(list('0123456789'))\n",
        "                    for i in range(np.random.randint(1, DIGITS + 1))))\n",
        "    a, b = f(), f()\n",
        "    # Skip any addition questions we've already seen\n",
        "    # Also skip any such that x+Y == Y+x (hence the sorting).\n",
        "    key = tuple(sorted((a, b)))\n",
        "    if key in seen:\n",
        "        continue\n",
        "    seen.add(key)\n",
        "    # Pad the data with spaces such that it is always MAXLEN.\n",
        "    q = '{}+{}'.format(a, b)\n",
        "    query = q + ' ' * (MAXLEN - len(q))\n",
        "    ans = str(a + b)\n",
        "    # Answers can be of maximum size DIGITS + 1.\n",
        "    ans += ' ' * (DIGITS + 1 - len(ans))\n",
        "    questions.append(query)\n",
        "    expected.append(ans)\n",
        "print('Total questions: ', len(questions),'\\n')\n",
        "\n",
        "for i in range(10):\n",
        "  n = np.random.randint(len(questions))\n",
        "  print('Random example :', questions[n], expected[n])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Generating data... \n",
            "\n",
            "Total questions:  50000 \n",
            "\n",
            "Random example : 755+38  793 \n",
            "Random example : 728+856 1584\n",
            "Random example : 392+85  477 \n",
            "Random example : 27+869  896 \n",
            "Random example : 89+258  347 \n",
            "Random example : 452+983 1435\n",
            "Random example : 678+77  755 \n",
            "Random example : 859+794 1653\n",
            "Random example : 29+26   55  \n",
            "Random example : 10+286  296 \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "PYkp-sR-I_Kw",
        "colab_type": "code",
        "outputId": "5bbae6fe-4298-4337-fc64-ca3ab99bd015",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 86
        }
      },
      "cell_type": "code",
      "source": [
        "print('One hot encoding... \\n')\n",
        "x = np.zeros((len(questions), MAXLEN, len(chars)), dtype=np.bool)\n",
        "y = np.zeros((len(questions), DIGITS + 1, len(chars)), dtype=np.bool)\n",
        "for i, sentence in enumerate(questions):\n",
        "    x[i] = ctable.encode(sentence, MAXLEN)\n",
        "for i, sentence in enumerate(expected):\n",
        "    y[i] = ctable.encode(sentence, DIGITS + 1)\n",
        "\n",
        "\n",
        "# Shuffle (x, y) \n",
        "indices = np.arange(len(y))\n",
        "np.random.shuffle(indices)\n",
        "x = x[indices]\n",
        "y = y[indices]\n",
        "\n",
        "# Explicitly set apart 10% for validation data that we never train over.\n",
        "split_at = len(x) - len(x) // 10\n",
        "(x_train, x_val) = x[:split_at], x[split_at:]\n",
        "(y_train, y_val) = y[:split_at], y[split_at:]\n",
        "\n",
        "print('Training Data  : '  , end=\"\")\n",
        "print(x_train.shape, end=\"\")\n",
        "print(y_train.shape)\n",
        "\n",
        "print('Validation Data: ', end=\"\")\n",
        "print(x_val.shape, end=\"\")\n",
        "print(y_val.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "One hot encoding... \n",
            "\n",
            "Training Data  : (45000, 7, 12)(45000, 4, 12)\n",
            "Validation Data: (5000, 7, 12)(5000, 4, 12)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "yAD-Ac3eYxZ3",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "We will use some utility layers:\n",
        "+ `RepeatVector` : it repeats its input `n` times.\n",
        "\n",
        "```\n",
        "   model = Sequential()\n",
        "   model.add(Dense(32, input_dim=32))\n",
        "   # now: model.output_shape == (None, 32)\n",
        "\n",
        "   model.add(RepeatVector(3))\n",
        "   # now: model.output_shape == (None, 3, 32)\n",
        "```\n",
        "\n",
        "+ `TimeDistributed`: In keras - while building a sequential model - usually the second dimension (one after sample dimension) - is related to a time dimension. This layer applies a layer to every temporal slice of an input.\n",
        "The input should be at least 3D, and the dimension of index one will be considered to be the temporal dimension.\n",
        "Consider a batch of 32 samples, where each sample is a sequence of 10 vectors of 16 dimensions. The batch input shape of the layer is then (32, 10, 16), and the input_shape, not including the samples dimension, is (10, 16).\n",
        "You can then use TimeDistributed to apply a Dense layer to each of the 10 time steps, independently:\n",
        "\n",
        "```\n",
        "    # as the first layer in a model\n",
        "    model = Sequential()\n",
        "    model.add(TimeDistributed(Dense(8), input_shape=(10, 16)))\n",
        "    # now model.output_shape == (None, 10, 8)\n",
        "```"
      ]
    },
    {
      "metadata": {
        "id": "YWkNSDa4O8qy",
        "colab_type": "code",
        "outputId": "053afd91-4a98-422f-c085-34d9d8d9a88e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51671
        }
      },
      "cell_type": "code",
      "source": [
        "# Try replacing GRU, or SimpleRNN.\n",
        "\n",
        "# keras.layers.LSTM(units, \n",
        "#                   activation='tanh', \n",
        "#                   use_bias=True, \n",
        "#                   recurrent_initializer='orthogonal', \n",
        "#                   bias_initializer='zeros', \n",
        "#                   ...\n",
        "#                   return_sequences=False, \n",
        "#                   return_state=False, \n",
        "#                   stateful=False)\n",
        "\n",
        "\n",
        "RNN = layers.LSTM\n",
        "HIDDEN_SIZE = 128\n",
        "BATCH_SIZE = 128\n",
        "LAYERS = 1\n",
        "\n",
        "print('Build model...')\n",
        "model = Sequential()\n",
        "\n",
        "# \"Encode\" the input sequence using an RNN, producing an output of HIDDEN_SIZE.\n",
        "# Note: In a situation where your input sequences have a variable length,\n",
        "# you can use input_shape=(None, num_feature).\n",
        "# Returns the last output in the output sequence\n",
        "model.add(RNN(HIDDEN_SIZE, input_shape=(MAXLEN, len(chars))))\n",
        "\n",
        "# As the decoder RNN's input, repeatedly provide with the last hidden state of\n",
        "# RNN for each time step. Repeat 'DIGITS + 1' times as that's the maximum\n",
        "# length of output, e.g., when DIGITS=3, max output is 999+999=1998.\n",
        "model.add(layers.RepeatVector(DIGITS + 1))\n",
        "\n",
        "# The decoder RNN could be multiple layers stacked or a single layer.\n",
        "for _ in range(LAYERS):\n",
        "    # By setting return_sequences to True, return not only the last output but\n",
        "    # all the outputs so far in the form of (num_samples, timesteps,\n",
        "    # output_dim). This is necessary as TimeDistributed in the below expects\n",
        "    # the first dimension to be the timesteps.\n",
        "    model.add(RNN(HIDDEN_SIZE, return_sequences=True))\n",
        "\n",
        "# Apply a dense layer to the every temporal slice of an input. For each of step\n",
        "# of the output sequence, decide which character should be chosen.\n",
        "model.add(layers.TimeDistributed(layers.Dense(len(chars))))\n",
        "model.add(layers.Activation('softmax'))\n",
        "model.compile(loss='categorical_crossentropy',\n",
        "              optimizer='adam',\n",
        "              metrics=['accuracy'])\n",
        "model.summary()\n",
        "\n",
        "# Train the model each generation and show predictions against the validation\n",
        "# dataset.\n",
        "for iteration in range(1, 200):\n",
        "    print()\n",
        "    print('-' * 50)\n",
        "    print('Iteration', iteration)\n",
        "    model.fit(x_train, y_train,\n",
        "              batch_size=BATCH_SIZE,\n",
        "              epochs=1,\n",
        "              validation_data=(x_val, y_val))\n",
        "    # Select 10 samples from the validation set at random so we can visualize\n",
        "    # errors.\n",
        "    for i in range(10):\n",
        "        ind = np.random.randint(0, len(x_val))\n",
        "        rowx, rowy = x_val[np.array([ind])], y_val[np.array([ind])]\n",
        "        preds = model.predict_classes(rowx, verbose=0)\n",
        "        q = ctable.decode(rowx[0])\n",
        "        correct = ctable.decode(rowy[0])\n",
        "        guess = ctable.decode(preds[0], calc_argmax=False)\n",
        "        print('Q', q, end=' ')\n",
        "        print('T', correct, end=' ')\n",
        "        if correct == guess:\n",
        "            print(colors.ok + '☑' + colors.close, end=' ')\n",
        "        else:\n",
        "            print(colors.fail + '☒' + colors.close, end=' ')\n",
        "        print(guess)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Build model...\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/resource_variable_ops.py:435: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Colocations handled automatically by placer.\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "lstm (LSTM)                  (None, 128)               72192     \n",
            "_________________________________________________________________\n",
            "repeat_vector (RepeatVector) (None, 4, 128)            0         \n",
            "_________________________________________________________________\n",
            "lstm_1 (LSTM)                (None, 4, 128)            131584    \n",
            "_________________________________________________________________\n",
            "time_distributed (TimeDistri (None, 4, 12)             1548      \n",
            "_________________________________________________________________\n",
            "activation (Activation)      (None, 4, 12)             0         \n",
            "=================================================================\n",
            "Total params: 205,324\n",
            "Trainable params: 205,324\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 1\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.cast instead.\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_grad.py:102: div (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Deprecated in favor of operator or tf.math.divide.\n",
            "45000/45000 [==============================] - 16s 361us/sample - loss: 1.8892 - acc: 0.3218 - val_loss: 1.8004 - val_acc: 0.3410\n",
            "Q 228+79  T 307  \u001b[91m☒\u001b[0m 121 \n",
            "Q 933+405 T 1338 \u001b[91m☒\u001b[0m 105 \n",
            "Q 398+40  T 438  \u001b[91m☒\u001b[0m 105 \n",
            "Q 326+59  T 385  \u001b[91m☒\u001b[0m 121 \n",
            "Q 990+368 T 1358 \u001b[91m☒\u001b[0m 1105\n",
            "Q 988+3   T 991  \u001b[91m☒\u001b[0m 100 \n",
            "Q 447+1   T 448  \u001b[91m☒\u001b[0m 12  \n",
            "Q 889+180 T 1069 \u001b[91m☒\u001b[0m 1105\n",
            "Q 198+73  T 271  \u001b[91m☒\u001b[0m 105 \n",
            "Q 604+36  T 640  \u001b[91m☒\u001b[0m 125 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 2\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 323us/sample - loss: 1.7683 - acc: 0.3465 - val_loss: 1.7257 - val_acc: 0.3586\n",
            "Q 622+825 T 1447 \u001b[91m☒\u001b[0m 101 \n",
            "Q 946+137 T 1083 \u001b[91m☒\u001b[0m 101 \n",
            "Q 46+809  T 855  \u001b[91m☒\u001b[0m 101 \n",
            "Q 562+145 T 707  \u001b[91m☒\u001b[0m 101 \n",
            "Q 641+381 T 1022 \u001b[91m☒\u001b[0m 101 \n",
            "Q 896+34  T 930  \u001b[91m☒\u001b[0m 105 \n",
            "Q 474+32  T 506  \u001b[91m☒\u001b[0m 306 \n",
            "Q 174+859 T 1033 \u001b[91m☒\u001b[0m 101 \n",
            "Q 65+26   T 91   \u001b[91m☒\u001b[0m 56  \n",
            "Q 175+407 T 582  \u001b[91m☒\u001b[0m 101 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 3\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 1.6386 - acc: 0.3905 - val_loss: 1.5599 - val_acc: 0.4172\n",
            "Q 213+603 T 816  \u001b[91m☒\u001b[0m 300 \n",
            "Q 207+77  T 284  \u001b[91m☒\u001b[0m 377 \n",
            "Q 12+305  T 317  \u001b[91m☒\u001b[0m 232 \n",
            "Q 757+79  T 836  \u001b[91m☒\u001b[0m 744 \n",
            "Q 213+0   T 213  \u001b[91m☒\u001b[0m 22  \n",
            "Q 69+937  T 1006 \u001b[91m☒\u001b[0m 900 \n",
            "Q 537+25  T 562  \u001b[91m☒\u001b[0m 497 \n",
            "Q 97+722  T 819  \u001b[91m☒\u001b[0m 800 \n",
            "Q 327+94  T 421  \u001b[91m☒\u001b[0m 300 \n",
            "Q 487+821 T 1308 \u001b[91m☒\u001b[0m 1207\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 4\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 317us/sample - loss: 1.5036 - acc: 0.4357 - val_loss: 1.4465 - val_acc: 0.4544\n",
            "Q 54+378  T 432  \u001b[91m☒\u001b[0m 522 \n",
            "Q 974+41  T 1015 \u001b[91m☒\u001b[0m 1012\n",
            "Q 563+900 T 1463 \u001b[91m☒\u001b[0m 1410\n",
            "Q 27+484  T 511  \u001b[91m☒\u001b[0m 410 \n",
            "Q 94+145  T 239  \u001b[91m☒\u001b[0m 442 \n",
            "Q 482+190 T 672  \u001b[91m☒\u001b[0m 522 \n",
            "Q 379+1   T 380  \u001b[91m☒\u001b[0m 10  \n",
            "Q 415+179 T 594  \u001b[91m☒\u001b[0m 559 \n",
            "Q 74+678  T 752  \u001b[91m☒\u001b[0m 741 \n",
            "Q 912+674 T 1586 \u001b[91m☒\u001b[0m 1035\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 5\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 1.3859 - acc: 0.4837 - val_loss: 1.3306 - val_acc: 0.5037\n",
            "Q 828+0   T 828  \u001b[91m☒\u001b[0m 829 \n",
            "Q 609+71  T 680  \u001b[91m☒\u001b[0m 777 \n",
            "Q 204+63  T 267  \u001b[91m☒\u001b[0m 294 \n",
            "Q 180+995 T 1175 \u001b[91m☒\u001b[0m 1100\n",
            "Q 364+94  T 458  \u001b[91m☒\u001b[0m 422 \n",
            "Q 757+58  T 815  \u001b[91m☒\u001b[0m 734 \n",
            "Q 397+83  T 480  \u001b[91m☒\u001b[0m 415 \n",
            "Q 31+573  T 604  \u001b[91m☒\u001b[0m 675 \n",
            "Q 288+52  T 340  \u001b[91m☒\u001b[0m 315 \n",
            "Q 10+371  T 381  \u001b[91m☒\u001b[0m 191 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 6\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 322us/sample - loss: 1.2601 - acc: 0.5321 - val_loss: 1.2070 - val_acc: 0.5516\n",
            "Q 373+18  T 391  \u001b[91m☒\u001b[0m 381 \n",
            "Q 327+18  T 345  \u001b[91m☒\u001b[0m 281 \n",
            "Q 495+799 T 1294 \u001b[91m☒\u001b[0m 1377\n",
            "Q 45+921  T 966  \u001b[91m☒\u001b[0m 977 \n",
            "Q 893+869 T 1762 \u001b[91m☒\u001b[0m 1765\n",
            "Q 179+2   T 181  \u001b[91m☒\u001b[0m 199 \n",
            "Q 227+91  T 318  \u001b[91m☒\u001b[0m 301 \n",
            "Q 704+80  T 784  \u001b[91m☒\u001b[0m 888 \n",
            "Q 107+845 T 952  \u001b[91m☒\u001b[0m 997 \n",
            "Q 777+75  T 852  \u001b[91m☒\u001b[0m 845 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 7\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 320us/sample - loss: 1.1522 - acc: 0.5692 - val_loss: 1.1060 - val_acc: 0.5832\n",
            "Q 954+0   T 954  \u001b[91m☒\u001b[0m 955 \n",
            "Q 303+29  T 332  \u001b[91m☒\u001b[0m 331 \n",
            "Q 984+724 T 1708 \u001b[91m☒\u001b[0m 1828\n",
            "Q 3+913   T 916  \u001b[91m☒\u001b[0m 923 \n",
            "Q 930+64  T 994  \u001b[91m☒\u001b[0m 988 \n",
            "Q 27+519  T 546  \u001b[91m☒\u001b[0m 539 \n",
            "Q 696+676 T 1372 \u001b[91m☒\u001b[0m 1333\n",
            "Q 660+75  T 735  \u001b[91m☒\u001b[0m 722 \n",
            "Q 53+828  T 881  \u001b[91m☒\u001b[0m 872 \n",
            "Q 958+315 T 1273 \u001b[91m☒\u001b[0m 1205\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 8\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 312us/sample - loss: 1.0449 - acc: 0.6056 - val_loss: 0.9940 - val_acc: 0.6266\n",
            "Q 73+164  T 237  \u001b[91m☒\u001b[0m 230 \n",
            "Q 824+79  T 903  \u001b[91m☒\u001b[0m 810 \n",
            "Q 20+23   T 43   \u001b[91m☒\u001b[0m 33  \n",
            "Q 23+861  T 884  \u001b[91m☒\u001b[0m 895 \n",
            "Q 738+29  T 767  \u001b[91m☒\u001b[0m 769 \n",
            "Q 834+792 T 1626 \u001b[91m☒\u001b[0m 1624\n",
            "Q 950+4   T 954  \u001b[91m☒\u001b[0m 956 \n",
            "Q 34+42   T 76   \u001b[91m☒\u001b[0m 66  \n",
            "Q 887+64  T 951  \u001b[91m☒\u001b[0m 942 \n",
            "Q 295+471 T 766  \u001b[91m☒\u001b[0m 765 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 9\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 321us/sample - loss: 0.9186 - acc: 0.6496 - val_loss: 0.8491 - val_acc: 0.6722\n",
            "Q 75+304  T 379  \u001b[91m☒\u001b[0m 390 \n",
            "Q 838+400 T 1238 \u001b[91m☒\u001b[0m 1111\n",
            "Q 75+946  T 1021 \u001b[91m☒\u001b[0m 1011\n",
            "Q 593+84  T 677  \u001b[91m☒\u001b[0m 667 \n",
            "Q 440+6   T 446  \u001b[91m☒\u001b[0m 447 \n",
            "Q 313+51  T 364  \u001b[92m☑\u001b[0m 364 \n",
            "Q 83+865  T 948  \u001b[92m☑\u001b[0m 948 \n",
            "Q 11+616  T 627  \u001b[91m☒\u001b[0m 632 \n",
            "Q 550+808 T 1358 \u001b[91m☒\u001b[0m 1370\n",
            "Q 148+273 T 421  \u001b[91m☒\u001b[0m 509 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 10\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 321us/sample - loss: 0.7423 - acc: 0.7232 - val_loss: 0.6707 - val_acc: 0.7537\n",
            "Q 61+665  T 726  \u001b[91m☒\u001b[0m 725 \n",
            "Q 415+833 T 1248 \u001b[91m☒\u001b[0m 1267\n",
            "Q 401+89  T 490  \u001b[91m☒\u001b[0m 400 \n",
            "Q 551+6   T 557  \u001b[91m☒\u001b[0m 566 \n",
            "Q 38+40   T 78   \u001b[91m☒\u001b[0m 88  \n",
            "Q 587+50  T 637  \u001b[91m☒\u001b[0m 628 \n",
            "Q 47+345  T 392  \u001b[92m☑\u001b[0m 392 \n",
            "Q 719+5   T 724  \u001b[92m☑\u001b[0m 724 \n",
            "Q 37+253  T 290  \u001b[91m☒\u001b[0m 291 \n",
            "Q 5+801   T 806  \u001b[92m☑\u001b[0m 806 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 11\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 325us/sample - loss: 0.5802 - acc: 0.7991 - val_loss: 0.5301 - val_acc: 0.8184\n",
            "Q 651+1   T 652  \u001b[92m☑\u001b[0m 652 \n",
            "Q 586+18  T 604  \u001b[92m☑\u001b[0m 604 \n",
            "Q 231+11  T 242  \u001b[92m☑\u001b[0m 242 \n",
            "Q 18+239  T 257  \u001b[92m☑\u001b[0m 257 \n",
            "Q 16+24   T 40   \u001b[91m☒\u001b[0m 49  \n",
            "Q 774+65  T 839  \u001b[91m☒\u001b[0m 849 \n",
            "Q 426+275 T 701  \u001b[91m☒\u001b[0m 711 \n",
            "Q 14+83   T 97   \u001b[91m☒\u001b[0m 98  \n",
            "Q 170+872 T 1042 \u001b[91m☒\u001b[0m 1012\n",
            "Q 643+88  T 731  \u001b[91m☒\u001b[0m 732 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 12\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 0.4612 - acc: 0.8523 - val_loss: 0.4225 - val_acc: 0.8644\n",
            "Q 63+93   T 156  \u001b[92m☑\u001b[0m 156 \n",
            "Q 590+577 T 1167 \u001b[92m☑\u001b[0m 1167\n",
            "Q 8+618   T 626  \u001b[92m☑\u001b[0m 626 \n",
            "Q 71+77   T 148  \u001b[91m☒\u001b[0m 158 \n",
            "Q 421+421 T 842  \u001b[92m☑\u001b[0m 842 \n",
            "Q 41+940  T 981  \u001b[92m☑\u001b[0m 981 \n",
            "Q 819+767 T 1586 \u001b[92m☑\u001b[0m 1586\n",
            "Q 467+193 T 660  \u001b[91m☒\u001b[0m 761 \n",
            "Q 921+516 T 1437 \u001b[91m☒\u001b[0m 1447\n",
            "Q 0+175   T 175  \u001b[92m☑\u001b[0m 175 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 13\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 330us/sample - loss: 0.3656 - acc: 0.8920 - val_loss: 0.3369 - val_acc: 0.8998\n",
            "Q 40+215  T 255  \u001b[92m☑\u001b[0m 255 \n",
            "Q 320+517 T 837  \u001b[91m☒\u001b[0m 847 \n",
            "Q 496+92  T 588  \u001b[92m☑\u001b[0m 588 \n",
            "Q 11+439  T 450  \u001b[92m☑\u001b[0m 450 \n",
            "Q 85+909  T 994  \u001b[92m☑\u001b[0m 994 \n",
            "Q 931+850 T 1781 \u001b[92m☑\u001b[0m 1781\n",
            "Q 109+964 T 1073 \u001b[91m☒\u001b[0m 1053\n",
            "Q 82+942  T 1024 \u001b[92m☑\u001b[0m 1024\n",
            "Q 326+6   T 332  \u001b[92m☑\u001b[0m 332 \n",
            "Q 18+466  T 484  \u001b[92m☑\u001b[0m 484 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 14\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 321us/sample - loss: 0.2951 - acc: 0.9178 - val_loss: 0.2848 - val_acc: 0.9149\n",
            "Q 520+35  T 555  \u001b[92m☑\u001b[0m 555 \n",
            "Q 505+427 T 932  \u001b[91m☒\u001b[0m 942 \n",
            "Q 2+914   T 916  \u001b[92m☑\u001b[0m 916 \n",
            "Q 42+575  T 617  \u001b[92m☑\u001b[0m 617 \n",
            "Q 975+27  T 1002 \u001b[92m☑\u001b[0m 1002\n",
            "Q 773+67  T 840  \u001b[92m☑\u001b[0m 840 \n",
            "Q 635+891 T 1526 \u001b[91m☒\u001b[0m 1535\n",
            "Q 798+907 T 1705 \u001b[91m☒\u001b[0m 1795\n",
            "Q 239+89  T 328  \u001b[92m☑\u001b[0m 328 \n",
            "Q 14+807  T 821  \u001b[91m☒\u001b[0m 811 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 15\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 0.2404 - acc: 0.9359 - val_loss: 0.2284 - val_acc: 0.9359\n",
            "Q 190+72  T 262  \u001b[91m☒\u001b[0m 252 \n",
            "Q 38+93   T 131  \u001b[92m☑\u001b[0m 131 \n",
            "Q 587+13  T 600  \u001b[91m☒\u001b[0m 690 \n",
            "Q 850+150 T 1000 \u001b[91m☒\u001b[0m 9000\n",
            "Q 99+80   T 179  \u001b[92m☑\u001b[0m 179 \n",
            "Q 36+78   T 114  \u001b[92m☑\u001b[0m 114 \n",
            "Q 803+65  T 868  \u001b[92m☑\u001b[0m 868 \n",
            "Q 327+95  T 422  \u001b[92m☑\u001b[0m 422 \n",
            "Q 59+54   T 113  \u001b[92m☑\u001b[0m 113 \n",
            "Q 635+323 T 958  \u001b[91m☒\u001b[0m 968 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 16\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 326us/sample - loss: 0.2023 - acc: 0.9473 - val_loss: 0.2092 - val_acc: 0.9374\n",
            "Q 100+0   T 100  \u001b[92m☑\u001b[0m 100 \n",
            "Q 2+874   T 876  \u001b[92m☑\u001b[0m 876 \n",
            "Q 516+208 T 724  \u001b[91m☒\u001b[0m 624 \n",
            "Q 15+277  T 292  \u001b[92m☑\u001b[0m 292 \n",
            "Q 805+298 T 1103 \u001b[91m☒\u001b[0m 1003\n",
            "Q 720+18  T 738  \u001b[92m☑\u001b[0m 738 \n",
            "Q 54+865  T 919  \u001b[92m☑\u001b[0m 919 \n",
            "Q 53+0    T 53   \u001b[92m☑\u001b[0m 53  \n",
            "Q 869+8   T 877  \u001b[92m☑\u001b[0m 877 \n",
            "Q 97+78   T 175  \u001b[92m☑\u001b[0m 175 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 17\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 325us/sample - loss: 0.1666 - acc: 0.9589 - val_loss: 0.1699 - val_acc: 0.9532\n",
            "Q 34+97   T 131  \u001b[92m☑\u001b[0m 131 \n",
            "Q 11+485  T 496  \u001b[92m☑\u001b[0m 496 \n",
            "Q 7+55    T 62   \u001b[92m☑\u001b[0m 62  \n",
            "Q 455+90  T 545  \u001b[92m☑\u001b[0m 545 \n",
            "Q 123+5   T 128  \u001b[92m☑\u001b[0m 128 \n",
            "Q 44+484  T 528  \u001b[92m☑\u001b[0m 528 \n",
            "Q 469+18  T 487  \u001b[92m☑\u001b[0m 487 \n",
            "Q 896+9   T 905  \u001b[92m☑\u001b[0m 905 \n",
            "Q 84+328  T 412  \u001b[92m☑\u001b[0m 412 \n",
            "Q 63+515  T 578  \u001b[92m☑\u001b[0m 578 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 18\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 0.1418 - acc: 0.9655 - val_loss: 0.1618 - val_acc: 0.9540\n",
            "Q 546+841 T 1387 \u001b[92m☑\u001b[0m 1387\n",
            "Q 32+314  T 346  \u001b[92m☑\u001b[0m 346 \n",
            "Q 875+259 T 1134 \u001b[92m☑\u001b[0m 1134\n",
            "Q 45+171  T 216  \u001b[92m☑\u001b[0m 216 \n",
            "Q 48+69   T 117  \u001b[92m☑\u001b[0m 117 \n",
            "Q 46+273  T 319  \u001b[92m☑\u001b[0m 319 \n",
            "Q 742+195 T 937  \u001b[92m☑\u001b[0m 937 \n",
            "Q 27+68   T 95   \u001b[92m☑\u001b[0m 95  \n",
            "Q 712+5   T 717  \u001b[92m☑\u001b[0m 717 \n",
            "Q 898+19  T 917  \u001b[92m☑\u001b[0m 917 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 19\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 0.1234 - acc: 0.9698 - val_loss: 0.1251 - val_acc: 0.9674\n",
            "Q 558+90  T 648  \u001b[92m☑\u001b[0m 648 \n",
            "Q 6+242   T 248  \u001b[92m☑\u001b[0m 248 \n",
            "Q 91+283  T 374  \u001b[92m☑\u001b[0m 374 \n",
            "Q 537+851 T 1388 \u001b[92m☑\u001b[0m 1388\n",
            "Q 523+281 T 804  \u001b[91m☒\u001b[0m 814 \n",
            "Q 160+310 T 470  \u001b[91m☒\u001b[0m 460 \n",
            "Q 82+464  T 546  \u001b[92m☑\u001b[0m 546 \n",
            "Q 730+992 T 1722 \u001b[91m☒\u001b[0m 1732\n",
            "Q 215+617 T 832  \u001b[92m☑\u001b[0m 832 \n",
            "Q 121+752 T 873  \u001b[92m☑\u001b[0m 873 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 20\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 0.1073 - acc: 0.9744 - val_loss: 0.1428 - val_acc: 0.9545\n",
            "Q 84+437  T 521  \u001b[92m☑\u001b[0m 521 \n",
            "Q 3+622   T 625  \u001b[92m☑\u001b[0m 625 \n",
            "Q 67+686  T 753  \u001b[92m☑\u001b[0m 753 \n",
            "Q 752+17  T 769  \u001b[91m☒\u001b[0m 768 \n",
            "Q 80+368  T 448  \u001b[92m☑\u001b[0m 448 \n",
            "Q 70+12   T 82   \u001b[92m☑\u001b[0m 82  \n",
            "Q 346+842 T 1188 \u001b[91m☒\u001b[0m 1187\n",
            "Q 34+489  T 523  \u001b[92m☑\u001b[0m 523 \n",
            "Q 592+569 T 1161 \u001b[92m☑\u001b[0m 1161\n",
            "Q 135+130 T 265  \u001b[91m☒\u001b[0m 365 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 21\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 329us/sample - loss: 0.1084 - acc: 0.9715 - val_loss: 0.1015 - val_acc: 0.9729\n",
            "Q 557+477 T 1034 \u001b[92m☑\u001b[0m 1034\n",
            "Q 7+501   T 508  \u001b[92m☑\u001b[0m 508 \n",
            "Q 98+695  T 793  \u001b[92m☑\u001b[0m 793 \n",
            "Q 747+455 T 1202 \u001b[92m☑\u001b[0m 1202\n",
            "Q 63+347  T 410  \u001b[92m☑\u001b[0m 410 \n",
            "Q 28+444  T 472  \u001b[92m☑\u001b[0m 472 \n",
            "Q 7+728   T 735  \u001b[92m☑\u001b[0m 735 \n",
            "Q 932+66  T 998  \u001b[92m☑\u001b[0m 998 \n",
            "Q 362+60  T 422  \u001b[92m☑\u001b[0m 422 \n",
            "Q 95+2    T 97   \u001b[92m☑\u001b[0m 97  \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 22\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 0.0793 - acc: 0.9827 - val_loss: 0.0870 - val_acc: 0.9783\n",
            "Q 204+63  T 267  \u001b[92m☑\u001b[0m 267 \n",
            "Q 713+62  T 775  \u001b[92m☑\u001b[0m 775 \n",
            "Q 529+41  T 570  \u001b[92m☑\u001b[0m 570 \n",
            "Q 220+946 T 1166 \u001b[92m☑\u001b[0m 1166\n",
            "Q 16+76   T 92   \u001b[92m☑\u001b[0m 92  \n",
            "Q 321+690 T 1011 \u001b[92m☑\u001b[0m 1011\n",
            "Q 82+942  T 1024 \u001b[92m☑\u001b[0m 1024\n",
            "Q 390+6   T 396  \u001b[92m☑\u001b[0m 396 \n",
            "Q 252+45  T 297  \u001b[92m☑\u001b[0m 297 \n",
            "Q 196+35  T 231  \u001b[92m☑\u001b[0m 231 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 23\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 325us/sample - loss: 0.0695 - acc: 0.9853 - val_loss: 0.0793 - val_acc: 0.9793\n",
            "Q 35+607  T 642  \u001b[92m☑\u001b[0m 642 \n",
            "Q 111+29  T 140  \u001b[92m☑\u001b[0m 140 \n",
            "Q 83+49   T 132  \u001b[92m☑\u001b[0m 132 \n",
            "Q 508+41  T 549  \u001b[92m☑\u001b[0m 549 \n",
            "Q 132+10  T 142  \u001b[92m☑\u001b[0m 142 \n",
            "Q 295+477 T 772  \u001b[91m☒\u001b[0m 762 \n",
            "Q 874+819 T 1693 \u001b[92m☑\u001b[0m 1693\n",
            "Q 198+73  T 271  \u001b[91m☒\u001b[0m 261 \n",
            "Q 887+96  T 983  \u001b[92m☑\u001b[0m 983 \n",
            "Q 442+234 T 676  \u001b[92m☑\u001b[0m 676 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 24\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 328us/sample - loss: 0.0674 - acc: 0.9843 - val_loss: 0.0721 - val_acc: 0.9817\n",
            "Q 30+490  T 520  \u001b[92m☑\u001b[0m 520 \n",
            "Q 433+0   T 433  \u001b[92m☑\u001b[0m 433 \n",
            "Q 51+583  T 634  \u001b[92m☑\u001b[0m 634 \n",
            "Q 94+11   T 105  \u001b[92m☑\u001b[0m 105 \n",
            "Q 7+55    T 62   \u001b[92m☑\u001b[0m 62  \n",
            "Q 498+763 T 1261 \u001b[92m☑\u001b[0m 1261\n",
            "Q 98+46   T 144  \u001b[92m☑\u001b[0m 144 \n",
            "Q 7+218   T 225  \u001b[92m☑\u001b[0m 225 \n",
            "Q 326+6   T 332  \u001b[92m☑\u001b[0m 332 \n",
            "Q 249+83  T 332  \u001b[92m☑\u001b[0m 332 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 25\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 330us/sample - loss: 0.0660 - acc: 0.9835 - val_loss: 0.0879 - val_acc: 0.9743\n",
            "Q 658+469 T 1127 \u001b[92m☑\u001b[0m 1127\n",
            "Q 7+261   T 268  \u001b[91m☒\u001b[0m 267 \n",
            "Q 68+689  T 757  \u001b[92m☑\u001b[0m 757 \n",
            "Q 353+8   T 361  \u001b[92m☑\u001b[0m 361 \n",
            "Q 306+580 T 886  \u001b[92m☑\u001b[0m 886 \n",
            "Q 15+499  T 514  \u001b[92m☑\u001b[0m 514 \n",
            "Q 91+316  T 407  \u001b[92m☑\u001b[0m 407 \n",
            "Q 479+6   T 485  \u001b[92m☑\u001b[0m 485 \n",
            "Q 906+98  T 1004 \u001b[92m☑\u001b[0m 1004\n",
            "Q 78+698  T 776  \u001b[92m☑\u001b[0m 776 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 26\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 0.0561 - acc: 0.9869 - val_loss: 0.0687 - val_acc: 0.9801\n",
            "Q 75+0    T 75   \u001b[92m☑\u001b[0m 75  \n",
            "Q 0+174   T 174  \u001b[92m☑\u001b[0m 174 \n",
            "Q 93+912  T 1005 \u001b[92m☑\u001b[0m 1005\n",
            "Q 49+89   T 138  \u001b[92m☑\u001b[0m 138 \n",
            "Q 0+471   T 471  \u001b[92m☑\u001b[0m 471 \n",
            "Q 446+93  T 539  \u001b[92m☑\u001b[0m 539 \n",
            "Q 265+134 T 399  \u001b[92m☑\u001b[0m 399 \n",
            "Q 630+9   T 639  \u001b[91m☒\u001b[0m 649 \n",
            "Q 687+32  T 719  \u001b[92m☑\u001b[0m 719 \n",
            "Q 286+522 T 808  \u001b[91m☒\u001b[0m 708 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 27\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 322us/sample - loss: 0.0490 - acc: 0.9890 - val_loss: 0.0539 - val_acc: 0.9852\n",
            "Q 565+10  T 575  \u001b[92m☑\u001b[0m 575 \n",
            "Q 54+37   T 91   \u001b[92m☑\u001b[0m 91  \n",
            "Q 855+710 T 1565 \u001b[92m☑\u001b[0m 1565\n",
            "Q 469+213 T 682  \u001b[92m☑\u001b[0m 682 \n",
            "Q 868+566 T 1434 \u001b[92m☑\u001b[0m 1434\n",
            "Q 367+633 T 1000 \u001b[91m☒\u001b[0m 1001\n",
            "Q 615+19  T 634  \u001b[92m☑\u001b[0m 634 \n",
            "Q 344+303 T 647  \u001b[92m☑\u001b[0m 647 \n",
            "Q 111+29  T 140  \u001b[92m☑\u001b[0m 140 \n",
            "Q 190+329 T 519  \u001b[92m☑\u001b[0m 519 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 28\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 328us/sample - loss: 0.0475 - acc: 0.9889 - val_loss: 0.0693 - val_acc: 0.9789\n",
            "Q 58+347  T 405  \u001b[92m☑\u001b[0m 405 \n",
            "Q 89+374  T 463  \u001b[92m☑\u001b[0m 463 \n",
            "Q 906+35  T 941  \u001b[92m☑\u001b[0m 941 \n",
            "Q 626+756 T 1382 \u001b[92m☑\u001b[0m 1382\n",
            "Q 320+33  T 353  \u001b[92m☑\u001b[0m 353 \n",
            "Q 362+260 T 622  \u001b[92m☑\u001b[0m 622 \n",
            "Q 297+9   T 306  \u001b[92m☑\u001b[0m 306 \n",
            "Q 727+81  T 808  \u001b[92m☑\u001b[0m 808 \n",
            "Q 33+890  T 923  \u001b[92m☑\u001b[0m 923 \n",
            "Q 70+678  T 748  \u001b[92m☑\u001b[0m 748 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 29\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 322us/sample - loss: 0.0441 - acc: 0.9897 - val_loss: 0.0429 - val_acc: 0.9891\n",
            "Q 602+853 T 1455 \u001b[92m☑\u001b[0m 1455\n",
            "Q 47+174  T 221  \u001b[92m☑\u001b[0m 221 \n",
            "Q 430+66  T 496  \u001b[92m☑\u001b[0m 496 \n",
            "Q 49+55   T 104  \u001b[92m☑\u001b[0m 104 \n",
            "Q 347+988 T 1335 \u001b[92m☑\u001b[0m 1335\n",
            "Q 149+653 T 802  \u001b[92m☑\u001b[0m 802 \n",
            "Q 771+9   T 780  \u001b[92m☑\u001b[0m 780 \n",
            "Q 674+83  T 757  \u001b[92m☑\u001b[0m 757 \n",
            "Q 857+17  T 874  \u001b[92m☑\u001b[0m 874 \n",
            "Q 42+900  T 942  \u001b[92m☑\u001b[0m 942 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 30\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 325us/sample - loss: 0.0300 - acc: 0.9944 - val_loss: 0.0483 - val_acc: 0.9858\n",
            "Q 655+89  T 744  \u001b[92m☑\u001b[0m 744 \n",
            "Q 456+192 T 648  \u001b[92m☑\u001b[0m 648 \n",
            "Q 23+261  T 284  \u001b[92m☑\u001b[0m 284 \n",
            "Q 13+94   T 107  \u001b[92m☑\u001b[0m 107 \n",
            "Q 993+29  T 1022 \u001b[92m☑\u001b[0m 1022\n",
            "Q 699+961 T 1660 \u001b[91m☒\u001b[0m 1650\n",
            "Q 810+38  T 848  \u001b[92m☑\u001b[0m 848 \n",
            "Q 364+493 T 857  \u001b[92m☑\u001b[0m 857 \n",
            "Q 180+216 T 396  \u001b[92m☑\u001b[0m 396 \n",
            "Q 401+89  T 490  \u001b[92m☑\u001b[0m 490 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 31\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 0.0514 - acc: 0.9858 - val_loss: 0.0585 - val_acc: 0.9801\n",
            "Q 685+42  T 727  \u001b[92m☑\u001b[0m 727 \n",
            "Q 18+22   T 40   \u001b[92m☑\u001b[0m 40  \n",
            "Q 760+871 T 1631 \u001b[92m☑\u001b[0m 1631\n",
            "Q 278+163 T 441  \u001b[92m☑\u001b[0m 441 \n",
            "Q 891+857 T 1748 \u001b[91m☒\u001b[0m 1758\n",
            "Q 834+52  T 886  \u001b[92m☑\u001b[0m 886 \n",
            "Q 678+98  T 776  \u001b[92m☑\u001b[0m 776 \n",
            "Q 324+48  T 372  \u001b[92m☑\u001b[0m 372 \n",
            "Q 198+450 T 648  \u001b[92m☑\u001b[0m 648 \n",
            "Q 46+205  T 251  \u001b[92m☑\u001b[0m 251 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 32\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 0.0303 - acc: 0.9936 - val_loss: 0.0390 - val_acc: 0.9894\n",
            "Q 601+62  T 663  \u001b[92m☑\u001b[0m 663 \n",
            "Q 98+31   T 129  \u001b[92m☑\u001b[0m 129 \n",
            "Q 71+38   T 109  \u001b[92m☑\u001b[0m 109 \n",
            "Q 72+404  T 476  \u001b[92m☑\u001b[0m 476 \n",
            "Q 847+68  T 915  \u001b[92m☑\u001b[0m 915 \n",
            "Q 796+2   T 798  \u001b[92m☑\u001b[0m 798 \n",
            "Q 38+58   T 96   \u001b[92m☑\u001b[0m 96  \n",
            "Q 22+550  T 572  \u001b[92m☑\u001b[0m 572 \n",
            "Q 3+764   T 767  \u001b[92m☑\u001b[0m 767 \n",
            "Q 34+77   T 111  \u001b[92m☑\u001b[0m 111 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 33\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 0.0213 - acc: 0.9967 - val_loss: 0.0436 - val_acc: 0.9863\n",
            "Q 76+39   T 115  \u001b[92m☑\u001b[0m 115 \n",
            "Q 898+175 T 1073 \u001b[92m☑\u001b[0m 1073\n",
            "Q 158+313 T 471  \u001b[92m☑\u001b[0m 471 \n",
            "Q 6+358   T 364  \u001b[92m☑\u001b[0m 364 \n",
            "Q 99+940  T 1039 \u001b[92m☑\u001b[0m 1039\n",
            "Q 62+66   T 128  \u001b[92m☑\u001b[0m 128 \n",
            "Q 89+988  T 1077 \u001b[91m☒\u001b[0m 1067\n",
            "Q 3+183   T 186  \u001b[92m☑\u001b[0m 186 \n",
            "Q 4+751   T 755  \u001b[92m☑\u001b[0m 755 \n",
            "Q 816+7   T 823  \u001b[92m☑\u001b[0m 823 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 34\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 328us/sample - loss: 0.0293 - acc: 0.9931 - val_loss: 0.0687 - val_acc: 0.9750\n",
            "Q 216+89  T 305  \u001b[92m☑\u001b[0m 305 \n",
            "Q 187+203 T 390  \u001b[92m☑\u001b[0m 390 \n",
            "Q 154+5   T 159  \u001b[91m☒\u001b[0m 169 \n",
            "Q 66+70   T 136  \u001b[92m☑\u001b[0m 136 \n",
            "Q 25+984  T 1009 \u001b[92m☑\u001b[0m 1009\n",
            "Q 539+916 T 1455 \u001b[92m☑\u001b[0m 1455\n",
            "Q 49+41   T 90   \u001b[92m☑\u001b[0m 90  \n",
            "Q 286+522 T 808  \u001b[92m☑\u001b[0m 808 \n",
            "Q 84+39   T 123  \u001b[91m☒\u001b[0m 133 \n",
            "Q 911+8   T 919  \u001b[92m☑\u001b[0m 919 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 35\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 320us/sample - loss: 0.0464 - acc: 0.9863 - val_loss: 0.0642 - val_acc: 0.9789\n",
            "Q 2+16    T 18   \u001b[92m☑\u001b[0m 18  \n",
            "Q 18+363  T 381  \u001b[92m☑\u001b[0m 381 \n",
            "Q 340+12  T 352  \u001b[92m☑\u001b[0m 352 \n",
            "Q 3+822   T 825  \u001b[92m☑\u001b[0m 825 \n",
            "Q 6+943   T 949  \u001b[92m☑\u001b[0m 949 \n",
            "Q 891+117 T 1008 \u001b[92m☑\u001b[0m 1008\n",
            "Q 94+414  T 508  \u001b[92m☑\u001b[0m 508 \n",
            "Q 697+947 T 1644 \u001b[91m☒\u001b[0m 1634\n",
            "Q 97+183  T 280  \u001b[92m☑\u001b[0m 280 \n",
            "Q 461+18  T 479  \u001b[92m☑\u001b[0m 479 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 36\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 319us/sample - loss: 0.0278 - acc: 0.9933 - val_loss: 0.0271 - val_acc: 0.9929\n",
            "Q 78+449  T 527  \u001b[92m☑\u001b[0m 527 \n",
            "Q 90+60   T 150  \u001b[92m☑\u001b[0m 150 \n",
            "Q 73+164  T 237  \u001b[92m☑\u001b[0m 237 \n",
            "Q 11+218  T 229  \u001b[92m☑\u001b[0m 229 \n",
            "Q 6+947   T 953  \u001b[92m☑\u001b[0m 953 \n",
            "Q 873+832 T 1705 \u001b[92m☑\u001b[0m 1705\n",
            "Q 55+638  T 693  \u001b[92m☑\u001b[0m 693 \n",
            "Q 46+225  T 271  \u001b[92m☑\u001b[0m 271 \n",
            "Q 24+436  T 460  \u001b[92m☑\u001b[0m 460 \n",
            "Q 68+800  T 868  \u001b[92m☑\u001b[0m 868 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 37\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 322us/sample - loss: 0.0132 - acc: 0.9985 - val_loss: 0.0237 - val_acc: 0.9936\n",
            "Q 0+471   T 471  \u001b[92m☑\u001b[0m 471 \n",
            "Q 71+140  T 211  \u001b[92m☑\u001b[0m 211 \n",
            "Q 251+147 T 398  \u001b[92m☑\u001b[0m 398 \n",
            "Q 387+53  T 440  \u001b[92m☑\u001b[0m 440 \n",
            "Q 26+638  T 664  \u001b[92m☑\u001b[0m 664 \n",
            "Q 391+219 T 610  \u001b[91m☒\u001b[0m 600 \n",
            "Q 480+154 T 634  \u001b[92m☑\u001b[0m 634 \n",
            "Q 686+578 T 1264 \u001b[92m☑\u001b[0m 1264\n",
            "Q 28+502  T 530  \u001b[92m☑\u001b[0m 530 \n",
            "Q 654+13  T 667  \u001b[92m☑\u001b[0m 667 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 38\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 322us/sample - loss: 0.0129 - acc: 0.9982 - val_loss: 0.0265 - val_acc: 0.9919\n",
            "Q 632+616 T 1248 \u001b[92m☑\u001b[0m 1248\n",
            "Q 246+1   T 247  \u001b[92m☑\u001b[0m 247 \n",
            "Q 122+430 T 552  \u001b[92m☑\u001b[0m 552 \n",
            "Q 661+52  T 713  \u001b[92m☑\u001b[0m 713 \n",
            "Q 811+9   T 820  \u001b[92m☑\u001b[0m 820 \n",
            "Q 252+543 T 795  \u001b[92m☑\u001b[0m 795 \n",
            "Q 44+760  T 804  \u001b[92m☑\u001b[0m 804 \n",
            "Q 912+674 T 1586 \u001b[92m☑\u001b[0m 1586\n",
            "Q 388+67  T 455  \u001b[92m☑\u001b[0m 455 \n",
            "Q 259+229 T 488  \u001b[92m☑\u001b[0m 488 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 39\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 0.0399 - acc: 0.9886 - val_loss: 0.1708 - val_acc: 0.9469\n",
            "Q 832+403 T 1235 \u001b[91m☒\u001b[0m 1335\n",
            "Q 8+272   T 280  \u001b[92m☑\u001b[0m 280 \n",
            "Q 88+398  T 486  \u001b[91m☒\u001b[0m 496 \n",
            "Q 3+443   T 446  \u001b[92m☑\u001b[0m 446 \n",
            "Q 474+611 T 1085 \u001b[91m☒\u001b[0m 1095\n",
            "Q 104+62  T 166  \u001b[92m☑\u001b[0m 166 \n",
            "Q 476+64  T 540  \u001b[92m☑\u001b[0m 540 \n",
            "Q 966+372 T 1338 \u001b[92m☑\u001b[0m 1338\n",
            "Q 100+47  T 147  \u001b[92m☑\u001b[0m 147 \n",
            "Q 437+137 T 574  \u001b[92m☑\u001b[0m 574 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 40\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 0.0314 - acc: 0.9912 - val_loss: 0.0378 - val_acc: 0.9883\n",
            "Q 93+804  T 897  \u001b[91m☒\u001b[0m 997 \n",
            "Q 679+457 T 1136 \u001b[92m☑\u001b[0m 1136\n",
            "Q 107+855 T 962  \u001b[92m☑\u001b[0m 962 \n",
            "Q 748+14  T 762  \u001b[92m☑\u001b[0m 762 \n",
            "Q 95+18   T 113  \u001b[92m☑\u001b[0m 113 \n",
            "Q 33+75   T 108  \u001b[92m☑\u001b[0m 108 \n",
            "Q 3+183   T 186  \u001b[92m☑\u001b[0m 186 \n",
            "Q 479+794 T 1273 \u001b[92m☑\u001b[0m 1273\n",
            "Q 2+964   T 966  \u001b[92m☑\u001b[0m 966 \n",
            "Q 160+82  T 242  \u001b[92m☑\u001b[0m 242 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 41\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 319us/sample - loss: 0.0124 - acc: 0.9982 - val_loss: 0.0211 - val_acc: 0.9944\n",
            "Q 7+237   T 244  \u001b[92m☑\u001b[0m 244 \n",
            "Q 75+374  T 449  \u001b[92m☑\u001b[0m 449 \n",
            "Q 638+265 T 903  \u001b[92m☑\u001b[0m 903 \n",
            "Q 60+907  T 967  \u001b[92m☑\u001b[0m 967 \n",
            "Q 4+656   T 660  \u001b[92m☑\u001b[0m 660 \n",
            "Q 59+690  T 749  \u001b[92m☑\u001b[0m 749 \n",
            "Q 492+10  T 502  \u001b[92m☑\u001b[0m 502 \n",
            "Q 636+946 T 1582 \u001b[92m☑\u001b[0m 1582\n",
            "Q 7+90    T 97   \u001b[92m☑\u001b[0m 97  \n",
            "Q 241+12  T 253  \u001b[92m☑\u001b[0m 253 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 42\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 0.0090 - acc: 0.9991 - val_loss: 0.0191 - val_acc: 0.9945\n",
            "Q 889+289 T 1178 \u001b[92m☑\u001b[0m 1178\n",
            "Q 288+5   T 293  \u001b[92m☑\u001b[0m 293 \n",
            "Q 23+53   T 76   \u001b[92m☑\u001b[0m 76  \n",
            "Q 523+591 T 1114 \u001b[92m☑\u001b[0m 1114\n",
            "Q 756+121 T 877  \u001b[92m☑\u001b[0m 877 \n",
            "Q 109+68  T 177  \u001b[92m☑\u001b[0m 177 \n",
            "Q 936+611 T 1547 \u001b[92m☑\u001b[0m 1547\n",
            "Q 373+18  T 391  \u001b[92m☑\u001b[0m 391 \n",
            "Q 276+626 T 902  \u001b[91m☒\u001b[0m 802 \n",
            "Q 38+15   T 53   \u001b[92m☑\u001b[0m 53  \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 43\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 323us/sample - loss: 0.0101 - acc: 0.9983 - val_loss: 0.0393 - val_acc: 0.9875\n",
            "Q 70+196  T 266  \u001b[92m☑\u001b[0m 266 \n",
            "Q 320+517 T 837  \u001b[92m☑\u001b[0m 837 \n",
            "Q 21+65   T 86   \u001b[92m☑\u001b[0m 86  \n",
            "Q 69+855  T 924  \u001b[92m☑\u001b[0m 924 \n",
            "Q 99+861  T 960  \u001b[92m☑\u001b[0m 960 \n",
            "Q 61+478  T 539  \u001b[92m☑\u001b[0m 539 \n",
            "Q 35+12   T 47   \u001b[92m☑\u001b[0m 47  \n",
            "Q 3+614   T 617  \u001b[92m☑\u001b[0m 617 \n",
            "Q 733+35  T 768  \u001b[92m☑\u001b[0m 768 \n",
            "Q 216+37  T 253  \u001b[92m☑\u001b[0m 253 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 44\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 329us/sample - loss: 0.0574 - acc: 0.9813 - val_loss: 0.0314 - val_acc: 0.9906\n",
            "Q 773+226 T 999  \u001b[92m☑\u001b[0m 999 \n",
            "Q 580+48  T 628  \u001b[92m☑\u001b[0m 628 \n",
            "Q 995+92  T 1087 \u001b[92m☑\u001b[0m 1087\n",
            "Q 36+157  T 193  \u001b[92m☑\u001b[0m 193 \n",
            "Q 242+540 T 782  \u001b[92m☑\u001b[0m 782 \n",
            "Q 2+930   T 932  \u001b[92m☑\u001b[0m 932 \n",
            "Q 959+9   T 968  \u001b[92m☑\u001b[0m 968 \n",
            "Q 27+63   T 90   \u001b[92m☑\u001b[0m 90  \n",
            "Q 933+13  T 946  \u001b[92m☑\u001b[0m 946 \n",
            "Q 49+8    T 57   \u001b[92m☑\u001b[0m 57  \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 45\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 329us/sample - loss: 0.0089 - acc: 0.9991 - val_loss: 0.0178 - val_acc: 0.9951\n",
            "Q 266+981 T 1247 \u001b[92m☑\u001b[0m 1247\n",
            "Q 41+200  T 241  \u001b[92m☑\u001b[0m 241 \n",
            "Q 3+22    T 25   \u001b[92m☑\u001b[0m 25  \n",
            "Q 29+329  T 358  \u001b[92m☑\u001b[0m 358 \n",
            "Q 3+633   T 636  \u001b[92m☑\u001b[0m 636 \n",
            "Q 920+45  T 965  \u001b[92m☑\u001b[0m 965 \n",
            "Q 69+600  T 669  \u001b[92m☑\u001b[0m 669 \n",
            "Q 334+81  T 415  \u001b[92m☑\u001b[0m 415 \n",
            "Q 578+78  T 656  \u001b[92m☑\u001b[0m 656 \n",
            "Q 886+93  T 979  \u001b[92m☑\u001b[0m 979 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 46\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 329us/sample - loss: 0.0067 - acc: 0.9994 - val_loss: 0.0173 - val_acc: 0.9947\n",
            "Q 477+630 T 1107 \u001b[92m☑\u001b[0m 1107\n",
            "Q 3+36    T 39   \u001b[92m☑\u001b[0m 39  \n",
            "Q 459+596 T 1055 \u001b[92m☑\u001b[0m 1055\n",
            "Q 888+47  T 935  \u001b[92m☑\u001b[0m 935 \n",
            "Q 431+17  T 448  \u001b[92m☑\u001b[0m 448 \n",
            "Q 903+2   T 905  \u001b[92m☑\u001b[0m 905 \n",
            "Q 422+61  T 483  \u001b[92m☑\u001b[0m 483 \n",
            "Q 192+25  T 217  \u001b[92m☑\u001b[0m 217 \n",
            "Q 17+23   T 40   \u001b[92m☑\u001b[0m 40  \n",
            "Q 395+60  T 455  \u001b[92m☑\u001b[0m 455 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 47\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 322us/sample - loss: 0.0119 - acc: 0.9975 - val_loss: 0.0644 - val_acc: 0.9785\n",
            "Q 74+231  T 305  \u001b[92m☑\u001b[0m 305 \n",
            "Q 190+63  T 253  \u001b[92m☑\u001b[0m 253 \n",
            "Q 383+552 T 935  \u001b[92m☑\u001b[0m 935 \n",
            "Q 892+320 T 1212 \u001b[92m☑\u001b[0m 1212\n",
            "Q 30+57   T 87   \u001b[92m☑\u001b[0m 87  \n",
            "Q 758+17  T 775  \u001b[92m☑\u001b[0m 775 \n",
            "Q 28+879  T 907  \u001b[91m☒\u001b[0m 917 \n",
            "Q 73+179  T 252  \u001b[92m☑\u001b[0m 252 \n",
            "Q 731+13  T 744  \u001b[92m☑\u001b[0m 744 \n",
            "Q 516+716 T 1232 \u001b[92m☑\u001b[0m 1232\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 48\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 0.0301 - acc: 0.9909 - val_loss: 0.0201 - val_acc: 0.9938\n",
            "Q 65+851  T 916  \u001b[92m☑\u001b[0m 916 \n",
            "Q 817+51  T 868  \u001b[92m☑\u001b[0m 868 \n",
            "Q 59+59   T 118  \u001b[92m☑\u001b[0m 118 \n",
            "Q 252+948 T 1200 \u001b[92m☑\u001b[0m 1200\n",
            "Q 223+333 T 556  \u001b[92m☑\u001b[0m 556 \n",
            "Q 76+230  T 306  \u001b[92m☑\u001b[0m 306 \n",
            "Q 608+8   T 616  \u001b[92m☑\u001b[0m 616 \n",
            "Q 8+906   T 914  \u001b[92m☑\u001b[0m 914 \n",
            "Q 459+93  T 552  \u001b[92m☑\u001b[0m 552 \n",
            "Q 52+379  T 431  \u001b[92m☑\u001b[0m 431 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 49\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 320us/sample - loss: 0.0072 - acc: 0.9991 - val_loss: 0.0164 - val_acc: 0.9953\n",
            "Q 483+74  T 557  \u001b[92m☑\u001b[0m 557 \n",
            "Q 88+293  T 381  \u001b[92m☑\u001b[0m 381 \n",
            "Q 884+54  T 938  \u001b[92m☑\u001b[0m 938 \n",
            "Q 140+3   T 143  \u001b[92m☑\u001b[0m 143 \n",
            "Q 29+792  T 821  \u001b[92m☑\u001b[0m 821 \n",
            "Q 97+744  T 841  \u001b[92m☑\u001b[0m 841 \n",
            "Q 447+922 T 1369 \u001b[92m☑\u001b[0m 1369\n",
            "Q 87+844  T 931  \u001b[92m☑\u001b[0m 931 \n",
            "Q 92+77   T 169  \u001b[92m☑\u001b[0m 169 \n",
            "Q 110+450 T 560  \u001b[92m☑\u001b[0m 560 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 50\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 0.0048 - acc: 0.9997 - val_loss: 0.0145 - val_acc: 0.9956\n",
            "Q 440+6   T 446  \u001b[92m☑\u001b[0m 446 \n",
            "Q 297+590 T 887  \u001b[92m☑\u001b[0m 887 \n",
            "Q 440+26  T 466  \u001b[92m☑\u001b[0m 466 \n",
            "Q 2+50    T 52   \u001b[92m☑\u001b[0m 52  \n",
            "Q 609+80  T 689  \u001b[92m☑\u001b[0m 689 \n",
            "Q 577+913 T 1490 \u001b[92m☑\u001b[0m 1490\n",
            "Q 47+185  T 232  \u001b[92m☑\u001b[0m 232 \n",
            "Q 30+513  T 543  \u001b[92m☑\u001b[0m 543 \n",
            "Q 30+513  T 543  \u001b[92m☑\u001b[0m 543 \n",
            "Q 0+327   T 327  \u001b[92m☑\u001b[0m 327 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 51\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 322us/sample - loss: 0.0046 - acc: 0.9997 - val_loss: 0.0154 - val_acc: 0.9951\n",
            "Q 50+115  T 165  \u001b[92m☑\u001b[0m 165 \n",
            "Q 226+79  T 305  \u001b[92m☑\u001b[0m 305 \n",
            "Q 939+97  T 1036 \u001b[92m☑\u001b[0m 1036\n",
            "Q 577+913 T 1490 \u001b[92m☑\u001b[0m 1490\n",
            "Q 813+522 T 1335 \u001b[92m☑\u001b[0m 1335\n",
            "Q 2+351   T 353  \u001b[92m☑\u001b[0m 353 \n",
            "Q 83+448  T 531  \u001b[92m☑\u001b[0m 531 \n",
            "Q 312+796 T 1108 \u001b[92m☑\u001b[0m 1108\n",
            "Q 670+20  T 690  \u001b[92m☑\u001b[0m 690 \n",
            "Q 429+132 T 561  \u001b[92m☑\u001b[0m 561 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 52\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 0.0655 - acc: 0.9792 - val_loss: 0.0296 - val_acc: 0.9909\n",
            "Q 379+968 T 1347 \u001b[92m☑\u001b[0m 1347\n",
            "Q 759+577 T 1336 \u001b[92m☑\u001b[0m 1336\n",
            "Q 8+288   T 296  \u001b[92m☑\u001b[0m 296 \n",
            "Q 789+83  T 872  \u001b[92m☑\u001b[0m 872 \n",
            "Q 77+938  T 1015 \u001b[92m☑\u001b[0m 1015\n",
            "Q 164+147 T 311  \u001b[92m☑\u001b[0m 311 \n",
            "Q 192+95  T 287  \u001b[92m☑\u001b[0m 287 \n",
            "Q 999+648 T 1647 \u001b[91m☒\u001b[0m 1637\n",
            "Q 963+34  T 997  \u001b[92m☑\u001b[0m 997 \n",
            "Q 120+31  T 151  \u001b[92m☑\u001b[0m 151 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 53\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 318us/sample - loss: 0.0127 - acc: 0.9972 - val_loss: 0.0200 - val_acc: 0.9938\n",
            "Q 27+8    T 35   \u001b[92m☑\u001b[0m 35  \n",
            "Q 143+6   T 149  \u001b[92m☑\u001b[0m 149 \n",
            "Q 212+3   T 215  \u001b[92m☑\u001b[0m 215 \n",
            "Q 235+908 T 1143 \u001b[92m☑\u001b[0m 1143\n",
            "Q 276+40  T 316  \u001b[92m☑\u001b[0m 316 \n",
            "Q 49+17   T 66   \u001b[92m☑\u001b[0m 66  \n",
            "Q 134+35  T 169  \u001b[92m☑\u001b[0m 169 \n",
            "Q 0+969   T 969  \u001b[92m☑\u001b[0m 969 \n",
            "Q 59+270  T 329  \u001b[92m☑\u001b[0m 329 \n",
            "Q 210+7   T 217  \u001b[92m☑\u001b[0m 217 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 54\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 322us/sample - loss: 0.0045 - acc: 0.9998 - val_loss: 0.0129 - val_acc: 0.9967\n",
            "Q 210+18  T 228  \u001b[92m☑\u001b[0m 228 \n",
            "Q 635+1   T 636  \u001b[92m☑\u001b[0m 636 \n",
            "Q 287+52  T 339  \u001b[92m☑\u001b[0m 339 \n",
            "Q 892+320 T 1212 \u001b[92m☑\u001b[0m 1212\n",
            "Q 0+155   T 155  \u001b[92m☑\u001b[0m 155 \n",
            "Q 94+436  T 530  \u001b[92m☑\u001b[0m 530 \n",
            "Q 3+731   T 734  \u001b[92m☑\u001b[0m 734 \n",
            "Q 378+20  T 398  \u001b[92m☑\u001b[0m 398 \n",
            "Q 767+20  T 787  \u001b[92m☑\u001b[0m 787 \n",
            "Q 370+18  T 388  \u001b[92m☑\u001b[0m 388 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 55\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 305us/sample - loss: 0.0036 - acc: 0.9999 - val_loss: 0.0126 - val_acc: 0.9967\n",
            "Q 0+969   T 969  \u001b[92m☑\u001b[0m 969 \n",
            "Q 55+317  T 372  \u001b[92m☑\u001b[0m 372 \n",
            "Q 91+960  T 1051 \u001b[92m☑\u001b[0m 1051\n",
            "Q 92+780  T 872  \u001b[92m☑\u001b[0m 872 \n",
            "Q 919+882 T 1801 \u001b[92m☑\u001b[0m 1801\n",
            "Q 207+9   T 216  \u001b[92m☑\u001b[0m 216 \n",
            "Q 28+816  T 844  \u001b[92m☑\u001b[0m 844 \n",
            "Q 92+780  T 872  \u001b[92m☑\u001b[0m 872 \n",
            "Q 27+63   T 90   \u001b[92m☑\u001b[0m 90  \n",
            "Q 693+12  T 705  \u001b[92m☑\u001b[0m 705 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 56\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 309us/sample - loss: 0.0033 - acc: 0.9999 - val_loss: 0.0120 - val_acc: 0.9968\n",
            "Q 485+60  T 545  \u001b[92m☑\u001b[0m 545 \n",
            "Q 26+902  T 928  \u001b[92m☑\u001b[0m 928 \n",
            "Q 625+371 T 996  \u001b[92m☑\u001b[0m 996 \n",
            "Q 732+9   T 741  \u001b[92m☑\u001b[0m 741 \n",
            "Q 81+430  T 511  \u001b[92m☑\u001b[0m 511 \n",
            "Q 42+641  T 683  \u001b[92m☑\u001b[0m 683 \n",
            "Q 440+26  T 466  \u001b[92m☑\u001b[0m 466 \n",
            "Q 785+13  T 798  \u001b[92m☑\u001b[0m 798 \n",
            "Q 126+20  T 146  \u001b[92m☑\u001b[0m 146 \n",
            "Q 126+64  T 190  \u001b[92m☑\u001b[0m 190 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 57\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 305us/sample - loss: 0.0217 - acc: 0.9942 - val_loss: 0.2479 - val_acc: 0.9287\n",
            "Q 346+59  T 405  \u001b[92m☑\u001b[0m 405 \n",
            "Q 470+9   T 479  \u001b[92m☑\u001b[0m 479 \n",
            "Q 45+44   T 89   \u001b[92m☑\u001b[0m 89  \n",
            "Q 912+674 T 1586 \u001b[92m☑\u001b[0m 1586\n",
            "Q 593+84  T 677  \u001b[91m☒\u001b[0m 678 \n",
            "Q 951+48  T 999  \u001b[91m☒\u001b[0m 1009\n",
            "Q 542+33  T 575  \u001b[92m☑\u001b[0m 575 \n",
            "Q 65+715  T 780  \u001b[92m☑\u001b[0m 780 \n",
            "Q 8+31    T 39   \u001b[92m☑\u001b[0m 39  \n",
            "Q 829+5   T 834  \u001b[92m☑\u001b[0m 834 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 58\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 309us/sample - loss: 0.0345 - acc: 0.9898 - val_loss: 0.0147 - val_acc: 0.9961\n",
            "Q 667+458 T 1125 \u001b[92m☑\u001b[0m 1125\n",
            "Q 5+818   T 823  \u001b[92m☑\u001b[0m 823 \n",
            "Q 246+1   T 247  \u001b[92m☑\u001b[0m 247 \n",
            "Q 21+141  T 162  \u001b[92m☑\u001b[0m 162 \n",
            "Q 769+688 T 1457 \u001b[92m☑\u001b[0m 1457\n",
            "Q 14+933  T 947  \u001b[92m☑\u001b[0m 947 \n",
            "Q 28+239  T 267  \u001b[92m☑\u001b[0m 267 \n",
            "Q 707+54  T 761  \u001b[92m☑\u001b[0m 761 \n",
            "Q 767+78  T 845  \u001b[92m☑\u001b[0m 845 \n",
            "Q 940+17  T 957  \u001b[92m☑\u001b[0m 957 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 59\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 302us/sample - loss: 0.0038 - acc: 0.9999 - val_loss: 0.0127 - val_acc: 0.9967\n",
            "Q 724+497 T 1221 \u001b[92m☑\u001b[0m 1221\n",
            "Q 858+903 T 1761 \u001b[92m☑\u001b[0m 1761\n",
            "Q 147+996 T 1143 \u001b[92m☑\u001b[0m 1143\n",
            "Q 4+483   T 487  \u001b[92m☑\u001b[0m 487 \n",
            "Q 201+357 T 558  \u001b[92m☑\u001b[0m 558 \n",
            "Q 510+78  T 588  \u001b[92m☑\u001b[0m 588 \n",
            "Q 68+790  T 858  \u001b[92m☑\u001b[0m 858 \n",
            "Q 180+492 T 672  \u001b[92m☑\u001b[0m 672 \n",
            "Q 11+108  T 119  \u001b[92m☑\u001b[0m 119 \n",
            "Q 463+56  T 519  \u001b[92m☑\u001b[0m 519 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 60\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 303us/sample - loss: 0.0030 - acc: 0.9999 - val_loss: 0.0124 - val_acc: 0.9963\n",
            "Q 326+796 T 1122 \u001b[92m☑\u001b[0m 1122\n",
            "Q 361+182 T 543  \u001b[92m☑\u001b[0m 543 \n",
            "Q 437+10  T 447  \u001b[92m☑\u001b[0m 447 \n",
            "Q 921+516 T 1437 \u001b[92m☑\u001b[0m 1437\n",
            "Q 963+34  T 997  \u001b[92m☑\u001b[0m 997 \n",
            "Q 548+99  T 647  \u001b[92m☑\u001b[0m 647 \n",
            "Q 302+294 T 596  \u001b[92m☑\u001b[0m 596 \n",
            "Q 937+723 T 1660 \u001b[92m☑\u001b[0m 1660\n",
            "Q 41+664  T 705  \u001b[92m☑\u001b[0m 705 \n",
            "Q 10+561  T 571  \u001b[92m☑\u001b[0m 571 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 61\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 309us/sample - loss: 0.0028 - acc: 0.9999 - val_loss: 0.0125 - val_acc: 0.9965\n",
            "Q 429+132 T 561  \u001b[92m☑\u001b[0m 561 \n",
            "Q 734+882 T 1616 \u001b[92m☑\u001b[0m 1616\n",
            "Q 195+582 T 777  \u001b[92m☑\u001b[0m 777 \n",
            "Q 142+47  T 189  \u001b[92m☑\u001b[0m 189 \n",
            "Q 31+330  T 361  \u001b[92m☑\u001b[0m 361 \n",
            "Q 372+59  T 431  \u001b[92m☑\u001b[0m 431 \n",
            "Q 7+188   T 195  \u001b[92m☑\u001b[0m 195 \n",
            "Q 252+45  T 297  \u001b[92m☑\u001b[0m 297 \n",
            "Q 184+757 T 941  \u001b[92m☑\u001b[0m 941 \n",
            "Q 6+361   T 367  \u001b[92m☑\u001b[0m 367 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 62\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 301us/sample - loss: 0.0361 - acc: 0.9883 - val_loss: 0.0322 - val_acc: 0.9894\n",
            "Q 770+727 T 1497 \u001b[92m☑\u001b[0m 1497\n",
            "Q 793+491 T 1284 \u001b[92m☑\u001b[0m 1284\n",
            "Q 449+57  T 506  \u001b[92m☑\u001b[0m 506 \n",
            "Q 499+2   T 501  \u001b[91m☒\u001b[0m 401 \n",
            "Q 789+0   T 789  \u001b[92m☑\u001b[0m 789 \n",
            "Q 452+47  T 499  \u001b[92m☑\u001b[0m 499 \n",
            "Q 8+506   T 514  \u001b[92m☑\u001b[0m 514 \n",
            "Q 83+96   T 179  \u001b[92m☑\u001b[0m 179 \n",
            "Q 88+874  T 962  \u001b[92m☑\u001b[0m 962 \n",
            "Q 142+548 T 690  \u001b[92m☑\u001b[0m 690 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 63\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 301us/sample - loss: 0.0064 - acc: 0.9990 - val_loss: 0.0137 - val_acc: 0.9958\n",
            "Q 863+495 T 1358 \u001b[92m☑\u001b[0m 1358\n",
            "Q 523+3   T 526  \u001b[92m☑\u001b[0m 526 \n",
            "Q 912+674 T 1586 \u001b[92m☑\u001b[0m 1586\n",
            "Q 164+792 T 956  \u001b[92m☑\u001b[0m 956 \n",
            "Q 1+627   T 628  \u001b[92m☑\u001b[0m 628 \n",
            "Q 642+272 T 914  \u001b[92m☑\u001b[0m 914 \n",
            "Q 20+78   T 98   \u001b[92m☑\u001b[0m 98  \n",
            "Q 936+1   T 937  \u001b[92m☑\u001b[0m 937 \n",
            "Q 87+941  T 1028 \u001b[92m☑\u001b[0m 1028\n",
            "Q 341+24  T 365  \u001b[92m☑\u001b[0m 365 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 64\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 305us/sample - loss: 0.0029 - acc: 0.9999 - val_loss: 0.0113 - val_acc: 0.9966\n",
            "Q 14+901  T 915  \u001b[92m☑\u001b[0m 915 \n",
            "Q 46+123  T 169  \u001b[92m☑\u001b[0m 169 \n",
            "Q 970+72  T 1042 \u001b[92m☑\u001b[0m 1042\n",
            "Q 26+775  T 801  \u001b[92m☑\u001b[0m 801 \n",
            "Q 42+442  T 484  \u001b[92m☑\u001b[0m 484 \n",
            "Q 6+410   T 416  \u001b[92m☑\u001b[0m 416 \n",
            "Q 583+990 T 1573 \u001b[92m☑\u001b[0m 1573\n",
            "Q 760+871 T 1631 \u001b[92m☑\u001b[0m 1631\n",
            "Q 968+266 T 1234 \u001b[92m☑\u001b[0m 1234\n",
            "Q 370+401 T 771  \u001b[92m☑\u001b[0m 771 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 65\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 309us/sample - loss: 0.0024 - acc: 1.0000 - val_loss: 0.0104 - val_acc: 0.9971\n",
            "Q 511+22  T 533  \u001b[92m☑\u001b[0m 533 \n",
            "Q 135+98  T 233  \u001b[92m☑\u001b[0m 233 \n",
            "Q 21+525  T 546  \u001b[92m☑\u001b[0m 546 \n",
            "Q 50+487  T 537  \u001b[92m☑\u001b[0m 537 \n",
            "Q 17+29   T 46   \u001b[92m☑\u001b[0m 46  \n",
            "Q 19+93   T 112  \u001b[92m☑\u001b[0m 112 \n",
            "Q 944+92  T 1036 \u001b[92m☑\u001b[0m 1036\n",
            "Q 177+144 T 321  \u001b[92m☑\u001b[0m 321 \n",
            "Q 891+18  T 909  \u001b[92m☑\u001b[0m 909 \n",
            "Q 6+776   T 782  \u001b[92m☑\u001b[0m 782 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 66\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 314us/sample - loss: 0.0023 - acc: 0.9999 - val_loss: 0.0143 - val_acc: 0.9952\n",
            "Q 934+69  T 1003 \u001b[92m☑\u001b[0m 1003\n",
            "Q 33+814  T 847  \u001b[92m☑\u001b[0m 847 \n",
            "Q 679+91  T 770  \u001b[92m☑\u001b[0m 770 \n",
            "Q 596+160 T 756  \u001b[92m☑\u001b[0m 756 \n",
            "Q 320+85  T 405  \u001b[92m☑\u001b[0m 405 \n",
            "Q 416+574 T 990  \u001b[92m☑\u001b[0m 990 \n",
            "Q 883+89  T 972  \u001b[92m☑\u001b[0m 972 \n",
            "Q 569+83  T 652  \u001b[92m☑\u001b[0m 652 \n",
            "Q 658+34  T 692  \u001b[92m☑\u001b[0m 692 \n",
            "Q 34+168  T 202  \u001b[92m☑\u001b[0m 202 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 67\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 316us/sample - loss: 0.0401 - acc: 0.9872 - val_loss: 0.0321 - val_acc: 0.9897\n",
            "Q 135+60  T 195  \u001b[92m☑\u001b[0m 195 \n",
            "Q 836+745 T 1581 \u001b[92m☑\u001b[0m 1581\n",
            "Q 66+809  T 875  \u001b[92m☑\u001b[0m 875 \n",
            "Q 9+995   T 1004 \u001b[91m☒\u001b[0m 1044\n",
            "Q 140+3   T 143  \u001b[92m☑\u001b[0m 143 \n",
            "Q 88+903  T 991  \u001b[92m☑\u001b[0m 991 \n",
            "Q 195+195 T 390  \u001b[91m☒\u001b[0m 380 \n",
            "Q 770+97  T 867  \u001b[92m☑\u001b[0m 867 \n",
            "Q 359+33  T 392  \u001b[92m☑\u001b[0m 392 \n",
            "Q 478+82  T 560  \u001b[92m☑\u001b[0m 560 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 68\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 308us/sample - loss: 0.0057 - acc: 0.9991 - val_loss: 0.0105 - val_acc: 0.9972\n",
            "Q 47+708  T 755  \u001b[92m☑\u001b[0m 755 \n",
            "Q 19+822  T 841  \u001b[92m☑\u001b[0m 841 \n",
            "Q 34+412  T 446  \u001b[92m☑\u001b[0m 446 \n",
            "Q 547+672 T 1219 \u001b[92m☑\u001b[0m 1219\n",
            "Q 557+83  T 640  \u001b[92m☑\u001b[0m 640 \n",
            "Q 446+82  T 528  \u001b[92m☑\u001b[0m 528 \n",
            "Q 47+345  T 392  \u001b[92m☑\u001b[0m 392 \n",
            "Q 661+3   T 664  \u001b[92m☑\u001b[0m 664 \n",
            "Q 9+273   T 282  \u001b[92m☑\u001b[0m 282 \n",
            "Q 580+8   T 588  \u001b[92m☑\u001b[0m 588 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 69\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 304us/sample - loss: 0.0022 - acc: 1.0000 - val_loss: 0.0095 - val_acc: 0.9973\n",
            "Q 697+705 T 1402 \u001b[92m☑\u001b[0m 1402\n",
            "Q 160+996 T 1156 \u001b[92m☑\u001b[0m 1156\n",
            "Q 55+105  T 160  \u001b[92m☑\u001b[0m 160 \n",
            "Q 192+611 T 803  \u001b[92m☑\u001b[0m 803 \n",
            "Q 354+77  T 431  \u001b[92m☑\u001b[0m 431 \n",
            "Q 404+928 T 1332 \u001b[92m☑\u001b[0m 1332\n",
            "Q 912+59  T 971  \u001b[92m☑\u001b[0m 971 \n",
            "Q 862+814 T 1676 \u001b[92m☑\u001b[0m 1676\n",
            "Q 687+999 T 1686 \u001b[92m☑\u001b[0m 1686\n",
            "Q 189+3   T 192  \u001b[92m☑\u001b[0m 192 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 70\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 303us/sample - loss: 0.0019 - acc: 1.0000 - val_loss: 0.0105 - val_acc: 0.9972\n",
            "Q 48+124  T 172  \u001b[92m☑\u001b[0m 172 \n",
            "Q 847+88  T 935  \u001b[92m☑\u001b[0m 935 \n",
            "Q 601+903 T 1504 \u001b[92m☑\u001b[0m 1504\n",
            "Q 70+678  T 748  \u001b[92m☑\u001b[0m 748 \n",
            "Q 55+139  T 194  \u001b[92m☑\u001b[0m 194 \n",
            "Q 211+120 T 331  \u001b[92m☑\u001b[0m 331 \n",
            "Q 92+52   T 144  \u001b[92m☑\u001b[0m 144 \n",
            "Q 47+400  T 447  \u001b[92m☑\u001b[0m 447 \n",
            "Q 220+98  T 318  \u001b[92m☑\u001b[0m 318 \n",
            "Q 78+798  T 876  \u001b[92m☑\u001b[0m 876 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 71\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 300us/sample - loss: 0.0019 - acc: 0.9999 - val_loss: 0.0094 - val_acc: 0.9972\n",
            "Q 6+184   T 190  \u001b[92m☑\u001b[0m 190 \n",
            "Q 83+894  T 977  \u001b[92m☑\u001b[0m 977 \n",
            "Q 438+0   T 438  \u001b[92m☑\u001b[0m 438 \n",
            "Q 628+99  T 727  \u001b[92m☑\u001b[0m 727 \n",
            "Q 439+610 T 1049 \u001b[92m☑\u001b[0m 1049\n",
            "Q 167+485 T 652  \u001b[92m☑\u001b[0m 652 \n",
            "Q 71+54   T 125  \u001b[92m☑\u001b[0m 125 \n",
            "Q 232+37  T 269  \u001b[92m☑\u001b[0m 269 \n",
            "Q 249+66  T 315  \u001b[92m☑\u001b[0m 315 \n",
            "Q 708+8   T 716  \u001b[92m☑\u001b[0m 716 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 72\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 310us/sample - loss: 0.0045 - acc: 0.9990 - val_loss: 0.1597 - val_acc: 0.9520\n",
            "Q 632+616 T 1248 \u001b[91m☒\u001b[0m 1147\n",
            "Q 2+178   T 180  \u001b[92m☑\u001b[0m 180 \n",
            "Q 7+90    T 97   \u001b[92m☑\u001b[0m 97  \n",
            "Q 239+38  T 277  \u001b[92m☑\u001b[0m 277 \n",
            "Q 326+59  T 385  \u001b[92m☑\u001b[0m 385 \n",
            "Q 379+968 T 1347 \u001b[92m☑\u001b[0m 1347\n",
            "Q 3+23    T 26   \u001b[92m☑\u001b[0m 26  \n",
            "Q 71+494  T 565  \u001b[92m☑\u001b[0m 565 \n",
            "Q 95+221  T 316  \u001b[92m☑\u001b[0m 316 \n",
            "Q 70+114  T 184  \u001b[92m☑\u001b[0m 184 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 73\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 306us/sample - loss: 0.0493 - acc: 0.9845 - val_loss: 0.0153 - val_acc: 0.9951\n",
            "Q 557+83  T 640  \u001b[92m☑\u001b[0m 640 \n",
            "Q 381+633 T 1014 \u001b[92m☑\u001b[0m 1014\n",
            "Q 48+43   T 91   \u001b[92m☑\u001b[0m 91  \n",
            "Q 564+42  T 606  \u001b[92m☑\u001b[0m 606 \n",
            "Q 310+77  T 387  \u001b[92m☑\u001b[0m 387 \n",
            "Q 225+2   T 227  \u001b[92m☑\u001b[0m 227 \n",
            "Q 94+393  T 487  \u001b[92m☑\u001b[0m 487 \n",
            "Q 348+270 T 618  \u001b[92m☑\u001b[0m 618 \n",
            "Q 93+41   T 134  \u001b[92m☑\u001b[0m 134 \n",
            "Q 112+743 T 855  \u001b[92m☑\u001b[0m 855 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 74\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 306us/sample - loss: 0.0034 - acc: 0.9997 - val_loss: 0.0100 - val_acc: 0.9971\n",
            "Q 665+630 T 1295 \u001b[92m☑\u001b[0m 1295\n",
            "Q 7+218   T 225  \u001b[92m☑\u001b[0m 225 \n",
            "Q 948+15  T 963  \u001b[92m☑\u001b[0m 963 \n",
            "Q 777+597 T 1374 \u001b[92m☑\u001b[0m 1374\n",
            "Q 754+498 T 1252 \u001b[92m☑\u001b[0m 1252\n",
            "Q 128+66  T 194  \u001b[92m☑\u001b[0m 194 \n",
            "Q 671+42  T 713  \u001b[92m☑\u001b[0m 713 \n",
            "Q 48+124  T 172  \u001b[92m☑\u001b[0m 172 \n",
            "Q 857+229 T 1086 \u001b[92m☑\u001b[0m 1086\n",
            "Q 205+29  T 234  \u001b[92m☑\u001b[0m 234 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 75\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 301us/sample - loss: 0.0019 - acc: 1.0000 - val_loss: 0.0089 - val_acc: 0.9974\n",
            "Q 970+39  T 1009 \u001b[92m☑\u001b[0m 1009\n",
            "Q 16+488  T 504  \u001b[92m☑\u001b[0m 504 \n",
            "Q 13+237  T 250  \u001b[92m☑\u001b[0m 250 \n",
            "Q 64+809  T 873  \u001b[92m☑\u001b[0m 873 \n",
            "Q 703+860 T 1563 \u001b[92m☑\u001b[0m 1563\n",
            "Q 61+198  T 259  \u001b[92m☑\u001b[0m 259 \n",
            "Q 52+63   T 115  \u001b[92m☑\u001b[0m 115 \n",
            "Q 173+54  T 227  \u001b[92m☑\u001b[0m 227 \n",
            "Q 382+83  T 465  \u001b[92m☑\u001b[0m 465 \n",
            "Q 97+894  T 991  \u001b[92m☑\u001b[0m 991 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 76\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 300us/sample - loss: 0.0016 - acc: 1.0000 - val_loss: 0.0088 - val_acc: 0.9976\n",
            "Q 605+0   T 605  \u001b[92m☑\u001b[0m 605 \n",
            "Q 35+449  T 484  \u001b[92m☑\u001b[0m 484 \n",
            "Q 833+921 T 1754 \u001b[92m☑\u001b[0m 1754\n",
            "Q 259+229 T 488  \u001b[92m☑\u001b[0m 488 \n",
            "Q 804+4   T 808  \u001b[92m☑\u001b[0m 808 \n",
            "Q 104+62  T 166  \u001b[92m☑\u001b[0m 166 \n",
            "Q 360+464 T 824  \u001b[92m☑\u001b[0m 824 \n",
            "Q 715+89  T 804  \u001b[92m☑\u001b[0m 804 \n",
            "Q 41+504  T 545  \u001b[92m☑\u001b[0m 545 \n",
            "Q 47+708  T 755  \u001b[92m☑\u001b[0m 755 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 77\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 300us/sample - loss: 0.0017 - acc: 0.9999 - val_loss: 0.0090 - val_acc: 0.9969\n",
            "Q 87+284  T 371  \u001b[92m☑\u001b[0m 371 \n",
            "Q 635+891 T 1526 \u001b[92m☑\u001b[0m 1526\n",
            "Q 9+340   T 349  \u001b[92m☑\u001b[0m 349 \n",
            "Q 29+581  T 610  \u001b[92m☑\u001b[0m 610 \n",
            "Q 28+375  T 403  \u001b[92m☑\u001b[0m 403 \n",
            "Q 471+5   T 476  \u001b[92m☑\u001b[0m 476 \n",
            "Q 990+6   T 996  \u001b[92m☑\u001b[0m 996 \n",
            "Q 38+93   T 131  \u001b[92m☑\u001b[0m 131 \n",
            "Q 19+1    T 20   \u001b[91m☒\u001b[0m 10  \n",
            "Q 770+199 T 969  \u001b[92m☑\u001b[0m 969 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 78\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 311us/sample - loss: 0.0369 - acc: 0.9888 - val_loss: 0.0395 - val_acc: 0.9869\n",
            "Q 694+93  T 787  \u001b[92m☑\u001b[0m 787 \n",
            "Q 730+992 T 1722 \u001b[92m☑\u001b[0m 1722\n",
            "Q 204+32  T 236  \u001b[92m☑\u001b[0m 236 \n",
            "Q 192+72  T 264  \u001b[92m☑\u001b[0m 264 \n",
            "Q 15+956  T 971  \u001b[92m☑\u001b[0m 971 \n",
            "Q 227+760 T 987  \u001b[92m☑\u001b[0m 987 \n",
            "Q 18+700  T 718  \u001b[92m☑\u001b[0m 718 \n",
            "Q 689+79  T 768  \u001b[91m☒\u001b[0m 758 \n",
            "Q 132+92  T 224  \u001b[92m☑\u001b[0m 224 \n",
            "Q 666+735 T 1401 \u001b[92m☑\u001b[0m 1401\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 79\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 304us/sample - loss: 0.0062 - acc: 0.9988 - val_loss: 0.0101 - val_acc: 0.9972\n",
            "Q 281+829 T 1110 \u001b[92m☑\u001b[0m 1110\n",
            "Q 91+0    T 91   \u001b[92m☑\u001b[0m 91  \n",
            "Q 35+958  T 993  \u001b[92m☑\u001b[0m 993 \n",
            "Q 916+20  T 936  \u001b[92m☑\u001b[0m 936 \n",
            "Q 52+71   T 123  \u001b[92m☑\u001b[0m 123 \n",
            "Q 180+216 T 396  \u001b[92m☑\u001b[0m 396 \n",
            "Q 583+65  T 648  \u001b[92m☑\u001b[0m 648 \n",
            "Q 51+527  T 578  \u001b[92m☑\u001b[0m 578 \n",
            "Q 606+40  T 646  \u001b[92m☑\u001b[0m 646 \n",
            "Q 24+52   T 76   \u001b[92m☑\u001b[0m 76  \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 80\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 307us/sample - loss: 0.0019 - acc: 1.0000 - val_loss: 0.0090 - val_acc: 0.9974\n",
            "Q 713+62  T 775  \u001b[92m☑\u001b[0m 775 \n",
            "Q 860+1   T 861  \u001b[92m☑\u001b[0m 861 \n",
            "Q 8+356   T 364  \u001b[92m☑\u001b[0m 364 \n",
            "Q 2+321   T 323  \u001b[92m☑\u001b[0m 323 \n",
            "Q 733+35  T 768  \u001b[92m☑\u001b[0m 768 \n",
            "Q 780+656 T 1436 \u001b[92m☑\u001b[0m 1436\n",
            "Q 0+826   T 826  \u001b[92m☑\u001b[0m 826 \n",
            "Q 77+42   T 119  \u001b[92m☑\u001b[0m 119 \n",
            "Q 51+527  T 578  \u001b[92m☑\u001b[0m 578 \n",
            "Q 66+923  T 989  \u001b[92m☑\u001b[0m 989 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 81\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 317us/sample - loss: 0.0015 - acc: 1.0000 - val_loss: 0.0082 - val_acc: 0.9977\n",
            "Q 137+826 T 963  \u001b[92m☑\u001b[0m 963 \n",
            "Q 875+53  T 928  \u001b[92m☑\u001b[0m 928 \n",
            "Q 9+995   T 1004 \u001b[92m☑\u001b[0m 1004\n",
            "Q 23+587  T 610  \u001b[92m☑\u001b[0m 610 \n",
            "Q 889+342 T 1231 \u001b[92m☑\u001b[0m 1231\n",
            "Q 68+263  T 331  \u001b[92m☑\u001b[0m 331 \n",
            "Q 0+663   T 663  \u001b[92m☑\u001b[0m 663 \n",
            "Q 5+503   T 508  \u001b[92m☑\u001b[0m 508 \n",
            "Q 588+64  T 652  \u001b[92m☑\u001b[0m 652 \n",
            "Q 127+212 T 339  \u001b[92m☑\u001b[0m 339 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 82\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 303us/sample - loss: 0.0013 - acc: 1.0000 - val_loss: 0.0078 - val_acc: 0.9979\n",
            "Q 5+789   T 794  \u001b[92m☑\u001b[0m 794 \n",
            "Q 4+412   T 416  \u001b[92m☑\u001b[0m 416 \n",
            "Q 21+608  T 629  \u001b[92m☑\u001b[0m 629 \n",
            "Q 16+485  T 501  \u001b[92m☑\u001b[0m 501 \n",
            "Q 11+485  T 496  \u001b[92m☑\u001b[0m 496 \n",
            "Q 2+351   T 353  \u001b[92m☑\u001b[0m 353 \n",
            "Q 334+81  T 415  \u001b[92m☑\u001b[0m 415 \n",
            "Q 23+267  T 290  \u001b[92m☑\u001b[0m 290 \n",
            "Q 91+283  T 374  \u001b[92m☑\u001b[0m 374 \n",
            "Q 57+117  T 174  \u001b[92m☑\u001b[0m 174 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 83\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 309us/sample - loss: 0.0012 - acc: 1.0000 - val_loss: 0.0078 - val_acc: 0.9978\n",
            "Q 742+195 T 937  \u001b[92m☑\u001b[0m 937 \n",
            "Q 4+412   T 416  \u001b[92m☑\u001b[0m 416 \n",
            "Q 505+427 T 932  \u001b[92m☑\u001b[0m 932 \n",
            "Q 2+233   T 235  \u001b[92m☑\u001b[0m 235 \n",
            "Q 653+931 T 1584 \u001b[92m☑\u001b[0m 1584\n",
            "Q 216+796 T 1012 \u001b[92m☑\u001b[0m 1012\n",
            "Q 86+518  T 604  \u001b[92m☑\u001b[0m 604 \n",
            "Q 727+142 T 869  \u001b[92m☑\u001b[0m 869 \n",
            "Q 1+61    T 62   \u001b[92m☑\u001b[0m 62  \n",
            "Q 621+462 T 1083 \u001b[92m☑\u001b[0m 1083\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 84\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 308us/sample - loss: 0.0426 - acc: 0.9875 - val_loss: 0.0646 - val_acc: 0.9783\n",
            "Q 196+451 T 647  \u001b[92m☑\u001b[0m 647 \n",
            "Q 71+945  T 1016 \u001b[92m☑\u001b[0m 1016\n",
            "Q 664+81  T 745  \u001b[92m☑\u001b[0m 745 \n",
            "Q 379+53  T 432  \u001b[92m☑\u001b[0m 432 \n",
            "Q 172+36  T 208  \u001b[92m☑\u001b[0m 208 \n",
            "Q 121+944 T 1065 \u001b[91m☒\u001b[0m 1055\n",
            "Q 83+915  T 998  \u001b[92m☑\u001b[0m 998 \n",
            "Q 30+241  T 271  \u001b[92m☑\u001b[0m 271 \n",
            "Q 990+822 T 1812 \u001b[92m☑\u001b[0m 1812\n",
            "Q 29+844  T 873  \u001b[92m☑\u001b[0m 873 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 85\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 316us/sample - loss: 0.0130 - acc: 0.9966 - val_loss: 0.0239 - val_acc: 0.9927\n",
            "Q 551+6   T 557  \u001b[92m☑\u001b[0m 557 \n",
            "Q 293+68  T 361  \u001b[92m☑\u001b[0m 361 \n",
            "Q 15+370  T 385  \u001b[92m☑\u001b[0m 385 \n",
            "Q 663+44  T 707  \u001b[92m☑\u001b[0m 707 \n",
            "Q 847+68  T 915  \u001b[92m☑\u001b[0m 915 \n",
            "Q 30+90   T 120  \u001b[92m☑\u001b[0m 120 \n",
            "Q 203+629 T 832  \u001b[92m☑\u001b[0m 832 \n",
            "Q 22+457  T 479  \u001b[92m☑\u001b[0m 479 \n",
            "Q 697+947 T 1644 \u001b[92m☑\u001b[0m 1644\n",
            "Q 605+390 T 995  \u001b[92m☑\u001b[0m 995 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 86\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 308us/sample - loss: 0.0054 - acc: 0.9989 - val_loss: 0.0098 - val_acc: 0.9973\n",
            "Q 899+80  T 979  \u001b[92m☑\u001b[0m 979 \n",
            "Q 99+69   T 168  \u001b[92m☑\u001b[0m 168 \n",
            "Q 124+931 T 1055 \u001b[92m☑\u001b[0m 1055\n",
            "Q 948+41  T 989  \u001b[92m☑\u001b[0m 989 \n",
            "Q 258+22  T 280  \u001b[92m☑\u001b[0m 280 \n",
            "Q 960+474 T 1434 \u001b[92m☑\u001b[0m 1434\n",
            "Q 121+4   T 125  \u001b[92m☑\u001b[0m 125 \n",
            "Q 521+5   T 526  \u001b[92m☑\u001b[0m 526 \n",
            "Q 28+502  T 530  \u001b[92m☑\u001b[0m 530 \n",
            "Q 210+18  T 228  \u001b[92m☑\u001b[0m 228 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 87\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 311us/sample - loss: 0.0016 - acc: 1.0000 - val_loss: 0.0087 - val_acc: 0.9972\n",
            "Q 597+444 T 1041 \u001b[92m☑\u001b[0m 1041\n",
            "Q 768+45  T 813  \u001b[92m☑\u001b[0m 813 \n",
            "Q 397+6   T 403  \u001b[92m☑\u001b[0m 403 \n",
            "Q 641+6   T 647  \u001b[92m☑\u001b[0m 647 \n",
            "Q 34+554  T 588  \u001b[92m☑\u001b[0m 588 \n",
            "Q 137+826 T 963  \u001b[92m☑\u001b[0m 963 \n",
            "Q 393+1   T 394  \u001b[92m☑\u001b[0m 394 \n",
            "Q 245+56  T 301  \u001b[92m☑\u001b[0m 301 \n",
            "Q 224+6   T 230  \u001b[92m☑\u001b[0m 230 \n",
            "Q 30+37   T 67   \u001b[92m☑\u001b[0m 67  \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 88\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 316us/sample - loss: 0.0012 - acc: 1.0000 - val_loss: 0.0076 - val_acc: 0.9977\n",
            "Q 883+230 T 1113 \u001b[92m☑\u001b[0m 1113\n",
            "Q 167+656 T 823  \u001b[92m☑\u001b[0m 823 \n",
            "Q 60+976  T 1036 \u001b[92m☑\u001b[0m 1036\n",
            "Q 957+462 T 1419 \u001b[92m☑\u001b[0m 1419\n",
            "Q 356+608 T 964  \u001b[92m☑\u001b[0m 964 \n",
            "Q 21+630  T 651  \u001b[92m☑\u001b[0m 651 \n",
            "Q 446+93  T 539  \u001b[92m☑\u001b[0m 539 \n",
            "Q 61+322  T 383  \u001b[92m☑\u001b[0m 383 \n",
            "Q 71+25   T 96   \u001b[92m☑\u001b[0m 96  \n",
            "Q 377+24  T 401  \u001b[92m☑\u001b[0m 401 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 89\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 309us/sample - loss: 0.0011 - acc: 1.0000 - val_loss: 0.0076 - val_acc: 0.9975\n",
            "Q 641+6   T 647  \u001b[92m☑\u001b[0m 647 \n",
            "Q 476+44  T 520  \u001b[92m☑\u001b[0m 520 \n",
            "Q 21+513  T 534  \u001b[92m☑\u001b[0m 534 \n",
            "Q 606+275 T 881  \u001b[92m☑\u001b[0m 881 \n",
            "Q 24+335  T 359  \u001b[92m☑\u001b[0m 359 \n",
            "Q 617+906 T 1523 \u001b[92m☑\u001b[0m 1523\n",
            "Q 53+828  T 881  \u001b[92m☑\u001b[0m 881 \n",
            "Q 578+367 T 945  \u001b[92m☑\u001b[0m 945 \n",
            "Q 52+992  T 1044 \u001b[92m☑\u001b[0m 1044\n",
            "Q 559+66  T 625  \u001b[92m☑\u001b[0m 625 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 90\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 309us/sample - loss: 9.6884e-04 - acc: 1.0000 - val_loss: 0.0070 - val_acc: 0.9980\n",
            "Q 37+863  T 900  \u001b[92m☑\u001b[0m 900 \n",
            "Q 47+571  T 618  \u001b[92m☑\u001b[0m 618 \n",
            "Q 24+52   T 76   \u001b[92m☑\u001b[0m 76  \n",
            "Q 1+61    T 62   \u001b[92m☑\u001b[0m 62  \n",
            "Q 903+2   T 905  \u001b[92m☑\u001b[0m 905 \n",
            "Q 48+180  T 228  \u001b[92m☑\u001b[0m 228 \n",
            "Q 27+792  T 819  \u001b[92m☑\u001b[0m 819 \n",
            "Q 37+28   T 65   \u001b[92m☑\u001b[0m 65  \n",
            "Q 47+932  T 979  \u001b[92m☑\u001b[0m 979 \n",
            "Q 71+77   T 148  \u001b[92m☑\u001b[0m 148 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 91\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 315us/sample - loss: 0.0010 - acc: 0.9999 - val_loss: 0.0126 - val_acc: 0.9959\n",
            "Q 861+33  T 894  \u001b[92m☑\u001b[0m 894 \n",
            "Q 99+510  T 609  \u001b[92m☑\u001b[0m 609 \n",
            "Q 230+8   T 238  \u001b[92m☑\u001b[0m 238 \n",
            "Q 68+585  T 653  \u001b[92m☑\u001b[0m 653 \n",
            "Q 63+72   T 135  \u001b[92m☑\u001b[0m 135 \n",
            "Q 588+46  T 634  \u001b[92m☑\u001b[0m 634 \n",
            "Q 93+595  T 688  \u001b[92m☑\u001b[0m 688 \n",
            "Q 48+853  T 901  \u001b[92m☑\u001b[0m 901 \n",
            "Q 94+393  T 487  \u001b[92m☑\u001b[0m 487 \n",
            "Q 949+779 T 1728 \u001b[92m☑\u001b[0m 1728\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 92\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 316us/sample - loss: 0.0569 - acc: 0.9830 - val_loss: 0.0159 - val_acc: 0.9947\n",
            "Q 63+384  T 447  \u001b[92m☑\u001b[0m 447 \n",
            "Q 446+332 T 778  \u001b[92m☑\u001b[0m 778 \n",
            "Q 697+893 T 1590 \u001b[92m☑\u001b[0m 1590\n",
            "Q 105+829 T 934  \u001b[92m☑\u001b[0m 934 \n",
            "Q 147+996 T 1143 \u001b[92m☑\u001b[0m 1143\n",
            "Q 61+98   T 159  \u001b[92m☑\u001b[0m 159 \n",
            "Q 661+5   T 666  \u001b[92m☑\u001b[0m 666 \n",
            "Q 195+582 T 777  \u001b[92m☑\u001b[0m 777 \n",
            "Q 522+90  T 612  \u001b[92m☑\u001b[0m 612 \n",
            "Q 69+63   T 132  \u001b[92m☑\u001b[0m 132 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 93\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 307us/sample - loss: 0.0029 - acc: 0.9997 - val_loss: 0.0088 - val_acc: 0.9976\n",
            "Q 814+26  T 840  \u001b[92m☑\u001b[0m 840 \n",
            "Q 49+683  T 732  \u001b[92m☑\u001b[0m 732 \n",
            "Q 829+49  T 878  \u001b[92m☑\u001b[0m 878 \n",
            "Q 62+129  T 191  \u001b[92m☑\u001b[0m 191 \n",
            "Q 227+91  T 318  \u001b[92m☑\u001b[0m 318 \n",
            "Q 538+902 T 1440 \u001b[92m☑\u001b[0m 1440\n",
            "Q 791+3   T 794  \u001b[92m☑\u001b[0m 794 \n",
            "Q 657+517 T 1174 \u001b[92m☑\u001b[0m 1174\n",
            "Q 145+414 T 559  \u001b[92m☑\u001b[0m 559 \n",
            "Q 411+611 T 1022 \u001b[92m☑\u001b[0m 1022\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 94\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 309us/sample - loss: 0.0015 - acc: 1.0000 - val_loss: 0.0080 - val_acc: 0.9977\n",
            "Q 883+89  T 972  \u001b[92m☑\u001b[0m 972 \n",
            "Q 195+195 T 390  \u001b[92m☑\u001b[0m 390 \n",
            "Q 268+26  T 294  \u001b[92m☑\u001b[0m 294 \n",
            "Q 474+611 T 1085 \u001b[92m☑\u001b[0m 1085\n",
            "Q 835+969 T 1804 \u001b[91m☒\u001b[0m 1704\n",
            "Q 356+608 T 964  \u001b[92m☑\u001b[0m 964 \n",
            "Q 551+6   T 557  \u001b[92m☑\u001b[0m 557 \n",
            "Q 640+619 T 1259 \u001b[92m☑\u001b[0m 1259\n",
            "Q 38+58   T 96   \u001b[92m☑\u001b[0m 96  \n",
            "Q 229+154 T 383  \u001b[92m☑\u001b[0m 383 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 95\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 319us/sample - loss: 0.0012 - acc: 1.0000 - val_loss: 0.0071 - val_acc: 0.9981\n",
            "Q 924+319 T 1243 \u001b[92m☑\u001b[0m 1243\n",
            "Q 702+756 T 1458 \u001b[92m☑\u001b[0m 1458\n",
            "Q 58+138  T 196  \u001b[92m☑\u001b[0m 196 \n",
            "Q 43+28   T 71   \u001b[92m☑\u001b[0m 71  \n",
            "Q 91+930  T 1021 \u001b[92m☑\u001b[0m 1021\n",
            "Q 198+14  T 212  \u001b[92m☑\u001b[0m 212 \n",
            "Q 229+154 T 383  \u001b[92m☑\u001b[0m 383 \n",
            "Q 522+3   T 525  \u001b[92m☑\u001b[0m 525 \n",
            "Q 48+937  T 985  \u001b[92m☑\u001b[0m 985 \n",
            "Q 28+933  T 961  \u001b[92m☑\u001b[0m 961 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 96\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 319us/sample - loss: 0.0010 - acc: 1.0000 - val_loss: 0.0074 - val_acc: 0.9979\n",
            "Q 907+3   T 910  \u001b[92m☑\u001b[0m 910 \n",
            "Q 94+981  T 1075 \u001b[92m☑\u001b[0m 1075\n",
            "Q 30+490  T 520  \u001b[92m☑\u001b[0m 520 \n",
            "Q 649+554 T 1203 \u001b[92m☑\u001b[0m 1203\n",
            "Q 88+90   T 178  \u001b[92m☑\u001b[0m 178 \n",
            "Q 803+695 T 1498 \u001b[91m☒\u001b[0m 1598\n",
            "Q 764+151 T 915  \u001b[92m☑\u001b[0m 915 \n",
            "Q 33+58   T 91   \u001b[92m☑\u001b[0m 91  \n",
            "Q 84+39   T 123  \u001b[92m☑\u001b[0m 123 \n",
            "Q 862+31  T 893  \u001b[92m☑\u001b[0m 893 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 97\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 323us/sample - loss: 9.3772e-04 - acc: 1.0000 - val_loss: 0.0069 - val_acc: 0.9981\n",
            "Q 373+18  T 391  \u001b[92m☑\u001b[0m 391 \n",
            "Q 19+981  T 1000 \u001b[92m☑\u001b[0m 1000\n",
            "Q 55+52   T 107  \u001b[92m☑\u001b[0m 107 \n",
            "Q 604+728 T 1332 \u001b[92m☑\u001b[0m 1332\n",
            "Q 440+471 T 911  \u001b[92m☑\u001b[0m 911 \n",
            "Q 524+382 T 906  \u001b[92m☑\u001b[0m 906 \n",
            "Q 20+301  T 321  \u001b[92m☑\u001b[0m 321 \n",
            "Q 734+208 T 942  \u001b[92m☑\u001b[0m 942 \n",
            "Q 472+515 T 987  \u001b[92m☑\u001b[0m 987 \n",
            "Q 177+707 T 884  \u001b[92m☑\u001b[0m 884 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 98\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 323us/sample - loss: 0.0011 - acc: 0.9999 - val_loss: 0.0113 - val_acc: 0.9963\n",
            "Q 451+64  T 515  \u001b[92m☑\u001b[0m 515 \n",
            "Q 855+79  T 934  \u001b[92m☑\u001b[0m 934 \n",
            "Q 607+9   T 616  \u001b[92m☑\u001b[0m 616 \n",
            "Q 635+58  T 693  \u001b[92m☑\u001b[0m 693 \n",
            "Q 624+29  T 653  \u001b[92m☑\u001b[0m 653 \n",
            "Q 2+642   T 644  \u001b[92m☑\u001b[0m 644 \n",
            "Q 569+27  T 596  \u001b[92m☑\u001b[0m 596 \n",
            "Q 276+0   T 276  \u001b[92m☑\u001b[0m 276 \n",
            "Q 978+99  T 1077 \u001b[92m☑\u001b[0m 1077\n",
            "Q 70+665  T 735  \u001b[92m☑\u001b[0m 735 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 99\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 0.0362 - acc: 0.9884 - val_loss: 0.0188 - val_acc: 0.9939\n",
            "Q 76+159  T 235  \u001b[92m☑\u001b[0m 235 \n",
            "Q 45+44   T 89   \u001b[92m☑\u001b[0m 89  \n",
            "Q 22+243  T 265  \u001b[92m☑\u001b[0m 265 \n",
            "Q 592+41  T 633  \u001b[92m☑\u001b[0m 633 \n",
            "Q 339+109 T 448  \u001b[92m☑\u001b[0m 448 \n",
            "Q 83+451  T 534  \u001b[92m☑\u001b[0m 534 \n",
            "Q 47+133  T 180  \u001b[92m☑\u001b[0m 180 \n",
            "Q 6+263   T 269  \u001b[92m☑\u001b[0m 269 \n",
            "Q 872+29  T 901  \u001b[92m☑\u001b[0m 901 \n",
            "Q 42+531  T 573  \u001b[92m☑\u001b[0m 573 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 100\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 0.0028 - acc: 0.9997 - val_loss: 0.0090 - val_acc: 0.9970\n",
            "Q 77+15   T 92   \u001b[92m☑\u001b[0m 92  \n",
            "Q 77+288  T 365  \u001b[92m☑\u001b[0m 365 \n",
            "Q 14+751  T 765  \u001b[92m☑\u001b[0m 765 \n",
            "Q 8+202   T 210  \u001b[92m☑\u001b[0m 210 \n",
            "Q 337+2   T 339  \u001b[92m☑\u001b[0m 339 \n",
            "Q 740+91  T 831  \u001b[92m☑\u001b[0m 831 \n",
            "Q 5+948   T 953  \u001b[92m☑\u001b[0m 953 \n",
            "Q 859+76  T 935  \u001b[92m☑\u001b[0m 935 \n",
            "Q 429+894 T 1323 \u001b[92m☑\u001b[0m 1323\n",
            "Q 122+60  T 182  \u001b[92m☑\u001b[0m 182 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 101\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 322us/sample - loss: 0.0014 - acc: 0.9999 - val_loss: 0.0081 - val_acc: 0.9973\n",
            "Q 136+422 T 558  \u001b[92m☑\u001b[0m 558 \n",
            "Q 714+82  T 796  \u001b[92m☑\u001b[0m 796 \n",
            "Q 479+794 T 1273 \u001b[92m☑\u001b[0m 1273\n",
            "Q 23+45   T 68   \u001b[92m☑\u001b[0m 68  \n",
            "Q 93+854  T 947  \u001b[92m☑\u001b[0m 947 \n",
            "Q 767+314 T 1081 \u001b[92m☑\u001b[0m 1081\n",
            "Q 180+216 T 396  \u001b[92m☑\u001b[0m 396 \n",
            "Q 86+248  T 334  \u001b[92m☑\u001b[0m 334 \n",
            "Q 23+861  T 884  \u001b[92m☑\u001b[0m 884 \n",
            "Q 362+92  T 454  \u001b[92m☑\u001b[0m 454 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 102\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 9.7463e-04 - acc: 1.0000 - val_loss: 0.0071 - val_acc: 0.9977\n",
            "Q 994+572 T 1566 \u001b[92m☑\u001b[0m 1566\n",
            "Q 657+4   T 661  \u001b[92m☑\u001b[0m 661 \n",
            "Q 69+81   T 150  \u001b[92m☑\u001b[0m 150 \n",
            "Q 897+895 T 1792 \u001b[92m☑\u001b[0m 1792\n",
            "Q 89+404  T 493  \u001b[92m☑\u001b[0m 493 \n",
            "Q 66+784  T 850  \u001b[92m☑\u001b[0m 850 \n",
            "Q 354+77  T 431  \u001b[92m☑\u001b[0m 431 \n",
            "Q 36+800  T 836  \u001b[92m☑\u001b[0m 836 \n",
            "Q 42+290  T 332  \u001b[92m☑\u001b[0m 332 \n",
            "Q 149+928 T 1077 \u001b[92m☑\u001b[0m 1077\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 103\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 320us/sample - loss: 8.5005e-04 - acc: 1.0000 - val_loss: 0.0084 - val_acc: 0.9974\n",
            "Q 876+3   T 879  \u001b[92m☑\u001b[0m 879 \n",
            "Q 833+96  T 929  \u001b[92m☑\u001b[0m 929 \n",
            "Q 58+415  T 473  \u001b[92m☑\u001b[0m 473 \n",
            "Q 872+29  T 901  \u001b[92m☑\u001b[0m 901 \n",
            "Q 372+91  T 463  \u001b[92m☑\u001b[0m 463 \n",
            "Q 33+912  T 945  \u001b[92m☑\u001b[0m 945 \n",
            "Q 47+798  T 845  \u001b[92m☑\u001b[0m 845 \n",
            "Q 251+17  T 268  \u001b[92m☑\u001b[0m 268 \n",
            "Q 179+27  T 206  \u001b[92m☑\u001b[0m 206 \n",
            "Q 330+21  T 351  \u001b[92m☑\u001b[0m 351 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 104\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 323us/sample - loss: 0.0010 - acc: 0.9999 - val_loss: 0.0105 - val_acc: 0.9968\n",
            "Q 594+210 T 804  \u001b[91m☒\u001b[0m 704 \n",
            "Q 2+991   T 993  \u001b[92m☑\u001b[0m 993 \n",
            "Q 89+7    T 96   \u001b[92m☑\u001b[0m 96  \n",
            "Q 60+494  T 554  \u001b[92m☑\u001b[0m 554 \n",
            "Q 863+495 T 1358 \u001b[92m☑\u001b[0m 1358\n",
            "Q 3+950   T 953  \u001b[92m☑\u001b[0m 953 \n",
            "Q 216+37  T 253  \u001b[92m☑\u001b[0m 253 \n",
            "Q 67+688  T 755  \u001b[92m☑\u001b[0m 755 \n",
            "Q 14+901  T 915  \u001b[92m☑\u001b[0m 915 \n",
            "Q 214+110 T 324  \u001b[92m☑\u001b[0m 324 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 105\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 328us/sample - loss: 0.0367 - acc: 0.9887 - val_loss: 0.0322 - val_acc: 0.9907\n",
            "Q 42+949  T 991  \u001b[92m☑\u001b[0m 991 \n",
            "Q 651+1   T 652  \u001b[92m☑\u001b[0m 652 \n",
            "Q 37+81   T 118  \u001b[92m☑\u001b[0m 118 \n",
            "Q 587+50  T 637  \u001b[92m☑\u001b[0m 637 \n",
            "Q 7+499   T 506  \u001b[92m☑\u001b[0m 506 \n",
            "Q 207+169 T 376  \u001b[92m☑\u001b[0m 376 \n",
            "Q 564+42  T 606  \u001b[92m☑\u001b[0m 606 \n",
            "Q 1+60    T 61   \u001b[92m☑\u001b[0m 61  \n",
            "Q 470+9   T 479  \u001b[92m☑\u001b[0m 479 \n",
            "Q 182+497 T 679  \u001b[91m☒\u001b[0m 689 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 106\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 322us/sample - loss: 0.0041 - acc: 0.9992 - val_loss: 0.0093 - val_acc: 0.9974\n",
            "Q 383+53  T 436  \u001b[92m☑\u001b[0m 436 \n",
            "Q 49+52   T 101  \u001b[92m☑\u001b[0m 101 \n",
            "Q 29+608  T 637  \u001b[92m☑\u001b[0m 637 \n",
            "Q 593+84  T 677  \u001b[92m☑\u001b[0m 677 \n",
            "Q 32+432  T 464  \u001b[92m☑\u001b[0m 464 \n",
            "Q 56+743  T 799  \u001b[92m☑\u001b[0m 799 \n",
            "Q 55+866  T 921  \u001b[92m☑\u001b[0m 921 \n",
            "Q 718+56  T 774  \u001b[92m☑\u001b[0m 774 \n",
            "Q 5+445   T 450  \u001b[92m☑\u001b[0m 450 \n",
            "Q 19+425  T 444  \u001b[92m☑\u001b[0m 444 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 107\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 325us/sample - loss: 0.0012 - acc: 1.0000 - val_loss: 0.0075 - val_acc: 0.9979\n",
            "Q 85+385  T 470  \u001b[92m☑\u001b[0m 470 \n",
            "Q 803+65  T 868  \u001b[92m☑\u001b[0m 868 \n",
            "Q 408+7   T 415  \u001b[92m☑\u001b[0m 415 \n",
            "Q 74+90   T 164  \u001b[92m☑\u001b[0m 164 \n",
            "Q 227+91  T 318  \u001b[92m☑\u001b[0m 318 \n",
            "Q 104+5   T 109  \u001b[92m☑\u001b[0m 109 \n",
            "Q 38+429  T 467  \u001b[92m☑\u001b[0m 467 \n",
            "Q 529+41  T 570  \u001b[92m☑\u001b[0m 570 \n",
            "Q 639+78  T 717  \u001b[92m☑\u001b[0m 717 \n",
            "Q 55+575  T 630  \u001b[92m☑\u001b[0m 630 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 108\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 8.6570e-04 - acc: 1.0000 - val_loss: 0.0070 - val_acc: 0.9980\n",
            "Q 79+253  T 332  \u001b[92m☑\u001b[0m 332 \n",
            "Q 461+278 T 739  \u001b[92m☑\u001b[0m 739 \n",
            "Q 393+1   T 394  \u001b[92m☑\u001b[0m 394 \n",
            "Q 428+4   T 432  \u001b[92m☑\u001b[0m 432 \n",
            "Q 625+68  T 693  \u001b[92m☑\u001b[0m 693 \n",
            "Q 17+501  T 518  \u001b[92m☑\u001b[0m 518 \n",
            "Q 604+175 T 779  \u001b[92m☑\u001b[0m 779 \n",
            "Q 549+4   T 553  \u001b[92m☑\u001b[0m 553 \n",
            "Q 273+56  T 329  \u001b[92m☑\u001b[0m 329 \n",
            "Q 13+998  T 1011 \u001b[92m☑\u001b[0m 1011\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 109\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 323us/sample - loss: 7.5224e-04 - acc: 1.0000 - val_loss: 0.0069 - val_acc: 0.9980\n",
            "Q 510+76  T 586  \u001b[92m☑\u001b[0m 586 \n",
            "Q 713+18  T 731  \u001b[92m☑\u001b[0m 731 \n",
            "Q 539+916 T 1455 \u001b[92m☑\u001b[0m 1455\n",
            "Q 285+660 T 945  \u001b[92m☑\u001b[0m 945 \n",
            "Q 9+340   T 349  \u001b[92m☑\u001b[0m 349 \n",
            "Q 62+609  T 671  \u001b[92m☑\u001b[0m 671 \n",
            "Q 653+34  T 687  \u001b[92m☑\u001b[0m 687 \n",
            "Q 3+614   T 617  \u001b[92m☑\u001b[0m 617 \n",
            "Q 9+995   T 1004 \u001b[92m☑\u001b[0m 1004\n",
            "Q 421+49  T 470  \u001b[92m☑\u001b[0m 470 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 110\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 6.9558e-04 - acc: 1.0000 - val_loss: 0.0067 - val_acc: 0.9981\n",
            "Q 249+657 T 906  \u001b[92m☑\u001b[0m 906 \n",
            "Q 327+31  T 358  \u001b[92m☑\u001b[0m 358 \n",
            "Q 3+353   T 356  \u001b[92m☑\u001b[0m 356 \n",
            "Q 6+805   T 811  \u001b[92m☑\u001b[0m 811 \n",
            "Q 725+3   T 728  \u001b[92m☑\u001b[0m 728 \n",
            "Q 61+665  T 726  \u001b[92m☑\u001b[0m 726 \n",
            "Q 975+0   T 975  \u001b[92m☑\u001b[0m 975 \n",
            "Q 223+89  T 312  \u001b[92m☑\u001b[0m 312 \n",
            "Q 28+877  T 905  \u001b[92m☑\u001b[0m 905 \n",
            "Q 317+973 T 1290 \u001b[92m☑\u001b[0m 1290\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 111\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 325us/sample - loss: 6.2256e-04 - acc: 1.0000 - val_loss: 0.0074 - val_acc: 0.9974\n",
            "Q 811+30  T 841  \u001b[92m☑\u001b[0m 841 \n",
            "Q 367+81  T 448  \u001b[92m☑\u001b[0m 448 \n",
            "Q 214+930 T 1144 \u001b[92m☑\u001b[0m 1144\n",
            "Q 896+34  T 930  \u001b[92m☑\u001b[0m 930 \n",
            "Q 324+48  T 372  \u001b[92m☑\u001b[0m 372 \n",
            "Q 98+614  T 712  \u001b[92m☑\u001b[0m 712 \n",
            "Q 56+473  T 529  \u001b[92m☑\u001b[0m 529 \n",
            "Q 66+723  T 789  \u001b[92m☑\u001b[0m 789 \n",
            "Q 606+275 T 881  \u001b[92m☑\u001b[0m 881 \n",
            "Q 1+793   T 794  \u001b[92m☑\u001b[0m 794 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 112\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 330us/sample - loss: 0.0400 - acc: 0.9881 - val_loss: 0.0242 - val_acc: 0.9928\n",
            "Q 895+759 T 1654 \u001b[92m☑\u001b[0m 1654\n",
            "Q 75+382  T 457  \u001b[91m☒\u001b[0m 456 \n",
            "Q 651+0   T 651  \u001b[92m☑\u001b[0m 651 \n",
            "Q 19+93   T 112  \u001b[92m☑\u001b[0m 112 \n",
            "Q 78+899  T 977  \u001b[92m☑\u001b[0m 977 \n",
            "Q 999+78  T 1077 \u001b[92m☑\u001b[0m 1077\n",
            "Q 413+2   T 415  \u001b[92m☑\u001b[0m 415 \n",
            "Q 30+875  T 905  \u001b[92m☑\u001b[0m 905 \n",
            "Q 848+201 T 1049 \u001b[91m☒\u001b[0m 900 \n",
            "Q 6+112   T 118  \u001b[92m☑\u001b[0m 118 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 113\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 326us/sample - loss: 0.0041 - acc: 0.9991 - val_loss: 0.0085 - val_acc: 0.9976\n",
            "Q 975+63  T 1038 \u001b[92m☑\u001b[0m 1038\n",
            "Q 598+351 T 949  \u001b[92m☑\u001b[0m 949 \n",
            "Q 855+23  T 878  \u001b[92m☑\u001b[0m 878 \n",
            "Q 15+957  T 972  \u001b[92m☑\u001b[0m 972 \n",
            "Q 4+818   T 822  \u001b[92m☑\u001b[0m 822 \n",
            "Q 22+591  T 613  \u001b[92m☑\u001b[0m 613 \n",
            "Q 805+888 T 1693 \u001b[92m☑\u001b[0m 1693\n",
            "Q 789+38  T 827  \u001b[92m☑\u001b[0m 827 \n",
            "Q 979+3   T 982  \u001b[92m☑\u001b[0m 982 \n",
            "Q 862+814 T 1676 \u001b[92m☑\u001b[0m 1676\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 114\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 324us/sample - loss: 0.0011 - acc: 1.0000 - val_loss: 0.0071 - val_acc: 0.9977\n",
            "Q 9+932   T 941  \u001b[92m☑\u001b[0m 941 \n",
            "Q 11+924  T 935  \u001b[92m☑\u001b[0m 935 \n",
            "Q 270+821 T 1091 \u001b[92m☑\u001b[0m 1091\n",
            "Q 215+617 T 832  \u001b[92m☑\u001b[0m 832 \n",
            "Q 933+405 T 1338 \u001b[92m☑\u001b[0m 1338\n",
            "Q 187+44  T 231  \u001b[92m☑\u001b[0m 231 \n",
            "Q 12+671  T 683  \u001b[92m☑\u001b[0m 683 \n",
            "Q 733+138 T 871  \u001b[92m☑\u001b[0m 871 \n",
            "Q 34+143  T 177  \u001b[92m☑\u001b[0m 177 \n",
            "Q 913+49  T 962  \u001b[92m☑\u001b[0m 962 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 115\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 326us/sample - loss: 8.2069e-04 - acc: 1.0000 - val_loss: 0.0065 - val_acc: 0.9981\n",
            "Q 845+5   T 850  \u001b[92m☑\u001b[0m 850 \n",
            "Q 97+1    T 98   \u001b[92m☑\u001b[0m 98  \n",
            "Q 602+853 T 1455 \u001b[92m☑\u001b[0m 1455\n",
            "Q 105+91  T 196  \u001b[92m☑\u001b[0m 196 \n",
            "Q 690+72  T 762  \u001b[92m☑\u001b[0m 762 \n",
            "Q 289+92  T 381  \u001b[92m☑\u001b[0m 381 \n",
            "Q 393+7   T 400  \u001b[92m☑\u001b[0m 400 \n",
            "Q 7+85    T 92   \u001b[92m☑\u001b[0m 92  \n",
            "Q 352+93  T 445  \u001b[92m☑\u001b[0m 445 \n",
            "Q 72+590  T 662  \u001b[92m☑\u001b[0m 662 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 116\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 321us/sample - loss: 7.0687e-04 - acc: 1.0000 - val_loss: 0.0063 - val_acc: 0.9979\n",
            "Q 685+42  T 727  \u001b[92m☑\u001b[0m 727 \n",
            "Q 685+42  T 727  \u001b[92m☑\u001b[0m 727 \n",
            "Q 672+398 T 1070 \u001b[92m☑\u001b[0m 1070\n",
            "Q 69+785  T 854  \u001b[92m☑\u001b[0m 854 \n",
            "Q 30+37   T 67   \u001b[92m☑\u001b[0m 67  \n",
            "Q 543+13  T 556  \u001b[92m☑\u001b[0m 556 \n",
            "Q 19+895  T 914  \u001b[92m☑\u001b[0m 914 \n",
            "Q 0+439   T 439  \u001b[92m☑\u001b[0m 439 \n",
            "Q 85+840  T 925  \u001b[92m☑\u001b[0m 925 \n",
            "Q 661+608 T 1269 \u001b[92m☑\u001b[0m 1269\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 117\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 325us/sample - loss: 6.1916e-04 - acc: 1.0000 - val_loss: 0.0062 - val_acc: 0.9983\n",
            "Q 606+40  T 646  \u001b[92m☑\u001b[0m 646 \n",
            "Q 5+801   T 806  \u001b[92m☑\u001b[0m 806 \n",
            "Q 115+595 T 710  \u001b[92m☑\u001b[0m 710 \n",
            "Q 504+68  T 572  \u001b[92m☑\u001b[0m 572 \n",
            "Q 10+74   T 84   \u001b[92m☑\u001b[0m 84  \n",
            "Q 249+98  T 347  \u001b[92m☑\u001b[0m 347 \n",
            "Q 77+68   T 145  \u001b[92m☑\u001b[0m 145 \n",
            "Q 314+9   T 323  \u001b[92m☑\u001b[0m 323 \n",
            "Q 238+96  T 334  \u001b[92m☑\u001b[0m 334 \n",
            "Q 9+189   T 198  \u001b[92m☑\u001b[0m 198 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 118\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 329us/sample - loss: 5.6821e-04 - acc: 1.0000 - val_loss: 0.0065 - val_acc: 0.9976\n",
            "Q 2+187   T 189  \u001b[92m☑\u001b[0m 189 \n",
            "Q 364+8   T 372  \u001b[92m☑\u001b[0m 372 \n",
            "Q 97+88   T 185  \u001b[92m☑\u001b[0m 185 \n",
            "Q 273+147 T 420  \u001b[91m☒\u001b[0m 410 \n",
            "Q 37+81   T 118  \u001b[92m☑\u001b[0m 118 \n",
            "Q 99+33   T 132  \u001b[92m☑\u001b[0m 132 \n",
            "Q 60+399  T 459  \u001b[92m☑\u001b[0m 459 \n",
            "Q 566+94  T 660  \u001b[92m☑\u001b[0m 660 \n",
            "Q 56+572  T 628  \u001b[92m☑\u001b[0m 628 \n",
            "Q 116+900 T 1016 \u001b[92m☑\u001b[0m 1016\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 119\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 323us/sample - loss: 0.0280 - acc: 0.9911 - val_loss: 0.0372 - val_acc: 0.9879\n",
            "Q 41+38   T 79   \u001b[92m☑\u001b[0m 79  \n",
            "Q 545+36  T 581  \u001b[92m☑\u001b[0m 581 \n",
            "Q 978+99  T 1077 \u001b[92m☑\u001b[0m 1077\n",
            "Q 22+570  T 592  \u001b[92m☑\u001b[0m 592 \n",
            "Q 25+589  T 614  \u001b[92m☑\u001b[0m 614 \n",
            "Q 96+869  T 965  \u001b[92m☑\u001b[0m 965 \n",
            "Q 60+293  T 353  \u001b[92m☑\u001b[0m 353 \n",
            "Q 825+748 T 1573 \u001b[92m☑\u001b[0m 1573\n",
            "Q 29+844  T 873  \u001b[92m☑\u001b[0m 873 \n",
            "Q 7+665   T 672  \u001b[92m☑\u001b[0m 672 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 120\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 322us/sample - loss: 0.0045 - acc: 0.9990 - val_loss: 0.0141 - val_acc: 0.9963\n",
            "Q 624+366 T 990  \u001b[92m☑\u001b[0m 990 \n",
            "Q 8+382   T 390  \u001b[92m☑\u001b[0m 390 \n",
            "Q 11+496  T 507  \u001b[92m☑\u001b[0m 507 \n",
            "Q 54+816  T 870  \u001b[92m☑\u001b[0m 870 \n",
            "Q 38+988  T 1026 \u001b[92m☑\u001b[0m 1026\n",
            "Q 27+897  T 924  \u001b[92m☑\u001b[0m 924 \n",
            "Q 70+646  T 716  \u001b[92m☑\u001b[0m 716 \n",
            "Q 26+932  T 958  \u001b[92m☑\u001b[0m 958 \n",
            "Q 763+265 T 1028 \u001b[92m☑\u001b[0m 1028\n",
            "Q 76+56   T 132  \u001b[92m☑\u001b[0m 132 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 121\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 329us/sample - loss: 0.0103 - acc: 0.9973 - val_loss: 0.0122 - val_acc: 0.9961\n",
            "Q 94+192  T 286  \u001b[92m☑\u001b[0m 286 \n",
            "Q 528+1   T 529  \u001b[92m☑\u001b[0m 529 \n",
            "Q 606+166 T 772  \u001b[92m☑\u001b[0m 772 \n",
            "Q 558+90  T 648  \u001b[92m☑\u001b[0m 648 \n",
            "Q 15+957  T 972  \u001b[92m☑\u001b[0m 972 \n",
            "Q 727+81  T 808  \u001b[92m☑\u001b[0m 808 \n",
            "Q 325+78  T 403  \u001b[92m☑\u001b[0m 403 \n",
            "Q 89+616  T 705  \u001b[92m☑\u001b[0m 705 \n",
            "Q 844+573 T 1417 \u001b[92m☑\u001b[0m 1417\n",
            "Q 11+439  T 450  \u001b[92m☑\u001b[0m 450 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 122\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 317us/sample - loss: 0.0017 - acc: 0.9998 - val_loss: 0.0071 - val_acc: 0.9979\n",
            "Q 594+210 T 804  \u001b[91m☒\u001b[0m 704 \n",
            "Q 4+667   T 671  \u001b[92m☑\u001b[0m 671 \n",
            "Q 77+50   T 127  \u001b[92m☑\u001b[0m 127 \n",
            "Q 605+390 T 995  \u001b[92m☑\u001b[0m 995 \n",
            "Q 421+421 T 842  \u001b[92m☑\u001b[0m 842 \n",
            "Q 595+94  T 689  \u001b[92m☑\u001b[0m 689 \n",
            "Q 6+518   T 524  \u001b[92m☑\u001b[0m 524 \n",
            "Q 138+658 T 796  \u001b[92m☑\u001b[0m 796 \n",
            "Q 9+612   T 621  \u001b[92m☑\u001b[0m 621 \n",
            "Q 544+7   T 551  \u001b[92m☑\u001b[0m 551 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 123\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 320us/sample - loss: 7.4414e-04 - acc: 1.0000 - val_loss: 0.0067 - val_acc: 0.9979\n",
            "Q 447+53  T 500  \u001b[92m☑\u001b[0m 500 \n",
            "Q 559+66  T 625  \u001b[92m☑\u001b[0m 625 \n",
            "Q 957+94  T 1051 \u001b[92m☑\u001b[0m 1051\n",
            "Q 43+996  T 1039 \u001b[92m☑\u001b[0m 1039\n",
            "Q 17+29   T 46   \u001b[92m☑\u001b[0m 46  \n",
            "Q 918+216 T 1134 \u001b[92m☑\u001b[0m 1134\n",
            "Q 868+57  T 925  \u001b[92m☑\u001b[0m 925 \n",
            "Q 937+633 T 1570 \u001b[92m☑\u001b[0m 1570\n",
            "Q 2+820   T 822  \u001b[92m☑\u001b[0m 822 \n",
            "Q 557+477 T 1034 \u001b[92m☑\u001b[0m 1034\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 124\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 330us/sample - loss: 5.7794e-04 - acc: 1.0000 - val_loss: 0.0061 - val_acc: 0.9981\n",
            "Q 99+80   T 179  \u001b[92m☑\u001b[0m 179 \n",
            "Q 9+386   T 395  \u001b[92m☑\u001b[0m 395 \n",
            "Q 704+80  T 784  \u001b[92m☑\u001b[0m 784 \n",
            "Q 34+901  T 935  \u001b[92m☑\u001b[0m 935 \n",
            "Q 164+792 T 956  \u001b[92m☑\u001b[0m 956 \n",
            "Q 713+18  T 731  \u001b[92m☑\u001b[0m 731 \n",
            "Q 1+327   T 328  \u001b[92m☑\u001b[0m 328 \n",
            "Q 97+894  T 991  \u001b[92m☑\u001b[0m 991 \n",
            "Q 511+22  T 533  \u001b[92m☑\u001b[0m 533 \n",
            "Q 2+696   T 698  \u001b[92m☑\u001b[0m 698 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 125\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 329us/sample - loss: 4.9742e-04 - acc: 1.0000 - val_loss: 0.0059 - val_acc: 0.9983\n",
            "Q 284+8   T 292  \u001b[92m☑\u001b[0m 292 \n",
            "Q 358+92  T 450  \u001b[92m☑\u001b[0m 450 \n",
            "Q 818+97  T 915  \u001b[92m☑\u001b[0m 915 \n",
            "Q 299+56  T 355  \u001b[92m☑\u001b[0m 355 \n",
            "Q 662+458 T 1120 \u001b[92m☑\u001b[0m 1120\n",
            "Q 99+594  T 693  \u001b[92m☑\u001b[0m 693 \n",
            "Q 46+17   T 63   \u001b[92m☑\u001b[0m 63  \n",
            "Q 721+43  T 764  \u001b[92m☑\u001b[0m 764 \n",
            "Q 262+26  T 288  \u001b[92m☑\u001b[0m 288 \n",
            "Q 184+757 T 941  \u001b[92m☑\u001b[0m 941 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 126\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 326us/sample - loss: 4.5223e-04 - acc: 1.0000 - val_loss: 0.0059 - val_acc: 0.9982\n",
            "Q 204+9   T 213  \u001b[92m☑\u001b[0m 213 \n",
            "Q 113+913 T 1026 \u001b[92m☑\u001b[0m 1026\n",
            "Q 430+69  T 499  \u001b[92m☑\u001b[0m 499 \n",
            "Q 45+791  T 836  \u001b[92m☑\u001b[0m 836 \n",
            "Q 295+40  T 335  \u001b[92m☑\u001b[0m 335 \n",
            "Q 212+83  T 295  \u001b[92m☑\u001b[0m 295 \n",
            "Q 851+42  T 893  \u001b[92m☑\u001b[0m 893 \n",
            "Q 85+275  T 360  \u001b[92m☑\u001b[0m 360 \n",
            "Q 43+672  T 715  \u001b[92m☑\u001b[0m 715 \n",
            "Q 761+124 T 885  \u001b[92m☑\u001b[0m 885 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 127\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 323us/sample - loss: 0.0041 - acc: 0.9988 - val_loss: 0.0872 - val_acc: 0.9739\n",
            "Q 28+616  T 644  \u001b[91m☒\u001b[0m 634 \n",
            "Q 811+78  T 889  \u001b[92m☑\u001b[0m 889 \n",
            "Q 838+400 T 1238 \u001b[92m☑\u001b[0m 1238\n",
            "Q 718+999 T 1717 \u001b[92m☑\u001b[0m 1717\n",
            "Q 43+15   T 58   \u001b[92m☑\u001b[0m 58  \n",
            "Q 131+30  T 161  \u001b[92m☑\u001b[0m 161 \n",
            "Q 871+66  T 937  \u001b[92m☑\u001b[0m 937 \n",
            "Q 437+10  T 447  \u001b[92m☑\u001b[0m 447 \n",
            "Q 984+161 T 1145 \u001b[92m☑\u001b[0m 1145\n",
            "Q 628+165 T 793  \u001b[92m☑\u001b[0m 793 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 128\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 330us/sample - loss: 0.0380 - acc: 0.9888 - val_loss: 0.0123 - val_acc: 0.9966\n",
            "Q 733+426 T 1159 \u001b[92m☑\u001b[0m 1159\n",
            "Q 310+77  T 387  \u001b[92m☑\u001b[0m 387 \n",
            "Q 12+495  T 507  \u001b[92m☑\u001b[0m 507 \n",
            "Q 286+758 T 1044 \u001b[92m☑\u001b[0m 1044\n",
            "Q 8+269   T 277  \u001b[92m☑\u001b[0m 277 \n",
            "Q 6+819   T 825  \u001b[92m☑\u001b[0m 825 \n",
            "Q 42+575  T 617  \u001b[92m☑\u001b[0m 617 \n",
            "Q 712+43  T 755  \u001b[92m☑\u001b[0m 755 \n",
            "Q 137+10  T 147  \u001b[92m☑\u001b[0m 147 \n",
            "Q 321+59  T 380  \u001b[92m☑\u001b[0m 380 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 129\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 330us/sample - loss: 0.0014 - acc: 0.9999 - val_loss: 0.0080 - val_acc: 0.9976\n",
            "Q 5+948   T 953  \u001b[92m☑\u001b[0m 953 \n",
            "Q 174+859 T 1033 \u001b[92m☑\u001b[0m 1033\n",
            "Q 804+4   T 808  \u001b[92m☑\u001b[0m 808 \n",
            "Q 237+5   T 242  \u001b[92m☑\u001b[0m 242 \n",
            "Q 12+671  T 683  \u001b[92m☑\u001b[0m 683 \n",
            "Q 690+72  T 762  \u001b[92m☑\u001b[0m 762 \n",
            "Q 58+954  T 1012 \u001b[92m☑\u001b[0m 1012\n",
            "Q 789+655 T 1444 \u001b[92m☑\u001b[0m 1444\n",
            "Q 966+372 T 1338 \u001b[92m☑\u001b[0m 1338\n",
            "Q 346+842 T 1188 \u001b[92m☑\u001b[0m 1188\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 130\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 321us/sample - loss: 9.2756e-04 - acc: 1.0000 - val_loss: 0.0065 - val_acc: 0.9980\n",
            "Q 798+907 T 1705 \u001b[92m☑\u001b[0m 1705\n",
            "Q 374+67  T 441  \u001b[92m☑\u001b[0m 441 \n",
            "Q 135+60  T 195  \u001b[92m☑\u001b[0m 195 \n",
            "Q 737+304 T 1041 \u001b[92m☑\u001b[0m 1041\n",
            "Q 60+744  T 804  \u001b[92m☑\u001b[0m 804 \n",
            "Q 100+0   T 100  \u001b[92m☑\u001b[0m 100 \n",
            "Q 63+949  T 1012 \u001b[92m☑\u001b[0m 1012\n",
            "Q 22+265  T 287  \u001b[92m☑\u001b[0m 287 \n",
            "Q 56+743  T 799  \u001b[92m☑\u001b[0m 799 \n",
            "Q 44+78   T 122  \u001b[92m☑\u001b[0m 122 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 131\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 329us/sample - loss: 6.1092e-04 - acc: 1.0000 - val_loss: 0.0065 - val_acc: 0.9980\n",
            "Q 93+2    T 95   \u001b[92m☑\u001b[0m 95  \n",
            "Q 861+133 T 994  \u001b[92m☑\u001b[0m 994 \n",
            "Q 704+30  T 734  \u001b[92m☑\u001b[0m 734 \n",
            "Q 974+702 T 1676 \u001b[92m☑\u001b[0m 1676\n",
            "Q 443+297 T 740  \u001b[92m☑\u001b[0m 740 \n",
            "Q 179+27  T 206  \u001b[92m☑\u001b[0m 206 \n",
            "Q 885+44  T 929  \u001b[92m☑\u001b[0m 929 \n",
            "Q 320+517 T 837  \u001b[92m☑\u001b[0m 837 \n",
            "Q 53+828  T 881  \u001b[92m☑\u001b[0m 881 \n",
            "Q 91+371  T 462  \u001b[92m☑\u001b[0m 462 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 132\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 329us/sample - loss: 5.3763e-04 - acc: 1.0000 - val_loss: 0.0059 - val_acc: 0.9984\n",
            "Q 30+490  T 520  \u001b[92m☑\u001b[0m 520 \n",
            "Q 795+296 T 1091 \u001b[92m☑\u001b[0m 1091\n",
            "Q 210+7   T 217  \u001b[92m☑\u001b[0m 217 \n",
            "Q 36+800  T 836  \u001b[92m☑\u001b[0m 836 \n",
            "Q 98+983  T 1081 \u001b[92m☑\u001b[0m 1081\n",
            "Q 60+418  T 478  \u001b[92m☑\u001b[0m 478 \n",
            "Q 23+207  T 230  \u001b[92m☑\u001b[0m 230 \n",
            "Q 523+88  T 611  \u001b[92m☑\u001b[0m 611 \n",
            "Q 695+20  T 715  \u001b[92m☑\u001b[0m 715 \n",
            "Q 9+381   T 390  \u001b[92m☑\u001b[0m 390 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 133\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 320us/sample - loss: 4.6945e-04 - acc: 1.0000 - val_loss: 0.0060 - val_acc: 0.9983\n",
            "Q 961+9   T 970  \u001b[92m☑\u001b[0m 970 \n",
            "Q 569+27  T 596  \u001b[92m☑\u001b[0m 596 \n",
            "Q 890+743 T 1633 \u001b[92m☑\u001b[0m 1633\n",
            "Q 764+5   T 769  \u001b[92m☑\u001b[0m 769 \n",
            "Q 855+45  T 900  \u001b[92m☑\u001b[0m 900 \n",
            "Q 918+216 T 1134 \u001b[92m☑\u001b[0m 1134\n",
            "Q 372+91  T 463  \u001b[92m☑\u001b[0m 463 \n",
            "Q 163+0   T 163  \u001b[92m☑\u001b[0m 163 \n",
            "Q 83+894  T 977  \u001b[92m☑\u001b[0m 977 \n",
            "Q 720+74  T 794  \u001b[92m☑\u001b[0m 794 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 134\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 326us/sample - loss: 4.3020e-04 - acc: 1.0000 - val_loss: 0.0057 - val_acc: 0.9981\n",
            "Q 891+18  T 909  \u001b[92m☑\u001b[0m 909 \n",
            "Q 307+6   T 313  \u001b[92m☑\u001b[0m 313 \n",
            "Q 1+793   T 794  \u001b[92m☑\u001b[0m 794 \n",
            "Q 862+6   T 868  \u001b[92m☑\u001b[0m 868 \n",
            "Q 567+44  T 611  \u001b[92m☑\u001b[0m 611 \n",
            "Q 396+449 T 845  \u001b[92m☑\u001b[0m 845 \n",
            "Q 25+497  T 522  \u001b[92m☑\u001b[0m 522 \n",
            "Q 818+97  T 915  \u001b[92m☑\u001b[0m 915 \n",
            "Q 672+48  T 720  \u001b[92m☑\u001b[0m 720 \n",
            "Q 681+3   T 684  \u001b[92m☑\u001b[0m 684 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 135\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 326us/sample - loss: 3.7880e-04 - acc: 1.0000 - val_loss: 0.0057 - val_acc: 0.9983\n",
            "Q 578+78  T 656  \u001b[92m☑\u001b[0m 656 \n",
            "Q 41+814  T 855  \u001b[92m☑\u001b[0m 855 \n",
            "Q 548+270 T 818  \u001b[92m☑\u001b[0m 818 \n",
            "Q 789+6   T 795  \u001b[92m☑\u001b[0m 795 \n",
            "Q 657+517 T 1174 \u001b[92m☑\u001b[0m 1174\n",
            "Q 23+451  T 474  \u001b[92m☑\u001b[0m 474 \n",
            "Q 295+40  T 335  \u001b[92m☑\u001b[0m 335 \n",
            "Q 476+44  T 520  \u001b[92m☑\u001b[0m 520 \n",
            "Q 975+27  T 1002 \u001b[92m☑\u001b[0m 1002\n",
            "Q 99+452  T 551  \u001b[92m☑\u001b[0m 551 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 136\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 3.7914e-04 - acc: 1.0000 - val_loss: 0.0053 - val_acc: 0.9985\n",
            "Q 437+10  T 447  \u001b[92m☑\u001b[0m 447 \n",
            "Q 70+387  T 457  \u001b[92m☑\u001b[0m 457 \n",
            "Q 184+940 T 1124 \u001b[92m☑\u001b[0m 1124\n",
            "Q 6+808   T 814  \u001b[92m☑\u001b[0m 814 \n",
            "Q 29+73   T 102  \u001b[92m☑\u001b[0m 102 \n",
            "Q 570+200 T 770  \u001b[92m☑\u001b[0m 770 \n",
            "Q 60+399  T 459  \u001b[92m☑\u001b[0m 459 \n",
            "Q 182+57  T 239  \u001b[92m☑\u001b[0m 239 \n",
            "Q 931+72  T 1003 \u001b[92m☑\u001b[0m 1003\n",
            "Q 303+31  T 334  \u001b[92m☑\u001b[0m 334 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 137\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 329us/sample - loss: 0.0351 - acc: 0.9891 - val_loss: 0.0145 - val_acc: 0.9951\n",
            "Q 639+87  T 726  \u001b[92m☑\u001b[0m 726 \n",
            "Q 2+438   T 440  \u001b[92m☑\u001b[0m 440 \n",
            "Q 835+43  T 878  \u001b[92m☑\u001b[0m 878 \n",
            "Q 113+6   T 119  \u001b[92m☑\u001b[0m 119 \n",
            "Q 31+214  T 245  \u001b[92m☑\u001b[0m 245 \n",
            "Q 46+273  T 319  \u001b[92m☑\u001b[0m 319 \n",
            "Q 17+271  T 288  \u001b[92m☑\u001b[0m 288 \n",
            "Q 33+91   T 124  \u001b[92m☑\u001b[0m 124 \n",
            "Q 194+53  T 247  \u001b[92m☑\u001b[0m 247 \n",
            "Q 72+650  T 722  \u001b[92m☑\u001b[0m 722 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 138\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 327us/sample - loss: 0.0020 - acc: 0.9997 - val_loss: 0.0064 - val_acc: 0.9981\n",
            "Q 372+32  T 404  \u001b[92m☑\u001b[0m 404 \n",
            "Q 42+24   T 66   \u001b[92m☑\u001b[0m 66  \n",
            "Q 99+80   T 179  \u001b[92m☑\u001b[0m 179 \n",
            "Q 14+329  T 343  \u001b[92m☑\u001b[0m 343 \n",
            "Q 91+601  T 692  \u001b[92m☑\u001b[0m 692 \n",
            "Q 954+11  T 965  \u001b[92m☑\u001b[0m 965 \n",
            "Q 639+80  T 719  \u001b[92m☑\u001b[0m 719 \n",
            "Q 81+674  T 755  \u001b[92m☑\u001b[0m 755 \n",
            "Q 34+462  T 496  \u001b[92m☑\u001b[0m 496 \n",
            "Q 134+900 T 1034 \u001b[92m☑\u001b[0m 1034\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 139\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 330us/sample - loss: 7.1409e-04 - acc: 1.0000 - val_loss: 0.0059 - val_acc: 0.9980\n",
            "Q 60+426  T 486  \u001b[92m☑\u001b[0m 486 \n",
            "Q 316+1   T 317  \u001b[92m☑\u001b[0m 317 \n",
            "Q 271+321 T 592  \u001b[92m☑\u001b[0m 592 \n",
            "Q 19+492  T 511  \u001b[92m☑\u001b[0m 511 \n",
            "Q 93+918  T 1011 \u001b[92m☑\u001b[0m 1011\n",
            "Q 238+892 T 1130 \u001b[92m☑\u001b[0m 1130\n",
            "Q 83+420  T 503  \u001b[92m☑\u001b[0m 503 \n",
            "Q 67+615  T 682  \u001b[92m☑\u001b[0m 682 \n",
            "Q 2+147   T 149  \u001b[92m☑\u001b[0m 149 \n",
            "Q 885+44  T 929  \u001b[92m☑\u001b[0m 929 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 140\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 328us/sample - loss: 5.4882e-04 - acc: 1.0000 - val_loss: 0.0057 - val_acc: 0.9984\n",
            "Q 91+316  T 407  \u001b[92m☑\u001b[0m 407 \n",
            "Q 9+331   T 340  \u001b[92m☑\u001b[0m 340 \n",
            "Q 589+564 T 1153 \u001b[92m☑\u001b[0m 1153\n",
            "Q 406+5   T 411  \u001b[92m☑\u001b[0m 411 \n",
            "Q 465+856 T 1321 \u001b[92m☑\u001b[0m 1321\n",
            "Q 14+79   T 93   \u001b[92m☑\u001b[0m 93  \n",
            "Q 933+865 T 1798 \u001b[92m☑\u001b[0m 1798\n",
            "Q 36+680  T 716  \u001b[92m☑\u001b[0m 716 \n",
            "Q 310+200 T 510  \u001b[92m☑\u001b[0m 510 \n",
            "Q 55+252  T 307  \u001b[92m☑\u001b[0m 307 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 141\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 321us/sample - loss: 4.7013e-04 - acc: 1.0000 - val_loss: 0.0053 - val_acc: 0.9984\n",
            "Q 21+56   T 77   \u001b[92m☑\u001b[0m 77  \n",
            "Q 870+204 T 1074 \u001b[92m☑\u001b[0m 1074\n",
            "Q 41+940  T 981  \u001b[92m☑\u001b[0m 981 \n",
            "Q 374+41  T 415  \u001b[92m☑\u001b[0m 415 \n",
            "Q 88+398  T 486  \u001b[92m☑\u001b[0m 486 \n",
            "Q 217+19  T 236  \u001b[92m☑\u001b[0m 236 \n",
            "Q 1+536   T 537  \u001b[92m☑\u001b[0m 537 \n",
            "Q 91+11   T 102  \u001b[92m☑\u001b[0m 102 \n",
            "Q 360+65  T 425  \u001b[92m☑\u001b[0m 425 \n",
            "Q 17+209  T 226  \u001b[92m☑\u001b[0m 226 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 142\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 318us/sample - loss: 4.1485e-04 - acc: 1.0000 - val_loss: 0.0050 - val_acc: 0.9985\n",
            "Q 714+91  T 805  \u001b[92m☑\u001b[0m 805 \n",
            "Q 296+27  T 323  \u001b[92m☑\u001b[0m 323 \n",
            "Q 734+321 T 1055 \u001b[92m☑\u001b[0m 1055\n",
            "Q 317+136 T 453  \u001b[92m☑\u001b[0m 453 \n",
            "Q 642+4   T 646  \u001b[92m☑\u001b[0m 646 \n",
            "Q 30+46   T 76   \u001b[92m☑\u001b[0m 76  \n",
            "Q 24+45   T 69   \u001b[92m☑\u001b[0m 69  \n",
            "Q 248+5   T 253  \u001b[92m☑\u001b[0m 253 \n",
            "Q 540+1   T 541  \u001b[92m☑\u001b[0m 541 \n",
            "Q 18+269  T 287  \u001b[92m☑\u001b[0m 287 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 143\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 302us/sample - loss: 3.8110e-04 - acc: 1.0000 - val_loss: 0.0060 - val_acc: 0.9983\n",
            "Q 187+96  T 283  \u001b[92m☑\u001b[0m 283 \n",
            "Q 366+99  T 465  \u001b[92m☑\u001b[0m 465 \n",
            "Q 735+11  T 746  \u001b[92m☑\u001b[0m 746 \n",
            "Q 82+795  T 877  \u001b[92m☑\u001b[0m 877 \n",
            "Q 238+858 T 1096 \u001b[92m☑\u001b[0m 1096\n",
            "Q 41+200  T 241  \u001b[92m☑\u001b[0m 241 \n",
            "Q 832+72  T 904  \u001b[92m☑\u001b[0m 904 \n",
            "Q 684+976 T 1660 \u001b[92m☑\u001b[0m 1660\n",
            "Q 67+215  T 282  \u001b[92m☑\u001b[0m 282 \n",
            "Q 718+999 T 1717 \u001b[92m☑\u001b[0m 1717\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 144\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 316us/sample - loss: 0.0229 - acc: 0.9926 - val_loss: 0.0275 - val_acc: 0.9910\n",
            "Q 359+320 T 679  \u001b[92m☑\u001b[0m 679 \n",
            "Q 394+587 T 981  \u001b[92m☑\u001b[0m 981 \n",
            "Q 1+206   T 207  \u001b[92m☑\u001b[0m 207 \n",
            "Q 984+0   T 984  \u001b[92m☑\u001b[0m 984 \n",
            "Q 355+367 T 722  \u001b[92m☑\u001b[0m 722 \n",
            "Q 6+74    T 80   \u001b[92m☑\u001b[0m 80  \n",
            "Q 62+129  T 191  \u001b[92m☑\u001b[0m 191 \n",
            "Q 9+317   T 326  \u001b[92m☑\u001b[0m 326 \n",
            "Q 276+6   T 282  \u001b[92m☑\u001b[0m 282 \n",
            "Q 265+74  T 339  \u001b[92m☑\u001b[0m 339 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 145\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 315us/sample - loss: 0.0203 - acc: 0.9944 - val_loss: 0.0107 - val_acc: 0.9966\n",
            "Q 559+901 T 1460 \u001b[92m☑\u001b[0m 1460\n",
            "Q 774+65  T 839  \u001b[92m☑\u001b[0m 839 \n",
            "Q 858+50  T 908  \u001b[92m☑\u001b[0m 908 \n",
            "Q 99+639  T 738  \u001b[92m☑\u001b[0m 738 \n",
            "Q 187+96  T 283  \u001b[92m☑\u001b[0m 283 \n",
            "Q 90+255  T 345  \u001b[92m☑\u001b[0m 345 \n",
            "Q 276+24  T 300  \u001b[91m☒\u001b[0m 200 \n",
            "Q 375+9   T 384  \u001b[92m☑\u001b[0m 384 \n",
            "Q 929+87  T 1016 \u001b[92m☑\u001b[0m 1016\n",
            "Q 481+40  T 521  \u001b[92m☑\u001b[0m 521 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 146\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 312us/sample - loss: 0.0014 - acc: 0.9999 - val_loss: 0.0070 - val_acc: 0.9977\n",
            "Q 8+535   T 543  \u001b[92m☑\u001b[0m 543 \n",
            "Q 545+36  T 581  \u001b[92m☑\u001b[0m 581 \n",
            "Q 10+29   T 39   \u001b[92m☑\u001b[0m 39  \n",
            "Q 9+111   T 120  \u001b[92m☑\u001b[0m 120 \n",
            "Q 90+80   T 170  \u001b[92m☑\u001b[0m 170 \n",
            "Q 372+91  T 463  \u001b[92m☑\u001b[0m 463 \n",
            "Q 990+6   T 996  \u001b[92m☑\u001b[0m 996 \n",
            "Q 815+70  T 885  \u001b[92m☑\u001b[0m 885 \n",
            "Q 965+715 T 1680 \u001b[92m☑\u001b[0m 1680\n",
            "Q 94+45   T 139  \u001b[92m☑\u001b[0m 139 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 147\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 316us/sample - loss: 6.8006e-04 - acc: 1.0000 - val_loss: 0.0057 - val_acc: 0.9984\n",
            "Q 100+47  T 147  \u001b[92m☑\u001b[0m 147 \n",
            "Q 796+25  T 821  \u001b[92m☑\u001b[0m 821 \n",
            "Q 391+2   T 393  \u001b[92m☑\u001b[0m 393 \n",
            "Q 560+4   T 564  \u001b[92m☑\u001b[0m 564 \n",
            "Q 750+74  T 824  \u001b[92m☑\u001b[0m 824 \n",
            "Q 352+8   T 360  \u001b[92m☑\u001b[0m 360 \n",
            "Q 300+30  T 330  \u001b[92m☑\u001b[0m 330 \n",
            "Q 38+92   T 130  \u001b[92m☑\u001b[0m 130 \n",
            "Q 510+385 T 895  \u001b[92m☑\u001b[0m 895 \n",
            "Q 42+290  T 332  \u001b[92m☑\u001b[0m 332 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 148\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 314us/sample - loss: 5.1164e-04 - acc: 1.0000 - val_loss: 0.0053 - val_acc: 0.9982\n",
            "Q 17+209  T 226  \u001b[92m☑\u001b[0m 226 \n",
            "Q 961+70  T 1031 \u001b[92m☑\u001b[0m 1031\n",
            "Q 144+57  T 201  \u001b[92m☑\u001b[0m 201 \n",
            "Q 11+698  T 709  \u001b[92m☑\u001b[0m 709 \n",
            "Q 27+402  T 429  \u001b[92m☑\u001b[0m 429 \n",
            "Q 249+66  T 315  \u001b[92m☑\u001b[0m 315 \n",
            "Q 8+202   T 210  \u001b[92m☑\u001b[0m 210 \n",
            "Q 675+47  T 722  \u001b[92m☑\u001b[0m 722 \n",
            "Q 271+6   T 277  \u001b[92m☑\u001b[0m 277 \n",
            "Q 41+504  T 545  \u001b[92m☑\u001b[0m 545 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 149\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 306us/sample - loss: 4.4117e-04 - acc: 1.0000 - val_loss: 0.0052 - val_acc: 0.9984\n",
            "Q 169+27  T 196  \u001b[92m☑\u001b[0m 196 \n",
            "Q 68+47   T 115  \u001b[92m☑\u001b[0m 115 \n",
            "Q 76+91   T 167  \u001b[92m☑\u001b[0m 167 \n",
            "Q 109+964 T 1073 \u001b[92m☑\u001b[0m 1073\n",
            "Q 595+17  T 612  \u001b[92m☑\u001b[0m 612 \n",
            "Q 90+101  T 191  \u001b[92m☑\u001b[0m 191 \n",
            "Q 322+68  T 390  \u001b[92m☑\u001b[0m 390 \n",
            "Q 294+89  T 383  \u001b[92m☑\u001b[0m 383 \n",
            "Q 28+502  T 530  \u001b[92m☑\u001b[0m 530 \n",
            "Q 159+60  T 219  \u001b[92m☑\u001b[0m 219 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 150\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 308us/sample - loss: 3.9044e-04 - acc: 1.0000 - val_loss: 0.0050 - val_acc: 0.9984\n",
            "Q 751+939 T 1690 \u001b[92m☑\u001b[0m 1690\n",
            "Q 51+858  T 909  \u001b[92m☑\u001b[0m 909 \n",
            "Q 989+201 T 1190 \u001b[92m☑\u001b[0m 1190\n",
            "Q 52+763  T 815  \u001b[92m☑\u001b[0m 815 \n",
            "Q 36+917  T 953  \u001b[92m☑\u001b[0m 953 \n",
            "Q 223+630 T 853  \u001b[92m☑\u001b[0m 853 \n",
            "Q 850+150 T 1000 \u001b[92m☑\u001b[0m 1000\n",
            "Q 328+34  T 362  \u001b[92m☑\u001b[0m 362 \n",
            "Q 379+1   T 380  \u001b[92m☑\u001b[0m 380 \n",
            "Q 421+28  T 449  \u001b[92m☑\u001b[0m 449 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 151\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 318us/sample - loss: 3.5244e-04 - acc: 1.0000 - val_loss: 0.0048 - val_acc: 0.9984\n",
            "Q 215+617 T 832  \u001b[92m☑\u001b[0m 832 \n",
            "Q 425+926 T 1351 \u001b[92m☑\u001b[0m 1351\n",
            "Q 3+950   T 953  \u001b[92m☑\u001b[0m 953 \n",
            "Q 789+38  T 827  \u001b[92m☑\u001b[0m 827 \n",
            "Q 131+18  T 149  \u001b[92m☑\u001b[0m 149 \n",
            "Q 932+66  T 998  \u001b[92m☑\u001b[0m 998 \n",
            "Q 706+24  T 730  \u001b[92m☑\u001b[0m 730 \n",
            "Q 42+120  T 162  \u001b[92m☑\u001b[0m 162 \n",
            "Q 244+68  T 312  \u001b[92m☑\u001b[0m 312 \n",
            "Q 222+92  T 314  \u001b[92m☑\u001b[0m 314 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 152\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 304us/sample - loss: 3.2395e-04 - acc: 1.0000 - val_loss: 0.0053 - val_acc: 0.9983\n",
            "Q 76+230  T 306  \u001b[92m☑\u001b[0m 306 \n",
            "Q 334+583 T 917  \u001b[92m☑\u001b[0m 917 \n",
            "Q 90+883  T 973  \u001b[92m☑\u001b[0m 973 \n",
            "Q 888+326 T 1214 \u001b[92m☑\u001b[0m 1214\n",
            "Q 270+19  T 289  \u001b[92m☑\u001b[0m 289 \n",
            "Q 630+954 T 1584 \u001b[92m☑\u001b[0m 1584\n",
            "Q 593+32  T 625  \u001b[92m☑\u001b[0m 625 \n",
            "Q 99+455  T 554  \u001b[92m☑\u001b[0m 554 \n",
            "Q 2+991   T 993  \u001b[92m☑\u001b[0m 993 \n",
            "Q 597+968 T 1565 \u001b[92m☑\u001b[0m 1565\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 153\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 308us/sample - loss: 3.3758e-04 - acc: 1.0000 - val_loss: 0.0055 - val_acc: 0.9980\n",
            "Q 1+4     T 5    \u001b[91m☒\u001b[0m 6   \n",
            "Q 94+145  T 239  \u001b[92m☑\u001b[0m 239 \n",
            "Q 31+932  T 963  \u001b[92m☑\u001b[0m 963 \n",
            "Q 97+607  T 704  \u001b[92m☑\u001b[0m 704 \n",
            "Q 66+86   T 152  \u001b[92m☑\u001b[0m 152 \n",
            "Q 769+38  T 807  \u001b[92m☑\u001b[0m 807 \n",
            "Q 563+380 T 943  \u001b[92m☑\u001b[0m 943 \n",
            "Q 970+93  T 1063 \u001b[92m☑\u001b[0m 1063\n",
            "Q 702+66  T 768  \u001b[92m☑\u001b[0m 768 \n",
            "Q 10+29   T 39   \u001b[92m☑\u001b[0m 39  \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 154\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 316us/sample - loss: 0.0316 - acc: 0.9904 - val_loss: 0.0194 - val_acc: 0.9935\n",
            "Q 306+300 T 606  \u001b[92m☑\u001b[0m 606 \n",
            "Q 399+392 T 791  \u001b[92m☑\u001b[0m 791 \n",
            "Q 281+24  T 305  \u001b[92m☑\u001b[0m 305 \n",
            "Q 880+859 T 1739 \u001b[92m☑\u001b[0m 1739\n",
            "Q 869+8   T 877  \u001b[92m☑\u001b[0m 877 \n",
            "Q 0+174   T 174  \u001b[92m☑\u001b[0m 174 \n",
            "Q 81+487  T 568  \u001b[92m☑\u001b[0m 568 \n",
            "Q 95+461  T 556  \u001b[92m☑\u001b[0m 556 \n",
            "Q 838+693 T 1531 \u001b[92m☑\u001b[0m 1531\n",
            "Q 606+40  T 646  \u001b[92m☑\u001b[0m 646 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 155\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 311us/sample - loss: 0.0037 - acc: 0.9992 - val_loss: 0.0084 - val_acc: 0.9976\n",
            "Q 6+921   T 927  \u001b[92m☑\u001b[0m 927 \n",
            "Q 469+18  T 487  \u001b[92m☑\u001b[0m 487 \n",
            "Q 73+876  T 949  \u001b[92m☑\u001b[0m 949 \n",
            "Q 223+922 T 1145 \u001b[92m☑\u001b[0m 1145\n",
            "Q 388+139 T 527  \u001b[92m☑\u001b[0m 527 \n",
            "Q 347+0   T 347  \u001b[92m☑\u001b[0m 347 \n",
            "Q 176+3   T 179  \u001b[92m☑\u001b[0m 179 \n",
            "Q 443+211 T 654  \u001b[92m☑\u001b[0m 654 \n",
            "Q 44+485  T 529  \u001b[92m☑\u001b[0m 529 \n",
            "Q 641+148 T 789  \u001b[92m☑\u001b[0m 789 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 156\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 304us/sample - loss: 8.0024e-04 - acc: 1.0000 - val_loss: 0.0070 - val_acc: 0.9976\n",
            "Q 440+6   T 446  \u001b[92m☑\u001b[0m 446 \n",
            "Q 334+0   T 334  \u001b[92m☑\u001b[0m 334 \n",
            "Q 229+8   T 237  \u001b[92m☑\u001b[0m 237 \n",
            "Q 123+11  T 134  \u001b[92m☑\u001b[0m 134 \n",
            "Q 85+108  T 193  \u001b[92m☑\u001b[0m 193 \n",
            "Q 836+7   T 843  \u001b[92m☑\u001b[0m 843 \n",
            "Q 24+91   T 115  \u001b[92m☑\u001b[0m 115 \n",
            "Q 665+630 T 1295 \u001b[92m☑\u001b[0m 1295\n",
            "Q 25+204  T 229  \u001b[92m☑\u001b[0m 229 \n",
            "Q 671+42  T 713  \u001b[92m☑\u001b[0m 713 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 157\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 310us/sample - loss: 5.2659e-04 - acc: 1.0000 - val_loss: 0.0057 - val_acc: 0.9981\n",
            "Q 4+751   T 755  \u001b[92m☑\u001b[0m 755 \n",
            "Q 6+943   T 949  \u001b[92m☑\u001b[0m 949 \n",
            "Q 714+82  T 796  \u001b[92m☑\u001b[0m 796 \n",
            "Q 42+442  T 484  \u001b[92m☑\u001b[0m 484 \n",
            "Q 1+61    T 62   \u001b[92m☑\u001b[0m 62  \n",
            "Q 475+231 T 706  \u001b[92m☑\u001b[0m 706 \n",
            "Q 51+84   T 135  \u001b[92m☑\u001b[0m 135 \n",
            "Q 26+775  T 801  \u001b[92m☑\u001b[0m 801 \n",
            "Q 97+183  T 280  \u001b[92m☑\u001b[0m 280 \n",
            "Q 15+609  T 624  \u001b[92m☑\u001b[0m 624 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 158\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 315us/sample - loss: 4.1487e-04 - acc: 1.0000 - val_loss: 0.0052 - val_acc: 0.9983\n",
            "Q 107+855 T 962  \u001b[92m☑\u001b[0m 962 \n",
            "Q 9+932   T 941  \u001b[92m☑\u001b[0m 941 \n",
            "Q 570+200 T 770  \u001b[92m☑\u001b[0m 770 \n",
            "Q 57+17   T 74   \u001b[92m☑\u001b[0m 74  \n",
            "Q 632+616 T 1248 \u001b[92m☑\u001b[0m 1248\n",
            "Q 40+172  T 212  \u001b[92m☑\u001b[0m 212 \n",
            "Q 50+161  T 211  \u001b[92m☑\u001b[0m 211 \n",
            "Q 1+795   T 796  \u001b[92m☑\u001b[0m 796 \n",
            "Q 949+6   T 955  \u001b[92m☑\u001b[0m 955 \n",
            "Q 223+495 T 718  \u001b[92m☑\u001b[0m 718 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 159\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 305us/sample - loss: 3.5895e-04 - acc: 1.0000 - val_loss: 0.0052 - val_acc: 0.9984\n",
            "Q 327+18  T 345  \u001b[92m☑\u001b[0m 345 \n",
            "Q 16+8    T 24   \u001b[92m☑\u001b[0m 24  \n",
            "Q 9+567   T 576  \u001b[92m☑\u001b[0m 576 \n",
            "Q 964+30  T 994  \u001b[92m☑\u001b[0m 994 \n",
            "Q 145+32  T 177  \u001b[92m☑\u001b[0m 177 \n",
            "Q 2+469   T 471  \u001b[92m☑\u001b[0m 471 \n",
            "Q 53+382  T 435  \u001b[92m☑\u001b[0m 435 \n",
            "Q 41+504  T 545  \u001b[92m☑\u001b[0m 545 \n",
            "Q 37+789  T 826  \u001b[92m☑\u001b[0m 826 \n",
            "Q 177+707 T 884  \u001b[92m☑\u001b[0m 884 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 160\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 307us/sample - loss: 3.1824e-04 - acc: 1.0000 - val_loss: 0.0050 - val_acc: 0.9983\n",
            "Q 508+31  T 539  \u001b[92m☑\u001b[0m 539 \n",
            "Q 480+42  T 522  \u001b[92m☑\u001b[0m 522 \n",
            "Q 90+787  T 877  \u001b[92m☑\u001b[0m 877 \n",
            "Q 6+935   T 941  \u001b[92m☑\u001b[0m 941 \n",
            "Q 953+15  T 968  \u001b[92m☑\u001b[0m 968 \n",
            "Q 257+44  T 301  \u001b[92m☑\u001b[0m 301 \n",
            "Q 367+81  T 448  \u001b[92m☑\u001b[0m 448 \n",
            "Q 529+410 T 939  \u001b[92m☑\u001b[0m 939 \n",
            "Q 67+945  T 1012 \u001b[92m☑\u001b[0m 1012\n",
            "Q 28+444  T 472  \u001b[92m☑\u001b[0m 472 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 161\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 318us/sample - loss: 2.8955e-04 - acc: 1.0000 - val_loss: 0.0047 - val_acc: 0.9985\n",
            "Q 97+744  T 841  \u001b[92m☑\u001b[0m 841 \n",
            "Q 274+941 T 1215 \u001b[92m☑\u001b[0m 1215\n",
            "Q 483+27  T 510  \u001b[92m☑\u001b[0m 510 \n",
            "Q 44+484  T 528  \u001b[92m☑\u001b[0m 528 \n",
            "Q 275+354 T 629  \u001b[92m☑\u001b[0m 629 \n",
            "Q 75+51   T 126  \u001b[92m☑\u001b[0m 126 \n",
            "Q 446+540 T 986  \u001b[92m☑\u001b[0m 986 \n",
            "Q 781+518 T 1299 \u001b[92m☑\u001b[0m 1299\n",
            "Q 372+17  T 389  \u001b[92m☑\u001b[0m 389 \n",
            "Q 882+761 T 1643 \u001b[92m☑\u001b[0m 1643\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 162\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 306us/sample - loss: 2.6040e-04 - acc: 1.0000 - val_loss: 0.0048 - val_acc: 0.9985\n",
            "Q 503+209 T 712  \u001b[92m☑\u001b[0m 712 \n",
            "Q 340+12  T 352  \u001b[92m☑\u001b[0m 352 \n",
            "Q 944+92  T 1036 \u001b[92m☑\u001b[0m 1036\n",
            "Q 348+270 T 618  \u001b[92m☑\u001b[0m 618 \n",
            "Q 256+93  T 349  \u001b[92m☑\u001b[0m 349 \n",
            "Q 877+3   T 880  \u001b[92m☑\u001b[0m 880 \n",
            "Q 48+420  T 468  \u001b[92m☑\u001b[0m 468 \n",
            "Q 82+545  T 627  \u001b[92m☑\u001b[0m 627 \n",
            "Q 8+818   T 826  \u001b[92m☑\u001b[0m 826 \n",
            "Q 9+933   T 942  \u001b[92m☑\u001b[0m 942 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 163\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 303us/sample - loss: 2.3497e-04 - acc: 1.0000 - val_loss: 0.0046 - val_acc: 0.9985\n",
            "Q 125+554 T 679  \u001b[92m☑\u001b[0m 679 \n",
            "Q 99+995  T 1094 \u001b[92m☑\u001b[0m 1094\n",
            "Q 38+625  T 663  \u001b[92m☑\u001b[0m 663 \n",
            "Q 89+344  T 433  \u001b[92m☑\u001b[0m 433 \n",
            "Q 307+6   T 313  \u001b[92m☑\u001b[0m 313 \n",
            "Q 168+55  T 223  \u001b[92m☑\u001b[0m 223 \n",
            "Q 277+439 T 716  \u001b[92m☑\u001b[0m 716 \n",
            "Q 372+59  T 431  \u001b[92m☑\u001b[0m 431 \n",
            "Q 757+66  T 823  \u001b[92m☑\u001b[0m 823 \n",
            "Q 30+548  T 578  \u001b[92m☑\u001b[0m 578 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 164\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 15s 323us/sample - loss: 0.0325 - acc: 0.9909 - val_loss: 0.0231 - val_acc: 0.9925\n",
            "Q 394+587 T 981  \u001b[92m☑\u001b[0m 981 \n",
            "Q 40+66   T 106  \u001b[92m☑\u001b[0m 106 \n",
            "Q 54+378  T 432  \u001b[92m☑\u001b[0m 432 \n",
            "Q 916+717 T 1633 \u001b[92m☑\u001b[0m 1633\n",
            "Q 371+61  T 432  \u001b[92m☑\u001b[0m 432 \n",
            "Q 35+852  T 887  \u001b[92m☑\u001b[0m 887 \n",
            "Q 307+25  T 332  \u001b[92m☑\u001b[0m 332 \n",
            "Q 149+10  T 159  \u001b[92m☑\u001b[0m 159 \n",
            "Q 18+414  T 432  \u001b[92m☑\u001b[0m 432 \n",
            "Q 251+86  T 337  \u001b[92m☑\u001b[0m 337 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 165\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 312us/sample - loss: 0.0042 - acc: 0.9990 - val_loss: 0.0088 - val_acc: 0.9975\n",
            "Q 550+1   T 551  \u001b[92m☑\u001b[0m 551 \n",
            "Q 206+31  T 237  \u001b[92m☑\u001b[0m 237 \n",
            "Q 288+17  T 305  \u001b[92m☑\u001b[0m 305 \n",
            "Q 253+374 T 627  \u001b[92m☑\u001b[0m 627 \n",
            "Q 424+4   T 428  \u001b[92m☑\u001b[0m 428 \n",
            "Q 504+628 T 1132 \u001b[92m☑\u001b[0m 1132\n",
            "Q 21+56   T 77   \u001b[92m☑\u001b[0m 77  \n",
            "Q 0+892   T 892  \u001b[92m☑\u001b[0m 892 \n",
            "Q 562+39  T 601  \u001b[92m☑\u001b[0m 601 \n",
            "Q 615+34  T 649  \u001b[92m☑\u001b[0m 649 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 166\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 303us/sample - loss: 8.7787e-04 - acc: 0.9999 - val_loss: 0.0062 - val_acc: 0.9980\n",
            "Q 518+2   T 520  \u001b[92m☑\u001b[0m 520 \n",
            "Q 339+109 T 448  \u001b[92m☑\u001b[0m 448 \n",
            "Q 604+92  T 696  \u001b[92m☑\u001b[0m 696 \n",
            "Q 2+505   T 507  \u001b[92m☑\u001b[0m 507 \n",
            "Q 170+55  T 225  \u001b[92m☑\u001b[0m 225 \n",
            "Q 130+85  T 215  \u001b[92m☑\u001b[0m 215 \n",
            "Q 740+91  T 831  \u001b[92m☑\u001b[0m 831 \n",
            "Q 767+20  T 787  \u001b[92m☑\u001b[0m 787 \n",
            "Q 800+90  T 890  \u001b[92m☑\u001b[0m 890 \n",
            "Q 6+711   T 717  \u001b[92m☑\u001b[0m 717 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 167\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 314us/sample - loss: 4.9987e-04 - acc: 1.0000 - val_loss: 0.0055 - val_acc: 0.9982\n",
            "Q 934+950 T 1884 \u001b[92m☑\u001b[0m 1884\n",
            "Q 163+0   T 163  \u001b[92m☑\u001b[0m 163 \n",
            "Q 332+882 T 1214 \u001b[92m☑\u001b[0m 1214\n",
            "Q 722+91  T 813  \u001b[92m☑\u001b[0m 813 \n",
            "Q 335+7   T 342  \u001b[92m☑\u001b[0m 342 \n",
            "Q 763+490 T 1253 \u001b[92m☑\u001b[0m 1253\n",
            "Q 795+296 T 1091 \u001b[92m☑\u001b[0m 1091\n",
            "Q 469+18  T 487  \u001b[92m☑\u001b[0m 487 \n",
            "Q 633+8   T 641  \u001b[92m☑\u001b[0m 641 \n",
            "Q 492+10  T 502  \u001b[92m☑\u001b[0m 502 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 168\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 317us/sample - loss: 3.9900e-04 - acc: 1.0000 - val_loss: 0.0054 - val_acc: 0.9981\n",
            "Q 58+501  T 559  \u001b[92m☑\u001b[0m 559 \n",
            "Q 29+844  T 873  \u001b[92m☑\u001b[0m 873 \n",
            "Q 104+5   T 109  \u001b[92m☑\u001b[0m 109 \n",
            "Q 33+91   T 124  \u001b[92m☑\u001b[0m 124 \n",
            "Q 88+236  T 324  \u001b[92m☑\u001b[0m 324 \n",
            "Q 307+2   T 309  \u001b[92m☑\u001b[0m 309 \n",
            "Q 104+62  T 166  \u001b[92m☑\u001b[0m 166 \n",
            "Q 674+732 T 1406 \u001b[92m☑\u001b[0m 1406\n",
            "Q 561+870 T 1431 \u001b[92m☑\u001b[0m 1431\n",
            "Q 368+66  T 434  \u001b[92m☑\u001b[0m 434 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 169\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 305us/sample - loss: 3.4378e-04 - acc: 1.0000 - val_loss: 0.0052 - val_acc: 0.9984\n",
            "Q 657+28  T 685  \u001b[92m☑\u001b[0m 685 \n",
            "Q 132+92  T 224  \u001b[92m☑\u001b[0m 224 \n",
            "Q 126+64  T 190  \u001b[92m☑\u001b[0m 190 \n",
            "Q 549+727 T 1276 \u001b[92m☑\u001b[0m 1276\n",
            "Q 75+931  T 1006 \u001b[92m☑\u001b[0m 1006\n",
            "Q 73+57   T 130  \u001b[92m☑\u001b[0m 130 \n",
            "Q 206+26  T 232  \u001b[92m☑\u001b[0m 232 \n",
            "Q 483+270 T 753  \u001b[92m☑\u001b[0m 753 \n",
            "Q 1+894   T 895  \u001b[92m☑\u001b[0m 895 \n",
            "Q 211+38  T 249  \u001b[92m☑\u001b[0m 249 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 170\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 310us/sample - loss: 3.0336e-04 - acc: 1.0000 - val_loss: 0.0050 - val_acc: 0.9984\n",
            "Q 52+449  T 501  \u001b[92m☑\u001b[0m 501 \n",
            "Q 28+42   T 70   \u001b[92m☑\u001b[0m 70  \n",
            "Q 562+189 T 751  \u001b[92m☑\u001b[0m 751 \n",
            "Q 4+787   T 791  \u001b[92m☑\u001b[0m 791 \n",
            "Q 20+909  T 929  \u001b[92m☑\u001b[0m 929 \n",
            "Q 99+647  T 746  \u001b[92m☑\u001b[0m 746 \n",
            "Q 185+86  T 271  \u001b[92m☑\u001b[0m 271 \n",
            "Q 12+965  T 977  \u001b[92m☑\u001b[0m 977 \n",
            "Q 354+5   T 359  \u001b[92m☑\u001b[0m 359 \n",
            "Q 318+10  T 328  \u001b[92m☑\u001b[0m 328 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 171\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 318us/sample - loss: 2.6862e-04 - acc: 1.0000 - val_loss: 0.0049 - val_acc: 0.9985\n",
            "Q 91+879  T 970  \u001b[92m☑\u001b[0m 970 \n",
            "Q 168+55  T 223  \u001b[92m☑\u001b[0m 223 \n",
            "Q 9+189   T 198  \u001b[92m☑\u001b[0m 198 \n",
            "Q 576+132 T 708  \u001b[92m☑\u001b[0m 708 \n",
            "Q 29+726  T 755  \u001b[92m☑\u001b[0m 755 \n",
            "Q 583+48  T 631  \u001b[92m☑\u001b[0m 631 \n",
            "Q 405+423 T 828  \u001b[92m☑\u001b[0m 828 \n",
            "Q 210+7   T 217  \u001b[92m☑\u001b[0m 217 \n",
            "Q 77+481  T 558  \u001b[92m☑\u001b[0m 558 \n",
            "Q 491+94  T 585  \u001b[92m☑\u001b[0m 585 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 172\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 305us/sample - loss: 2.4217e-04 - acc: 1.0000 - val_loss: 0.0047 - val_acc: 0.9985\n",
            "Q 793+2   T 795  \u001b[92m☑\u001b[0m 795 \n",
            "Q 974+36  T 1010 \u001b[92m☑\u001b[0m 1010\n",
            "Q 432+6   T 438  \u001b[92m☑\u001b[0m 438 \n",
            "Q 909+93  T 1002 \u001b[92m☑\u001b[0m 1002\n",
            "Q 147+603 T 750  \u001b[92m☑\u001b[0m 750 \n",
            "Q 410+92  T 502  \u001b[92m☑\u001b[0m 502 \n",
            "Q 26+242  T 268  \u001b[92m☑\u001b[0m 268 \n",
            "Q 541+651 T 1192 \u001b[92m☑\u001b[0m 1192\n",
            "Q 464+295 T 759  \u001b[92m☑\u001b[0m 759 \n",
            "Q 293+1   T 294  \u001b[92m☑\u001b[0m 294 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 173\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 305us/sample - loss: 2.6110e-04 - acc: 1.0000 - val_loss: 0.0058 - val_acc: 0.9983\n",
            "Q 618+75  T 693  \u001b[92m☑\u001b[0m 693 \n",
            "Q 325+11  T 336  \u001b[92m☑\u001b[0m 336 \n",
            "Q 90+883  T 973  \u001b[92m☑\u001b[0m 973 \n",
            "Q 713+48  T 761  \u001b[92m☑\u001b[0m 761 \n",
            "Q 699+99  T 798  \u001b[92m☑\u001b[0m 798 \n",
            "Q 810+593 T 1403 \u001b[92m☑\u001b[0m 1403\n",
            "Q 19+188  T 207  \u001b[92m☑\u001b[0m 207 \n",
            "Q 74+231  T 305  \u001b[92m☑\u001b[0m 305 \n",
            "Q 354+0   T 354  \u001b[92m☑\u001b[0m 354 \n",
            "Q 529+106 T 635  \u001b[92m☑\u001b[0m 635 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 174\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 318us/sample - loss: 0.0429 - acc: 0.9883 - val_loss: 0.0131 - val_acc: 0.9964\n",
            "Q 3+116   T 119  \u001b[92m☑\u001b[0m 119 \n",
            "Q 18+700  T 718  \u001b[92m☑\u001b[0m 718 \n",
            "Q 459+596 T 1055 \u001b[92m☑\u001b[0m 1055\n",
            "Q 75+712  T 787  \u001b[92m☑\u001b[0m 787 \n",
            "Q 3+23    T 26   \u001b[92m☑\u001b[0m 26  \n",
            "Q 282+5   T 287  \u001b[92m☑\u001b[0m 287 \n",
            "Q 29+844  T 873  \u001b[92m☑\u001b[0m 873 \n",
            "Q 932+66  T 998  \u001b[92m☑\u001b[0m 998 \n",
            "Q 35+8    T 43   \u001b[92m☑\u001b[0m 43  \n",
            "Q 54+378  T 432  \u001b[92m☑\u001b[0m 432 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 175\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 309us/sample - loss: 0.0012 - acc: 0.9999 - val_loss: 0.0075 - val_acc: 0.9980\n",
            "Q 586+18  T 604  \u001b[92m☑\u001b[0m 604 \n",
            "Q 63+515  T 578  \u001b[92m☑\u001b[0m 578 \n",
            "Q 515+89  T 604  \u001b[92m☑\u001b[0m 604 \n",
            "Q 557+477 T 1034 \u001b[92m☑\u001b[0m 1034\n",
            "Q 288+9   T 297  \u001b[92m☑\u001b[0m 297 \n",
            "Q 230+66  T 296  \u001b[92m☑\u001b[0m 296 \n",
            "Q 407+8   T 415  \u001b[92m☑\u001b[0m 415 \n",
            "Q 43+28   T 71   \u001b[92m☑\u001b[0m 71  \n",
            "Q 350+20  T 370  \u001b[92m☑\u001b[0m 370 \n",
            "Q 727+158 T 885  \u001b[92m☑\u001b[0m 885 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 176\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 309us/sample - loss: 5.8327e-04 - acc: 1.0000 - val_loss: 0.0064 - val_acc: 0.9980\n",
            "Q 866+898 T 1764 \u001b[92m☑\u001b[0m 1764\n",
            "Q 181+85  T 266  \u001b[92m☑\u001b[0m 266 \n",
            "Q 516+802 T 1318 \u001b[92m☑\u001b[0m 1318\n",
            "Q 83+448  T 531  \u001b[92m☑\u001b[0m 531 \n",
            "Q 42+496  T 538  \u001b[92m☑\u001b[0m 538 \n",
            "Q 9+514   T 523  \u001b[92m☑\u001b[0m 523 \n",
            "Q 21+753  T 774  \u001b[92m☑\u001b[0m 774 \n",
            "Q 75+998  T 1073 \u001b[92m☑\u001b[0m 1073\n",
            "Q 345+4   T 349  \u001b[92m☑\u001b[0m 349 \n",
            "Q 429+513 T 942  \u001b[92m☑\u001b[0m 942 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 177\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 313us/sample - loss: 4.4774e-04 - acc: 1.0000 - val_loss: 0.0060 - val_acc: 0.9983\n",
            "Q 19+455  T 474  \u001b[92m☑\u001b[0m 474 \n",
            "Q 644+653 T 1297 \u001b[92m☑\u001b[0m 1297\n",
            "Q 6+361   T 367  \u001b[92m☑\u001b[0m 367 \n",
            "Q 17+17   T 34   \u001b[92m☑\u001b[0m 34  \n",
            "Q 19+981  T 1000 \u001b[92m☑\u001b[0m 1000\n",
            "Q 19+277  T 296  \u001b[92m☑\u001b[0m 296 \n",
            "Q 76+736  T 812  \u001b[92m☑\u001b[0m 812 \n",
            "Q 63+945  T 1008 \u001b[92m☑\u001b[0m 1008\n",
            "Q 560+22  T 582  \u001b[92m☑\u001b[0m 582 \n",
            "Q 51+858  T 909  \u001b[92m☑\u001b[0m 909 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 178\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 315us/sample - loss: 3.7741e-04 - acc: 1.0000 - val_loss: 0.0057 - val_acc: 0.9983\n",
            "Q 618+285 T 903  \u001b[92m☑\u001b[0m 903 \n",
            "Q 628+99  T 727  \u001b[92m☑\u001b[0m 727 \n",
            "Q 727+81  T 808  \u001b[92m☑\u001b[0m 808 \n",
            "Q 97+183  T 280  \u001b[92m☑\u001b[0m 280 \n",
            "Q 58+858  T 916  \u001b[92m☑\u001b[0m 916 \n",
            "Q 3+950   T 953  \u001b[92m☑\u001b[0m 953 \n",
            "Q 35+12   T 47   \u001b[92m☑\u001b[0m 47  \n",
            "Q 128+70  T 198  \u001b[92m☑\u001b[0m 198 \n",
            "Q 2+125   T 127  \u001b[92m☑\u001b[0m 127 \n",
            "Q 47+185  T 232  \u001b[92m☑\u001b[0m 232 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 179\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 307us/sample - loss: 3.2800e-04 - acc: 1.0000 - val_loss: 0.0056 - val_acc: 0.9984\n",
            "Q 4+787   T 791  \u001b[92m☑\u001b[0m 791 \n",
            "Q 59+975  T 1034 \u001b[92m☑\u001b[0m 1034\n",
            "Q 134+35  T 169  \u001b[92m☑\u001b[0m 169 \n",
            "Q 83+451  T 534  \u001b[92m☑\u001b[0m 534 \n",
            "Q 576+31  T 607  \u001b[92m☑\u001b[0m 607 \n",
            "Q 291+80  T 371  \u001b[92m☑\u001b[0m 371 \n",
            "Q 49+46   T 95   \u001b[92m☑\u001b[0m 95  \n",
            "Q 74+54   T 128  \u001b[92m☑\u001b[0m 128 \n",
            "Q 8+791   T 799  \u001b[92m☑\u001b[0m 799 \n",
            "Q 870+204 T 1074 \u001b[92m☑\u001b[0m 1074\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 180\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 311us/sample - loss: 2.8838e-04 - acc: 1.0000 - val_loss: 0.0053 - val_acc: 0.9983\n",
            "Q 53+746  T 799  \u001b[92m☑\u001b[0m 799 \n",
            "Q 4+794   T 798  \u001b[92m☑\u001b[0m 798 \n",
            "Q 251+151 T 402  \u001b[92m☑\u001b[0m 402 \n",
            "Q 98+173  T 271  \u001b[92m☑\u001b[0m 271 \n",
            "Q 549+4   T 553  \u001b[92m☑\u001b[0m 553 \n",
            "Q 86+93   T 179  \u001b[92m☑\u001b[0m 179 \n",
            "Q 63+716  T 779  \u001b[92m☑\u001b[0m 779 \n",
            "Q 214+56  T 270  \u001b[92m☑\u001b[0m 270 \n",
            "Q 646+5   T 651  \u001b[92m☑\u001b[0m 651 \n",
            "Q 825+183 T 1008 \u001b[92m☑\u001b[0m 1008\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 181\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 322us/sample - loss: 2.5692e-04 - acc: 1.0000 - val_loss: 0.0051 - val_acc: 0.9983\n",
            "Q 507+56  T 563  \u001b[92m☑\u001b[0m 563 \n",
            "Q 213+292 T 505  \u001b[92m☑\u001b[0m 505 \n",
            "Q 55+814  T 869  \u001b[92m☑\u001b[0m 869 \n",
            "Q 67+551  T 618  \u001b[92m☑\u001b[0m 618 \n",
            "Q 740+100 T 840  \u001b[92m☑\u001b[0m 840 \n",
            "Q 24+66   T 90   \u001b[92m☑\u001b[0m 90  \n",
            "Q 14+433  T 447  \u001b[92m☑\u001b[0m 447 \n",
            "Q 628+932 T 1560 \u001b[92m☑\u001b[0m 1560\n",
            "Q 638+824 T 1462 \u001b[92m☑\u001b[0m 1462\n",
            "Q 332+549 T 881  \u001b[92m☑\u001b[0m 881 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 182\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 305us/sample - loss: 2.2975e-04 - acc: 1.0000 - val_loss: 0.0050 - val_acc: 0.9983\n",
            "Q 653+98  T 751  \u001b[92m☑\u001b[0m 751 \n",
            "Q 399+47  T 446  \u001b[92m☑\u001b[0m 446 \n",
            "Q 56+602  T 658  \u001b[92m☑\u001b[0m 658 \n",
            "Q 9+651   T 660  \u001b[92m☑\u001b[0m 660 \n",
            "Q 4+210   T 214  \u001b[92m☑\u001b[0m 214 \n",
            "Q 170+55  T 225  \u001b[92m☑\u001b[0m 225 \n",
            "Q 43+92   T 135  \u001b[92m☑\u001b[0m 135 \n",
            "Q 334+76  T 410  \u001b[92m☑\u001b[0m 410 \n",
            "Q 40+11   T 51   \u001b[92m☑\u001b[0m 51  \n",
            "Q 661+78  T 739  \u001b[92m☑\u001b[0m 739 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 183\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 307us/sample - loss: 2.0750e-04 - acc: 1.0000 - val_loss: 0.0048 - val_acc: 0.9982\n",
            "Q 223+333 T 556  \u001b[92m☑\u001b[0m 556 \n",
            "Q 8+561   T 569  \u001b[92m☑\u001b[0m 569 \n",
            "Q 28+52   T 80   \u001b[92m☑\u001b[0m 80  \n",
            "Q 33+912  T 945  \u001b[92m☑\u001b[0m 945 \n",
            "Q 41+322  T 363  \u001b[92m☑\u001b[0m 363 \n",
            "Q 556+58  T 614  \u001b[92m☑\u001b[0m 614 \n",
            "Q 2+147   T 149  \u001b[92m☑\u001b[0m 149 \n",
            "Q 62+683  T 745  \u001b[92m☑\u001b[0m 745 \n",
            "Q 68+800  T 868  \u001b[92m☑\u001b[0m 868 \n",
            "Q 142+548 T 690  \u001b[92m☑\u001b[0m 690 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 184\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 296us/sample - loss: 1.8863e-04 - acc: 1.0000 - val_loss: 0.0047 - val_acc: 0.9983\n",
            "Q 24+91   T 115  \u001b[92m☑\u001b[0m 115 \n",
            "Q 35+879  T 914  \u001b[92m☑\u001b[0m 914 \n",
            "Q 786+10  T 796  \u001b[92m☑\u001b[0m 796 \n",
            "Q 73+179  T 252  \u001b[92m☑\u001b[0m 252 \n",
            "Q 223+495 T 718  \u001b[92m☑\u001b[0m 718 \n",
            "Q 464+295 T 759  \u001b[92m☑\u001b[0m 759 \n",
            "Q 235+309 T 544  \u001b[92m☑\u001b[0m 544 \n",
            "Q 61+416  T 477  \u001b[92m☑\u001b[0m 477 \n",
            "Q 815+6   T 821  \u001b[92m☑\u001b[0m 821 \n",
            "Q 91+930  T 1021 \u001b[92m☑\u001b[0m 1021\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 185\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 296us/sample - loss: 1.7097e-04 - acc: 1.0000 - val_loss: 0.0046 - val_acc: 0.9984\n",
            "Q 217+909 T 1126 \u001b[92m☑\u001b[0m 1126\n",
            "Q 181+14  T 195  \u001b[92m☑\u001b[0m 195 \n",
            "Q 369+208 T 577  \u001b[92m☑\u001b[0m 577 \n",
            "Q 121+34  T 155  \u001b[92m☑\u001b[0m 155 \n",
            "Q 752+78  T 830  \u001b[92m☑\u001b[0m 830 \n",
            "Q 389+7   T 396  \u001b[92m☑\u001b[0m 396 \n",
            "Q 3+943   T 946  \u001b[92m☑\u001b[0m 946 \n",
            "Q 199+54  T 253  \u001b[92m☑\u001b[0m 253 \n",
            "Q 55+866  T 921  \u001b[92m☑\u001b[0m 921 \n",
            "Q 993+27  T 1020 \u001b[92m☑\u001b[0m 1020\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 186\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 299us/sample - loss: 1.5877e-04 - acc: 1.0000 - val_loss: 0.0045 - val_acc: 0.9984\n",
            "Q 0+833   T 833  \u001b[92m☑\u001b[0m 833 \n",
            "Q 21+753  T 774  \u001b[92m☑\u001b[0m 774 \n",
            "Q 967+77  T 1044 \u001b[92m☑\u001b[0m 1044\n",
            "Q 726+85  T 811  \u001b[92m☑\u001b[0m 811 \n",
            "Q 731+814 T 1545 \u001b[92m☑\u001b[0m 1545\n",
            "Q 77+436  T 513  \u001b[92m☑\u001b[0m 513 \n",
            "Q 404+228 T 632  \u001b[92m☑\u001b[0m 632 \n",
            "Q 523+3   T 526  \u001b[92m☑\u001b[0m 526 \n",
            "Q 95+418  T 513  \u001b[92m☑\u001b[0m 513 \n",
            "Q 805+298 T 1103 \u001b[92m☑\u001b[0m 1103\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 187\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 308us/sample - loss: 0.0387 - acc: 0.9888 - val_loss: 0.0130 - val_acc: 0.9958\n",
            "Q 172+13  T 185  \u001b[92m☑\u001b[0m 185 \n",
            "Q 21+513  T 534  \u001b[92m☑\u001b[0m 534 \n",
            "Q 12+368  T 380  \u001b[92m☑\u001b[0m 380 \n",
            "Q 23+333  T 356  \u001b[92m☑\u001b[0m 356 \n",
            "Q 422+859 T 1281 \u001b[92m☑\u001b[0m 1281\n",
            "Q 601+306 T 907  \u001b[92m☑\u001b[0m 907 \n",
            "Q 36+899  T 935  \u001b[92m☑\u001b[0m 935 \n",
            "Q 7+14    T 21   \u001b[92m☑\u001b[0m 21  \n",
            "Q 597+968 T 1565 \u001b[92m☑\u001b[0m 1565\n",
            "Q 706+24  T 730  \u001b[92m☑\u001b[0m 730 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 188\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 307us/sample - loss: 0.0032 - acc: 0.9993 - val_loss: 0.0072 - val_acc: 0.9981\n",
            "Q 551+823 T 1374 \u001b[92m☑\u001b[0m 1374\n",
            "Q 30+863  T 893  \u001b[92m☑\u001b[0m 893 \n",
            "Q 909+93  T 1002 \u001b[92m☑\u001b[0m 1002\n",
            "Q 961+9   T 970  \u001b[92m☑\u001b[0m 970 \n",
            "Q 93+804  T 897  \u001b[92m☑\u001b[0m 897 \n",
            "Q 909+80  T 989  \u001b[92m☑\u001b[0m 989 \n",
            "Q 24+335  T 359  \u001b[92m☑\u001b[0m 359 \n",
            "Q 974+702 T 1676 \u001b[92m☑\u001b[0m 1676\n",
            "Q 234+4   T 238  \u001b[92m☑\u001b[0m 238 \n",
            "Q 983+72  T 1055 \u001b[92m☑\u001b[0m 1055\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 189\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 299us/sample - loss: 7.0390e-04 - acc: 1.0000 - val_loss: 0.0048 - val_acc: 0.9986\n",
            "Q 37+48   T 85   \u001b[92m☑\u001b[0m 85  \n",
            "Q 578+724 T 1302 \u001b[92m☑\u001b[0m 1302\n",
            "Q 668+515 T 1183 \u001b[92m☑\u001b[0m 1183\n",
            "Q 800+43  T 843  \u001b[92m☑\u001b[0m 843 \n",
            "Q 129+49  T 178  \u001b[92m☑\u001b[0m 178 \n",
            "Q 74+678  T 752  \u001b[92m☑\u001b[0m 752 \n",
            "Q 199+54  T 253  \u001b[92m☑\u001b[0m 253 \n",
            "Q 1+795   T 796  \u001b[92m☑\u001b[0m 796 \n",
            "Q 878+4   T 882  \u001b[92m☑\u001b[0m 882 \n",
            "Q 95+221  T 316  \u001b[92m☑\u001b[0m 316 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 190\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 307us/sample - loss: 4.2407e-04 - acc: 1.0000 - val_loss: 0.0045 - val_acc: 0.9985\n",
            "Q 270+367 T 637  \u001b[92m☑\u001b[0m 637 \n",
            "Q 27+244  T 271  \u001b[92m☑\u001b[0m 271 \n",
            "Q 274+167 T 441  \u001b[92m☑\u001b[0m 441 \n",
            "Q 10+164  T 174  \u001b[92m☑\u001b[0m 174 \n",
            "Q 731+13  T 744  \u001b[92m☑\u001b[0m 744 \n",
            "Q 85+54   T 139  \u001b[92m☑\u001b[0m 139 \n",
            "Q 794+93  T 887  \u001b[92m☑\u001b[0m 887 \n",
            "Q 366+99  T 465  \u001b[92m☑\u001b[0m 465 \n",
            "Q 76+530  T 606  \u001b[92m☑\u001b[0m 606 \n",
            "Q 38+58   T 96   \u001b[92m☑\u001b[0m 96  \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 191\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 306us/sample - loss: 3.3276e-04 - acc: 1.0000 - val_loss: 0.0043 - val_acc: 0.9988\n",
            "Q 47+113  T 160  \u001b[92m☑\u001b[0m 160 \n",
            "Q 338+7   T 345  \u001b[92m☑\u001b[0m 345 \n",
            "Q 138+658 T 796  \u001b[92m☑\u001b[0m 796 \n",
            "Q 526+912 T 1438 \u001b[92m☑\u001b[0m 1438\n",
            "Q 585+652 T 1237 \u001b[92m☑\u001b[0m 1237\n",
            "Q 195+29  T 224  \u001b[92m☑\u001b[0m 224 \n",
            "Q 7+499   T 506  \u001b[92m☑\u001b[0m 506 \n",
            "Q 168+60  T 228  \u001b[92m☑\u001b[0m 228 \n",
            "Q 688+21  T 709  \u001b[92m☑\u001b[0m 709 \n",
            "Q 565+988 T 1553 \u001b[92m☑\u001b[0m 1553\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 192\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 292us/sample - loss: 2.8670e-04 - acc: 1.0000 - val_loss: 0.0042 - val_acc: 0.9987\n",
            "Q 313+815 T 1128 \u001b[92m☑\u001b[0m 1128\n",
            "Q 206+277 T 483  \u001b[92m☑\u001b[0m 483 \n",
            "Q 569+1   T 570  \u001b[92m☑\u001b[0m 570 \n",
            "Q 196+240 T 436  \u001b[92m☑\u001b[0m 436 \n",
            "Q 365+54  T 419  \u001b[92m☑\u001b[0m 419 \n",
            "Q 16+485  T 501  \u001b[92m☑\u001b[0m 501 \n",
            "Q 861+33  T 894  \u001b[92m☑\u001b[0m 894 \n",
            "Q 702+38  T 740  \u001b[92m☑\u001b[0m 740 \n",
            "Q 57+308  T 365  \u001b[92m☑\u001b[0m 365 \n",
            "Q 677+7   T 684  \u001b[92m☑\u001b[0m 684 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 193\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 305us/sample - loss: 2.5255e-04 - acc: 1.0000 - val_loss: 0.0041 - val_acc: 0.9988\n",
            "Q 404+901 T 1305 \u001b[92m☑\u001b[0m 1305\n",
            "Q 8+714   T 722  \u001b[92m☑\u001b[0m 722 \n",
            "Q 396+449 T 845  \u001b[92m☑\u001b[0m 845 \n",
            "Q 253+92  T 345  \u001b[92m☑\u001b[0m 345 \n",
            "Q 421+37  T 458  \u001b[92m☑\u001b[0m 458 \n",
            "Q 583+65  T 648  \u001b[92m☑\u001b[0m 648 \n",
            "Q 472+83  T 555  \u001b[92m☑\u001b[0m 555 \n",
            "Q 40+809  T 849  \u001b[92m☑\u001b[0m 849 \n",
            "Q 23+274  T 297  \u001b[92m☑\u001b[0m 297 \n",
            "Q 72+801  T 873  \u001b[92m☑\u001b[0m 873 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 194\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 297us/sample - loss: 2.2470e-04 - acc: 1.0000 - val_loss: 0.0041 - val_acc: 0.9987\n",
            "Q 654+381 T 1035 \u001b[92m☑\u001b[0m 1035\n",
            "Q 92+780  T 872  \u001b[92m☑\u001b[0m 872 \n",
            "Q 36+283  T 319  \u001b[92m☑\u001b[0m 319 \n",
            "Q 8+903   T 911  \u001b[92m☑\u001b[0m 911 \n",
            "Q 734+321 T 1055 \u001b[92m☑\u001b[0m 1055\n",
            "Q 275+34  T 309  \u001b[92m☑\u001b[0m 309 \n",
            "Q 236+58  T 294  \u001b[92m☑\u001b[0m 294 \n",
            "Q 757+79  T 836  \u001b[92m☑\u001b[0m 836 \n",
            "Q 65+887  T 952  \u001b[92m☑\u001b[0m 952 \n",
            "Q 559+901 T 1460 \u001b[92m☑\u001b[0m 1460\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 195\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 295us/sample - loss: 2.0228e-04 - acc: 1.0000 - val_loss: 0.0039 - val_acc: 0.9987\n",
            "Q 86+769  T 855  \u001b[92m☑\u001b[0m 855 \n",
            "Q 893+754 T 1647 \u001b[92m☑\u001b[0m 1647\n",
            "Q 668+111 T 779  \u001b[92m☑\u001b[0m 779 \n",
            "Q 288+290 T 578  \u001b[92m☑\u001b[0m 578 \n",
            "Q 686+578 T 1264 \u001b[92m☑\u001b[0m 1264\n",
            "Q 195+582 T 777  \u001b[92m☑\u001b[0m 777 \n",
            "Q 0+833   T 833  \u001b[92m☑\u001b[0m 833 \n",
            "Q 940+17  T 957  \u001b[92m☑\u001b[0m 957 \n",
            "Q 288+9   T 297  \u001b[92m☑\u001b[0m 297 \n",
            "Q 9+749   T 758  \u001b[92m☑\u001b[0m 758 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 196\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 301us/sample - loss: 1.8319e-04 - acc: 1.0000 - val_loss: 0.0040 - val_acc: 0.9987\n",
            "Q 133+81  T 214  \u001b[92m☑\u001b[0m 214 \n",
            "Q 632+19  T 651  \u001b[92m☑\u001b[0m 651 \n",
            "Q 501+61  T 562  \u001b[92m☑\u001b[0m 562 \n",
            "Q 58+393  T 451  \u001b[92m☑\u001b[0m 451 \n",
            "Q 13+252  T 265  \u001b[92m☑\u001b[0m 265 \n",
            "Q 252+796 T 1048 \u001b[92m☑\u001b[0m 1048\n",
            "Q 185+49  T 234  \u001b[92m☑\u001b[0m 234 \n",
            "Q 718+56  T 774  \u001b[92m☑\u001b[0m 774 \n",
            "Q 531+67  T 598  \u001b[92m☑\u001b[0m 598 \n",
            "Q 63+999  T 1062 \u001b[92m☑\u001b[0m 1062\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 197\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 14s 310us/sample - loss: 1.6661e-04 - acc: 1.0000 - val_loss: 0.0039 - val_acc: 0.9988\n",
            "Q 163+0   T 163  \u001b[92m☑\u001b[0m 163 \n",
            "Q 30+241  T 271  \u001b[92m☑\u001b[0m 271 \n",
            "Q 51+529  T 580  \u001b[92m☑\u001b[0m 580 \n",
            "Q 281+42  T 323  \u001b[92m☑\u001b[0m 323 \n",
            "Q 365+768 T 1133 \u001b[92m☑\u001b[0m 1133\n",
            "Q 5+600   T 605  \u001b[92m☑\u001b[0m 605 \n",
            "Q 44+65   T 109  \u001b[92m☑\u001b[0m 109 \n",
            "Q 75+382  T 457  \u001b[92m☑\u001b[0m 457 \n",
            "Q 886+93  T 979  \u001b[92m☑\u001b[0m 979 \n",
            "Q 5+445   T 450  \u001b[92m☑\u001b[0m 450 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 198\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 297us/sample - loss: 1.7571e-04 - acc: 1.0000 - val_loss: 0.0075 - val_acc: 0.9978\n",
            "Q 61+524  T 585  \u001b[92m☑\u001b[0m 585 \n",
            "Q 672+90  T 762  \u001b[92m☑\u001b[0m 762 \n",
            "Q 242+540 T 782  \u001b[92m☑\u001b[0m 782 \n",
            "Q 295+471 T 766  \u001b[92m☑\u001b[0m 766 \n",
            "Q 486+81  T 567  \u001b[92m☑\u001b[0m 567 \n",
            "Q 657+709 T 1366 \u001b[92m☑\u001b[0m 1366\n",
            "Q 38+40   T 78   \u001b[92m☑\u001b[0m 78  \n",
            "Q 94+640  T 734  \u001b[92m☑\u001b[0m 734 \n",
            "Q 666+735 T 1401 \u001b[92m☑\u001b[0m 1401\n",
            "Q 315+359 T 674  \u001b[92m☑\u001b[0m 674 \n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 199\n",
            "Train on 45000 samples, validate on 5000 samples\n",
            "45000/45000 [==============================] - 13s 298us/sample - loss: 0.0402 - acc: 0.9883 - val_loss: 0.0078 - val_acc: 0.9972\n",
            "Q 27+670  T 697  \u001b[92m☑\u001b[0m 697 \n",
            "Q 213+0   T 213  \u001b[92m☑\u001b[0m 213 \n",
            "Q 79+356  T 435  \u001b[92m☑\u001b[0m 435 \n",
            "Q 27+54   T 81   \u001b[92m☑\u001b[0m 81  \n",
            "Q 191+60  T 251  \u001b[92m☑\u001b[0m 251 \n",
            "Q 121+4   T 125  \u001b[92m☑\u001b[0m 125 \n",
            "Q 3+822   T 825  \u001b[92m☑\u001b[0m 825 \n",
            "Q 37+198  T 235  \u001b[92m☑\u001b[0m 235 \n",
            "Q 61+416  T 477  \u001b[92m☑\u001b[0m 477 \n",
            "Q 193+406 T 599  \u001b[91m☒\u001b[0m 699 \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "CxYKXAoyI_Kz",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## The general case: canonical sequence-to-sequence\n",
        "\n",
        "In the general case, input sequences and output sequences have different lengths (e.g. machine translation) and the entire input sequence is required in order to start predicting the target. Here's how it works:\n",
        "\n",
        "+ A RNN layer (or stack thereof) acts as \"encoder\": it processes the input sequence and returns its own internal state. Note that we discard the outputs of the encoder RNN, only recovering the state. This state will serve as the \"context\", or \"conditioning\", of the decoder in the next step.\n",
        "+ Another RNN layer (or stack thereof) acts as \"decoder\": it is trained to predict the next elements of the target sequence, given previous elements of the target sequence. Specifically, it is trained to turn the target sequences into the same sequences but offset by one timestep in the future, a training process called **teacher forcing** in this context. Importantly, the encoder uses as initial state the state vectors from the encoder, which is how the decoder obtains information about what it is supposed to generate. Effectively, the decoder learns to generate ``targets[t+1...]`` given ``targets[...t]``, conditioned on the input sequence."
      ]
    },
    {
      "metadata": {
        "id": "Oq362fcAI_K0",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "In inference mode, i.e. when we want to decode unknown input sequences, we go through a slightly different process:\n",
        "\n",
        "+ Encode the input sequence into state vectors.\n",
        "+ Start with a target sequence of size 1 (just the start-of-sequence element).\n",
        "+ Feed the state vectors and 1-char target sequence to the decoder to produce predictions for the next element.\n",
        "+ Sample the next element using these predictions (we simply use ``argmax``).\n",
        "+ Append the sampled element to the target sequence\n",
        "+ Repeat until we generate the end-of-sequence element or we hit the character limit."
      ]
    },
    {
      "metadata": {
        "id": "L3jf0qFHI_K1",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "For our example implementation, we will use a dataset of pairs of English sentences and their Catalan translation (which can be downloaded from [manythings.org/anki](http://manythings.org/anki)). \n",
        "\n",
        "We will implement a **character-level sequence-to-sequence model**, processing the input character-by-character and generating the output character-by-character. Another option would be a word-level model, which tends to be more common for machine translation. \n",
        "\n",
        "Here's a summary of our process:\n",
        "\n",
        "1) Turn the sentences into 3 Numpy arrays, ``encoder_input_data``, ``decoder_input_data``, ``decoder_target_data``:\n",
        " + ``encoder_input_data`` is a 3D array of shape ``(num_pairs, max_english_sentence_length, num_english_characters)`` containing a one-hot vectorization of the English sentences.\n",
        " + ``decoder_input_data`` is a 3D array of shape ``(num_pairs, max_cat_sentence_length, num_cat_characters)`` containg a one-hot vectorization of the Catalan sentences.\n",
        " + ``decoder_target_data`` is the same as ``decoder_input_data`` but offset by one timestep. ``decoder_target_data[:, t, :]`` will be the same as ``decoder_input_data[:, t + 1, :]``.\n",
        "\n",
        "2) Train a basic LSTM-based ``Seq2Seq`` model to predict ``decoder_target_data`` given ``encoder_input_data`` and ``decoder_input_data``. Our model uses teacher forcing.\n",
        "\n",
        "3) Decode some sentences to check that the model is working (i.e. turn samples from `encoder_input_data` into corresponding samples from `decoder_target_data`).\n",
        "\n",
        "Because the training process and inference process (decoding sentences) are quite different, we use different models for both, albeit they all leverage the same inner layers.\n",
        "\n",
        "This is our training model. It leverages three key features of Keras RNNs:\n",
        "\n",
        "+ The ``return_state`` constructor argument, configuring a RNN layer to return a list where the first entry is the outputs and the next entries are the internal RNN states. This is used to recover the states of the encoder.\n",
        "+ The ``inital_state`` call argument, specifying the initial state(s) of a RNN. This is used to pass the encoder states to the decoder as initial states.\n",
        "+ The ``return_sequences`` constructor argument, configuring a RNN to return its full sequence of outputs (instead of just the last output, which the defaults behavior). This is used in the decoder."
      ]
    },
    {
      "metadata": {
        "id": "6GxHOXWy2rVv",
        "colab_type": "code",
        "outputId": "76c4f68a-4e1a-4875-e079-c28bfd2254f2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "from tensorflow.keras.models import Model\n",
        "from tensorflow.keras.layers import Input, LSTM, Dense\n",
        "import numpy as np\n",
        "import requests\n",
        "\n",
        "batch_size = 64   \n",
        "latent_dim = 256  # Latent dimensionality of the encoding space.\n",
        "num_samples = 10000  \n",
        "\n",
        "data_path = 'https://raw.githubusercontent.com/DataScienceUB/DeepLearningMaster2019/master/data/cat.txt'\n",
        "\n",
        "input_texts = []\n",
        "target_texts = []\n",
        "input_characters = set()\n",
        "target_characters = set()\n",
        "\n",
        "response = requests.get(data_path)\n",
        "data = response.text\n",
        "\n",
        "lines = data.split('\\n')\n",
        "print('There are',len(lines),'sequences such as ',end='')\n",
        "lines[11]"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "There are 607 sequences such as "
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "\"I'm at home.\\tEstic a casa.\""
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 1
        }
      ]
    },
    {
      "metadata": {
        "id": "mehRtLHKI_K2",
        "colab_type": "code",
        "outputId": "550c9f3a-64da-439d-b296-d2de00c3925e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 343
        }
      },
      "cell_type": "code",
      "source": [
        "# Vectorize the data.\n",
        "\n",
        "for line in lines[: min(num_samples, len(lines) - 1)]:\n",
        "    input_text, target_text = line.split('\\t')\n",
        "    \n",
        "    # We use \"tab\" as the \"start sequence\" character\n",
        "    # for the targets, and \"\\n\" as \"end sequence\" character.\n",
        "    target_text = '\\t' + target_text + '\\n'\n",
        "    input_texts.append(input_text)\n",
        "    target_texts.append(target_text)\n",
        "    for char in input_text:\n",
        "        if char not in input_characters:\n",
        "            input_characters.add(char)\n",
        "    for char in target_text:\n",
        "        if char not in target_characters:\n",
        "            target_characters.add(char)\n",
        "\n",
        "input_characters = sorted(list(input_characters))\n",
        "target_characters = sorted(list(target_characters))\n",
        "num_encoder_tokens = len(input_characters)\n",
        "num_decoder_tokens = len(target_characters)\n",
        "max_encoder_seq_length = max([len(txt) for txt in input_texts])\n",
        "max_decoder_seq_length = max([len(txt) for txt in target_texts])\n",
        "\n",
        "print('Number of samples:', len(input_texts))\n",
        "print('Number of unique input tokens:', num_encoder_tokens)\n",
        "print('Number of unique output tokens:', num_decoder_tokens)\n",
        "print('Max sequence length for inputs:', max_encoder_seq_length)\n",
        "print('Max sequence length for outputs:', max_decoder_seq_length)\n",
        "\n",
        "input_token_index = dict(\n",
        "    [(char, i) for i, char in enumerate(input_characters)])\n",
        "target_token_index = dict(\n",
        "    [(char, i) for i, char in enumerate(target_characters)])\n",
        "\n",
        "encoder_input_data = np.zeros(\n",
        "    (len(input_texts), max_encoder_seq_length, num_encoder_tokens),\n",
        "    dtype='float32')\n",
        "decoder_input_data = np.zeros(\n",
        "    (len(input_texts), max_decoder_seq_length, num_decoder_tokens),\n",
        "    dtype='float32')\n",
        "decoder_target_data = np.zeros(\n",
        "    (len(input_texts), max_decoder_seq_length, num_decoder_tokens),\n",
        "    dtype='float32')\n",
        "\n",
        "for i, (input_text, target_text) in enumerate(zip(input_texts, target_texts)):\n",
        "    for t, char in enumerate(input_text):\n",
        "        encoder_input_data[i, t, input_token_index[char]] = 1.\n",
        "    for t, char in enumerate(target_text):\n",
        "        # decoder_target_data is ahead of decoder_input_data by one timestep\n",
        "        decoder_input_data[i, t, target_token_index[char]] = 1.\n",
        "        if t > 0:\n",
        "            # decoder_target_data will be ahead by one timestep\n",
        "            # and will not include the start character.\n",
        "            decoder_target_data[i, t - 1, target_token_index[char]] = 1.\n",
        "\n",
        "print('\\nInput sequence: \\n',encoder_input_data[0])\n",
        "print('\\nInput sequence first letter: \\n',encoder_input_data[0][0])\n"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of samples: 606\n",
            "Number of unique input tokens: 68\n",
            "Number of unique output tokens: 81\n",
            "Max sequence length for inputs: 107\n",
            "Max sequence length for outputs: 106\n",
            "\n",
            "Input sequence: \n",
            " [[0. 0. 0. ... 0. 0. 0.]\n",
            " [0. 0. 0. ... 0. 0. 0.]\n",
            " [0. 0. 0. ... 0. 0. 0.]\n",
            " ...\n",
            " [0. 0. 0. ... 0. 0. 0.]\n",
            " [0. 0. 0. ... 0. 0. 0.]\n",
            " [0. 0. 0. ... 0. 0. 0.]]\n",
            "\n",
            "Input sequence first letter: \n",
            " [0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.\n",
            " 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0.\n",
            " 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "-KnoPkIB4XJ4",
        "colab_type": "code",
        "outputId": "dd6bfa97-3e92-49ed-cf6b-28953305bf88",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34320
        }
      },
      "cell_type": "code",
      "source": [
        "epochs = 1000\n",
        "\n",
        "# Define an input sequence and process it.\n",
        "encoder_inputs = Input(shape=(None, num_encoder_tokens))\n",
        "encoder = LSTM(latent_dim, return_state=True)\n",
        "encoder_outputs, state_h, state_c = encoder(encoder_inputs)\n",
        "\n",
        "# We discard `encoder_outputs` and only keep the states.\n",
        "encoder_states = [state_h, state_c]\n",
        "\n",
        "# Set up the decoder\n",
        "decoder_inputs = Input(shape=(None, num_decoder_tokens))\n",
        "\n",
        "# We set up our decoder to return full output sequences,\n",
        "# and to return internal states as well. We don't use the\n",
        "# return states in the training model, but we will use them in inference.\n",
        "# We use `encoder_states` as initial state.\n",
        "decoder_lstm = LSTM(latent_dim, return_sequences=True, return_state=True)\n",
        "decoder_outputs, _, _ = decoder_lstm(decoder_inputs,\n",
        "                                     initial_state=encoder_states)\n",
        "decoder_dense = Dense(num_decoder_tokens, activation='softmax')\n",
        "decoder_outputs = decoder_dense(decoder_outputs)\n",
        "\n",
        "# Define the model\n",
        "model = Model([encoder_inputs, decoder_inputs], decoder_outputs)\n",
        "\n",
        "# Run training\n",
        "model.compile(optimizer='rmsprop', loss='categorical_crossentropy')\n",
        "model.fit([encoder_input_data, decoder_input_data], decoder_target_data,\n",
        "          batch_size=batch_size,\n",
        "          epochs=epochs,\n",
        "          validation_split=0.2)\n",
        "# Save model\n",
        "model.save('s2s.h5')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 484 samples, validate on 122 samples\n",
            "Epoch 1/1000\n",
            "484/484 [==============================] - 6s 12ms/sample - loss: 0.9601 - val_loss: 1.4164\n",
            "Epoch 2/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.8033 - val_loss: 1.3877\n",
            "Epoch 3/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.7934 - val_loss: 1.3837\n",
            "Epoch 4/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.7857 - val_loss: 1.4039\n",
            "Epoch 5/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.7792 - val_loss: 1.3940\n",
            "Epoch 6/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.7867 - val_loss: 1.3846\n",
            "Epoch 7/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.7681 - val_loss: 1.3654\n",
            "Epoch 8/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.7632 - val_loss: 1.3675\n",
            "Epoch 9/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.7589 - val_loss: 1.3473\n",
            "Epoch 10/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.7486 - val_loss: 1.3209\n",
            "Epoch 11/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.7472 - val_loss: 1.3133\n",
            "Epoch 12/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.7336 - val_loss: 1.2883\n",
            "Epoch 13/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.7277 - val_loss: 1.3479\n",
            "Epoch 14/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.7173 - val_loss: 1.2934\n",
            "Epoch 15/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.7108 - val_loss: 1.2753\n",
            "Epoch 16/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.7011 - val_loss: 1.2633\n",
            "Epoch 17/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.6922 - val_loss: 1.2472\n",
            "Epoch 18/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.6837 - val_loss: 1.2264\n",
            "Epoch 19/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.6786 - val_loss: 1.2337\n",
            "Epoch 20/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.6667 - val_loss: 1.2018\n",
            "Epoch 21/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.6633 - val_loss: 1.1910\n",
            "Epoch 22/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.6545 - val_loss: 1.1704\n",
            "Epoch 23/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.6485 - val_loss: 1.1690\n",
            "Epoch 24/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.6384 - val_loss: 1.1558\n",
            "Epoch 25/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.6362 - val_loss: 1.1419\n",
            "Epoch 26/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.6281 - val_loss: 1.1504\n",
            "Epoch 27/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.6245 - val_loss: 1.1293\n",
            "Epoch 28/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.6152 - val_loss: 1.1291\n",
            "Epoch 29/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.6128 - val_loss: 1.1376\n",
            "Epoch 30/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.6047 - val_loss: 1.1364\n",
            "Epoch 31/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.6044 - val_loss: 1.1250\n",
            "Epoch 32/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5957 - val_loss: 1.0943\n",
            "Epoch 33/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5942 - val_loss: 1.0907\n",
            "Epoch 34/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5859 - val_loss: 1.0833\n",
            "Epoch 35/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5872 - val_loss: 1.1026\n",
            "Epoch 36/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5795 - val_loss: 1.0823\n",
            "Epoch 37/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5764 - val_loss: 1.0613\n",
            "Epoch 38/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5733 - val_loss: 1.0678\n",
            "Epoch 39/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5691 - val_loss: 1.0560\n",
            "Epoch 40/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5653 - val_loss: 1.0586\n",
            "Epoch 41/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5614 - val_loss: 1.0767\n",
            "Epoch 42/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5610 - val_loss: 1.0605\n",
            "Epoch 43/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5550 - val_loss: 1.0584\n",
            "Epoch 44/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5526 - val_loss: 1.0553\n",
            "Epoch 45/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5526 - val_loss: 1.0612\n",
            "Epoch 46/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5456 - val_loss: 1.0538\n",
            "Epoch 47/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5424 - val_loss: 1.0409\n",
            "Epoch 48/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5427 - val_loss: 1.0444\n",
            "Epoch 49/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5349 - val_loss: 1.0487\n",
            "Epoch 50/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5344 - val_loss: 1.0309\n",
            "Epoch 51/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5307 - val_loss: 1.0521\n",
            "Epoch 52/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5278 - val_loss: 1.0441\n",
            "Epoch 53/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5246 - val_loss: 1.0315\n",
            "Epoch 54/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5193 - val_loss: 1.0326\n",
            "Epoch 55/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.5197 - val_loss: 1.0462\n",
            "Epoch 56/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5129 - val_loss: 1.0301\n",
            "Epoch 57/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5104 - val_loss: 1.0279\n",
            "Epoch 58/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.5082 - val_loss: 1.0235\n",
            "Epoch 59/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5060 - val_loss: 1.0215\n",
            "Epoch 60/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.5002 - val_loss: 1.0326\n",
            "Epoch 61/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4995 - val_loss: 1.0185\n",
            "Epoch 62/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4971 - val_loss: 1.0378\n",
            "Epoch 63/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4910 - val_loss: 1.0164\n",
            "Epoch 64/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4882 - val_loss: 1.0175\n",
            "Epoch 65/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4844 - val_loss: 1.0261\n",
            "Epoch 66/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4834 - val_loss: 1.0177\n",
            "Epoch 67/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4806 - val_loss: 1.0209\n",
            "Epoch 68/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4731 - val_loss: 1.0227\n",
            "Epoch 69/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4723 - val_loss: 1.0202\n",
            "Epoch 70/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4681 - val_loss: 1.0267\n",
            "Epoch 71/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4655 - val_loss: 1.0295\n",
            "Epoch 72/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4612 - val_loss: 1.0166\n",
            "Epoch 73/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4583 - val_loss: 1.0270\n",
            "Epoch 74/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4540 - val_loss: 1.0130\n",
            "Epoch 75/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4506 - val_loss: 1.0304\n",
            "Epoch 76/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4475 - val_loss: 1.0385\n",
            "Epoch 77/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4428 - val_loss: 1.0495\n",
            "Epoch 78/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4423 - val_loss: 1.0457\n",
            "Epoch 79/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4355 - val_loss: 1.0325\n",
            "Epoch 80/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4335 - val_loss: 1.0577\n",
            "Epoch 81/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4264 - val_loss: 1.0503\n",
            "Epoch 82/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.4249 - val_loss: 1.0453\n",
            "Epoch 83/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4225 - val_loss: 1.0643\n",
            "Epoch 84/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4174 - val_loss: 1.0412\n",
            "Epoch 85/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4172 - val_loss: 1.0487\n",
            "Epoch 86/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4101 - val_loss: 1.0663\n",
            "Epoch 87/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4364 - val_loss: 1.0355\n",
            "Epoch 88/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.4017 - val_loss: 1.0509\n",
            "Epoch 89/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3990 - val_loss: 1.0490\n",
            "Epoch 90/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3942 - val_loss: 1.0761\n",
            "Epoch 91/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3892 - val_loss: 1.0667\n",
            "Epoch 92/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3878 - val_loss: 1.0766\n",
            "Epoch 93/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3809 - val_loss: 1.0820\n",
            "Epoch 94/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3779 - val_loss: 1.0852\n",
            "Epoch 95/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.4199 - val_loss: 1.0661\n",
            "Epoch 96/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4295 - val_loss: 1.0713\n",
            "Epoch 97/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.4337 - val_loss: 1.0876\n",
            "Epoch 98/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3983 - val_loss: 1.0763\n",
            "Epoch 99/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3915 - val_loss: 1.0503\n",
            "Epoch 100/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3731 - val_loss: 1.0690\n",
            "Epoch 101/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3677 - val_loss: 1.0857\n",
            "Epoch 102/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3727 - val_loss: 1.0777\n",
            "Epoch 103/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3597 - val_loss: 1.0877\n",
            "Epoch 104/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3479 - val_loss: 1.0961\n",
            "Epoch 105/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3450 - val_loss: 1.0925\n",
            "Epoch 106/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3383 - val_loss: 1.1237\n",
            "Epoch 107/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3369 - val_loss: 1.1077\n",
            "Epoch 108/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3305 - val_loss: 1.1179\n",
            "Epoch 109/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3286 - val_loss: 1.1364\n",
            "Epoch 110/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3236 - val_loss: 1.1262\n",
            "Epoch 111/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3221 - val_loss: 1.1266\n",
            "Epoch 112/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3124 - val_loss: 1.1455\n",
            "Epoch 113/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.3127 - val_loss: 1.1435\n",
            "Epoch 114/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3053 - val_loss: 1.1472\n",
            "Epoch 115/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3064 - val_loss: 1.1620\n",
            "Epoch 116/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3001 - val_loss: 1.1761\n",
            "Epoch 117/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2916 - val_loss: 1.1667\n",
            "Epoch 118/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2899 - val_loss: 1.1647\n",
            "Epoch 119/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2902 - val_loss: 1.1732\n",
            "Epoch 120/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2814 - val_loss: 1.1855\n",
            "Epoch 121/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2833 - val_loss: 1.2057\n",
            "Epoch 122/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2763 - val_loss: 1.1930\n",
            "Epoch 123/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2708 - val_loss: 1.2025\n",
            "Epoch 124/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2698 - val_loss: 1.1978\n",
            "Epoch 125/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2710 - val_loss: 1.2088\n",
            "Epoch 126/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2595 - val_loss: 1.2248\n",
            "Epoch 127/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2579 - val_loss: 1.2350\n",
            "Epoch 128/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2571 - val_loss: 1.2317\n",
            "Epoch 129/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3404 - val_loss: 1.1941\n",
            "Epoch 130/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3513 - val_loss: 1.2390\n",
            "Epoch 131/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.3165 - val_loss: 1.2179\n",
            "Epoch 132/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2673 - val_loss: 1.2158\n",
            "Epoch 133/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2695 - val_loss: 1.2564\n",
            "Epoch 134/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2717 - val_loss: 1.2400\n",
            "Epoch 135/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2462 - val_loss: 1.2489\n",
            "Epoch 136/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2575 - val_loss: 1.2559\n",
            "Epoch 137/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2388 - val_loss: 1.2710\n",
            "Epoch 138/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2497 - val_loss: 1.2814\n",
            "Epoch 139/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2331 - val_loss: 1.2893\n",
            "Epoch 140/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.2304 - val_loss: 1.3140\n",
            "Epoch 141/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2363 - val_loss: 1.3013\n",
            "Epoch 142/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2213 - val_loss: 1.2934\n",
            "Epoch 143/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2307 - val_loss: 1.3172\n",
            "Epoch 144/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2194 - val_loss: 1.3203\n",
            "Epoch 145/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2107 - val_loss: 1.3365\n",
            "Epoch 146/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2116 - val_loss: 1.3346\n",
            "Epoch 147/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2125 - val_loss: 1.3359\n",
            "Epoch 148/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2070 - val_loss: 1.3680\n",
            "Epoch 149/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.2035 - val_loss: 1.3554\n",
            "Epoch 150/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2024 - val_loss: 1.3498\n",
            "Epoch 151/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1956 - val_loss: 1.3588\n",
            "Epoch 152/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1927 - val_loss: 1.3821\n",
            "Epoch 153/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1907 - val_loss: 1.3836\n",
            "Epoch 154/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1845 - val_loss: 1.4223\n",
            "Epoch 155/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1818 - val_loss: 1.3953\n",
            "Epoch 156/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1849 - val_loss: 1.4180\n",
            "Epoch 157/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1791 - val_loss: 1.4238\n",
            "Epoch 158/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1755 - val_loss: 1.4354\n",
            "Epoch 159/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1728 - val_loss: 1.4374\n",
            "Epoch 160/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1710 - val_loss: 1.4131\n",
            "Epoch 161/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1706 - val_loss: 1.4571\n",
            "Epoch 162/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1655 - val_loss: 1.4464\n",
            "Epoch 163/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1596 - val_loss: 1.4783\n",
            "Epoch 164/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1643 - val_loss: 1.4548\n",
            "Epoch 165/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1568 - val_loss: 1.4595\n",
            "Epoch 166/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1568 - val_loss: 1.5177\n",
            "Epoch 167/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1574 - val_loss: 1.5076\n",
            "Epoch 168/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1510 - val_loss: 1.5072\n",
            "Epoch 169/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1517 - val_loss: 1.5334\n",
            "Epoch 170/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1473 - val_loss: 1.5582\n",
            "Epoch 171/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1530 - val_loss: 1.5372\n",
            "Epoch 172/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1416 - val_loss: 1.5300\n",
            "Epoch 173/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1429 - val_loss: 1.5369\n",
            "Epoch 174/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1415 - val_loss: 1.5523\n",
            "Epoch 175/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1390 - val_loss: 1.5571\n",
            "Epoch 176/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1365 - val_loss: 1.5848\n",
            "Epoch 177/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1365 - val_loss: 1.5995\n",
            "Epoch 178/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1347 - val_loss: 1.6048\n",
            "Epoch 179/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1314 - val_loss: 1.6043\n",
            "Epoch 180/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1288 - val_loss: 1.6292\n",
            "Epoch 181/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1314 - val_loss: 1.6023\n",
            "Epoch 182/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1267 - val_loss: 1.6343\n",
            "Epoch 183/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1255 - val_loss: 1.6166\n",
            "Epoch 184/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1218 - val_loss: 1.6237\n",
            "Epoch 185/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1249 - val_loss: 1.6379\n",
            "Epoch 186/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1176 - val_loss: 1.6316\n",
            "Epoch 187/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1272 - val_loss: 1.6489\n",
            "Epoch 188/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1148 - val_loss: 1.6776\n",
            "Epoch 189/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1226 - val_loss: 1.6698\n",
            "Epoch 190/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1124 - val_loss: 1.6728\n",
            "Epoch 191/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1151 - val_loss: 1.6769\n",
            "Epoch 192/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2160 - val_loss: 1.5673\n",
            "Epoch 193/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1692 - val_loss: 1.6244\n",
            "Epoch 194/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.1066 - val_loss: 1.6248\n",
            "Epoch 195/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1074 - val_loss: 1.6559\n",
            "Epoch 196/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1120 - val_loss: 1.6629\n",
            "Epoch 197/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1096 - val_loss: 1.6850\n",
            "Epoch 198/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1065 - val_loss: 1.6816\n",
            "Epoch 199/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1120 - val_loss: 1.6939\n",
            "Epoch 200/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1021 - val_loss: 1.7037\n",
            "Epoch 201/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1072 - val_loss: 1.7440\n",
            "Epoch 202/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1036 - val_loss: 1.7399\n",
            "Epoch 203/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1049 - val_loss: 1.7343\n",
            "Epoch 204/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.1040 - val_loss: 1.7409\n",
            "Epoch 205/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0992 - val_loss: 1.7517\n",
            "Epoch 206/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1017 - val_loss: 1.7716\n",
            "Epoch 207/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.1020 - val_loss: 1.7669\n",
            "Epoch 208/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0967 - val_loss: 1.7681\n",
            "Epoch 209/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0993 - val_loss: 1.7941\n",
            "Epoch 210/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0961 - val_loss: 1.7793\n",
            "Epoch 211/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0967 - val_loss: 1.7995\n",
            "Epoch 212/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0969 - val_loss: 1.7988\n",
            "Epoch 213/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0952 - val_loss: 1.8372\n",
            "Epoch 214/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0964 - val_loss: 1.8049\n",
            "Epoch 215/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0914 - val_loss: 1.8270\n",
            "Epoch 216/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0946 - val_loss: 1.8387\n",
            "Epoch 217/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.1036 - val_loss: 1.8129\n",
            "Epoch 218/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0907 - val_loss: 1.8145\n",
            "Epoch 219/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0882 - val_loss: 1.8314\n",
            "Epoch 220/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0908 - val_loss: 1.8684\n",
            "Epoch 221/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0911 - val_loss: 1.8575\n",
            "Epoch 222/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0945 - val_loss: 1.8022\n",
            "Epoch 223/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0899 - val_loss: 1.8555\n",
            "Epoch 224/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0859 - val_loss: 1.8589\n",
            "Epoch 225/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0861 - val_loss: 1.8881\n",
            "Epoch 226/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0922 - val_loss: 1.8515\n",
            "Epoch 227/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0881 - val_loss: 1.8859\n",
            "Epoch 228/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0847 - val_loss: 1.9132\n",
            "Epoch 229/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0855 - val_loss: 1.9155\n",
            "Epoch 230/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0882 - val_loss: 1.8945\n",
            "Epoch 231/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0852 - val_loss: 1.9010\n",
            "Epoch 232/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0843 - val_loss: 1.9380\n",
            "Epoch 233/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0844 - val_loss: 1.9053\n",
            "Epoch 234/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0857 - val_loss: 1.9503\n",
            "Epoch 235/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0833 - val_loss: 2.0325\n",
            "Epoch 236/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.2062 - val_loss: 1.8350\n",
            "Epoch 237/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0803 - val_loss: 1.8596\n",
            "Epoch 238/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0783 - val_loss: 1.8727\n",
            "Epoch 239/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0788 - val_loss: 1.8985\n",
            "Epoch 240/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0802 - val_loss: 1.9053\n",
            "Epoch 241/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0811 - val_loss: 1.9176\n",
            "Epoch 242/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0834 - val_loss: 1.9173\n",
            "Epoch 243/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0807 - val_loss: 1.9186\n",
            "Epoch 244/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0810 - val_loss: 1.9455\n",
            "Epoch 245/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0805 - val_loss: 1.9452\n",
            "Epoch 246/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0816 - val_loss: 1.9405\n",
            "Epoch 247/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0796 - val_loss: 1.9362\n",
            "Epoch 248/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0801 - val_loss: 1.9709\n",
            "Epoch 249/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0829 - val_loss: 1.9568\n",
            "Epoch 250/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0788 - val_loss: 1.9417\n",
            "Epoch 251/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0768 - val_loss: 1.9721\n",
            "Epoch 252/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0782 - val_loss: 1.9823\n",
            "Epoch 253/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0839 - val_loss: 1.9737\n",
            "Epoch 254/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0773 - val_loss: 1.9955\n",
            "Epoch 255/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0776 - val_loss: 1.9868\n",
            "Epoch 256/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0783 - val_loss: 2.0112\n",
            "Epoch 257/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0927 - val_loss: 1.9795\n",
            "Epoch 258/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0747 - val_loss: 1.9803\n",
            "Epoch 259/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0741 - val_loss: 1.9973\n",
            "Epoch 260/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0746 - val_loss: 2.0086\n",
            "Epoch 261/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0756 - val_loss: 1.9977\n",
            "Epoch 262/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0772 - val_loss: 2.0222\n",
            "Epoch 263/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0755 - val_loss: 2.0356\n",
            "Epoch 264/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0756 - val_loss: 2.0173\n",
            "Epoch 265/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0755 - val_loss: 2.0220\n",
            "Epoch 266/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0759 - val_loss: 2.0514\n",
            "Epoch 267/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0768 - val_loss: 2.0052\n",
            "Epoch 268/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0777 - val_loss: 2.0133\n",
            "Epoch 269/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0766 - val_loss: 2.0248\n",
            "Epoch 270/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0727 - val_loss: 2.0438\n",
            "Epoch 271/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0749 - val_loss: 2.0127\n",
            "Epoch 272/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0747 - val_loss: 2.0724\n",
            "Epoch 273/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0758 - val_loss: 2.0651\n",
            "Epoch 274/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0739 - val_loss: 2.0594\n",
            "Epoch 275/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0738 - val_loss: 2.0738\n",
            "Epoch 276/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0746 - val_loss: 2.0616\n",
            "Epoch 277/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0746 - val_loss: 2.0613\n",
            "Epoch 278/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0748 - val_loss: 2.0715\n",
            "Epoch 279/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0737 - val_loss: 2.0774\n",
            "Epoch 280/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0734 - val_loss: 2.1005\n",
            "Epoch 281/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0738 - val_loss: 2.0853\n",
            "Epoch 282/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0777 - val_loss: 2.0619\n",
            "Epoch 283/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0735 - val_loss: 2.0859\n",
            "Epoch 284/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0726 - val_loss: 2.0382\n",
            "Epoch 285/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0786 - val_loss: 2.0550\n",
            "Epoch 286/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0715 - val_loss: 2.0484\n",
            "Epoch 287/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0715 - val_loss: 2.0808\n",
            "Epoch 288/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0714 - val_loss: 2.0702\n",
            "Epoch 289/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0718 - val_loss: 2.0862\n",
            "Epoch 290/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0736 - val_loss: 2.0979\n",
            "Epoch 291/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0727 - val_loss: 2.1005\n",
            "Epoch 292/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0726 - val_loss: 2.1118\n",
            "Epoch 293/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0734 - val_loss: 2.0887\n",
            "Epoch 294/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0721 - val_loss: 2.0980\n",
            "Epoch 295/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0720 - val_loss: 2.1052\n",
            "Epoch 296/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0719 - val_loss: 2.1129\n",
            "Epoch 297/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0720 - val_loss: 2.1260\n",
            "Epoch 298/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0716 - val_loss: 2.1216\n",
            "Epoch 299/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0715 - val_loss: 2.1445\n",
            "Epoch 300/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0822 - val_loss: 2.0440\n",
            "Epoch 301/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0758 - val_loss: 2.0917\n",
            "Epoch 302/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0689 - val_loss: 2.1036\n",
            "Epoch 303/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0722 - val_loss: 2.1189\n",
            "Epoch 304/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0710 - val_loss: 2.1088\n",
            "Epoch 305/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0721 - val_loss: 2.1254\n",
            "Epoch 306/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0709 - val_loss: 2.1472\n",
            "Epoch 307/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0711 - val_loss: 2.1455\n",
            "Epoch 308/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0711 - val_loss: 2.1682\n",
            "Epoch 309/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0720 - val_loss: 2.1580\n",
            "Epoch 310/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0702 - val_loss: 2.1426\n",
            "Epoch 311/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0702 - val_loss: 2.1930\n",
            "Epoch 312/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0715 - val_loss: 2.1444\n",
            "Epoch 313/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0706 - val_loss: 2.1398\n",
            "Epoch 314/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0703 - val_loss: 2.1723\n",
            "Epoch 315/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0704 - val_loss: 2.1573\n",
            "Epoch 316/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.2282 - val_loss: 2.0046\n",
            "Epoch 317/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0702 - val_loss: 2.0413\n",
            "Epoch 318/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0671 - val_loss: 2.0606\n",
            "Epoch 319/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0674 - val_loss: 2.0628\n",
            "Epoch 320/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0674 - val_loss: 2.0728\n",
            "Epoch 321/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0682 - val_loss: 2.0698\n",
            "Epoch 322/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0687 - val_loss: 2.0785\n",
            "Epoch 323/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0694 - val_loss: 2.1004\n",
            "Epoch 324/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0698 - val_loss: 2.0945\n",
            "Epoch 325/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0707 - val_loss: 2.1314\n",
            "Epoch 326/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0695 - val_loss: 2.0877\n",
            "Epoch 327/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0698 - val_loss: 2.1164\n",
            "Epoch 328/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0699 - val_loss: 2.1065\n",
            "Epoch 329/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0700 - val_loss: 2.1474\n",
            "Epoch 330/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0719 - val_loss: 2.1071\n",
            "Epoch 331/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0701 - val_loss: 2.0979\n",
            "Epoch 332/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0694 - val_loss: 2.1149\n",
            "Epoch 333/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0682 - val_loss: 2.1154\n",
            "Epoch 334/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0681 - val_loss: 2.1387\n",
            "Epoch 335/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0688 - val_loss: 2.1467\n",
            "Epoch 336/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0691 - val_loss: 2.1565\n",
            "Epoch 337/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0691 - val_loss: 2.1789\n",
            "Epoch 338/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.1230 - val_loss: 2.1083\n",
            "Epoch 339/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0662 - val_loss: 2.1215\n",
            "Epoch 340/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0660 - val_loss: 2.1368\n",
            "Epoch 341/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0669 - val_loss: 2.1449\n",
            "Epoch 342/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0669 - val_loss: 2.1406\n",
            "Epoch 343/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0673 - val_loss: 2.1521\n",
            "Epoch 344/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0676 - val_loss: 2.1524\n",
            "Epoch 345/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0680 - val_loss: 2.1416\n",
            "Epoch 346/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0674 - val_loss: 2.1596\n",
            "Epoch 347/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0670 - val_loss: 2.1759\n",
            "Epoch 348/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0665 - val_loss: 2.1807\n",
            "Epoch 349/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0665 - val_loss: 2.1671\n",
            "Epoch 350/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0667 - val_loss: 2.1838\n",
            "Epoch 351/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0665 - val_loss: 2.1940\n",
            "Epoch 352/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0658 - val_loss: 2.1639\n",
            "Epoch 353/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0657 - val_loss: 2.1590\n",
            "Epoch 354/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0646 - val_loss: 2.1829\n",
            "Epoch 355/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0650 - val_loss: 2.1729\n",
            "Epoch 356/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0665 - val_loss: 2.1476\n",
            "Epoch 357/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0666 - val_loss: 2.1651\n",
            "Epoch 358/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0631 - val_loss: 2.1579\n",
            "Epoch 359/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0638 - val_loss: 2.1934\n",
            "Epoch 360/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0629 - val_loss: 2.1763\n",
            "Epoch 361/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0634 - val_loss: 2.1764\n",
            "Epoch 362/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0622 - val_loss: 2.1537\n",
            "Epoch 363/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0630 - val_loss: 2.1733\n",
            "Epoch 364/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0681 - val_loss: 2.1751\n",
            "Epoch 365/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0654 - val_loss: 2.1858\n",
            "Epoch 366/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0639 - val_loss: 2.1756\n",
            "Epoch 367/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0643 - val_loss: 2.1999\n",
            "Epoch 368/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0637 - val_loss: 2.2181\n",
            "Epoch 369/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0632 - val_loss: 2.2110\n",
            "Epoch 370/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0639 - val_loss: 2.2311\n",
            "Epoch 371/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.1485 - val_loss: 2.1188\n",
            "Epoch 372/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0594 - val_loss: 2.1315\n",
            "Epoch 373/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0590 - val_loss: 2.1420\n",
            "Epoch 374/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0600 - val_loss: 2.1584\n",
            "Epoch 375/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0598 - val_loss: 2.1729\n",
            "Epoch 376/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0601 - val_loss: 2.1657\n",
            "Epoch 377/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0596 - val_loss: 2.1894\n",
            "Epoch 378/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0602 - val_loss: 2.1964\n",
            "Epoch 379/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0600 - val_loss: 2.1856\n",
            "Epoch 380/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0595 - val_loss: 2.1875\n",
            "Epoch 381/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0598 - val_loss: 2.2111\n",
            "Epoch 382/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0594 - val_loss: 2.2281\n",
            "Epoch 383/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0592 - val_loss: 2.1992\n",
            "Epoch 384/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0596 - val_loss: 2.2137\n",
            "Epoch 385/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0586 - val_loss: 2.2078\n",
            "Epoch 386/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0581 - val_loss: 2.2338\n",
            "Epoch 387/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0594 - val_loss: 2.2338\n",
            "Epoch 388/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0580 - val_loss: 2.2423\n",
            "Epoch 389/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0596 - val_loss: 2.2189\n",
            "Epoch 390/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0591 - val_loss: 2.2284\n",
            "Epoch 391/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0590 - val_loss: 2.2376\n",
            "Epoch 392/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0579 - val_loss: 2.2342\n",
            "Epoch 393/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0586 - val_loss: 2.2530\n",
            "Epoch 394/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0579 - val_loss: 2.2348\n",
            "Epoch 395/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0571 - val_loss: 2.2305\n",
            "Epoch 396/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0562 - val_loss: 2.2544\n",
            "Epoch 397/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0576 - val_loss: 2.2105\n",
            "Epoch 398/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0562 - val_loss: 2.2744\n",
            "Epoch 399/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0642 - val_loss: 2.2080\n",
            "Epoch 400/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0561 - val_loss: 2.1925\n",
            "Epoch 401/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0543 - val_loss: 2.2225\n",
            "Epoch 402/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0537 - val_loss: 2.2282\n",
            "Epoch 403/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0544 - val_loss: 2.2504\n",
            "Epoch 404/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0546 - val_loss: 2.2334\n",
            "Epoch 405/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0552 - val_loss: 2.2638\n",
            "Epoch 406/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0540 - val_loss: 2.2519\n",
            "Epoch 407/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0557 - val_loss: 2.2722\n",
            "Epoch 408/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0545 - val_loss: 2.2641\n",
            "Epoch 409/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0547 - val_loss: 2.2706\n",
            "Epoch 410/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0560 - val_loss: 2.2469\n",
            "Epoch 411/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0541 - val_loss: 2.2700\n",
            "Epoch 412/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0537 - val_loss: 2.2808\n",
            "Epoch 413/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0534 - val_loss: 2.2740\n",
            "Epoch 414/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0534 - val_loss: 2.2995\n",
            "Epoch 415/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0542 - val_loss: 2.2917\n",
            "Epoch 416/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0548 - val_loss: 2.2601\n",
            "Epoch 417/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0522 - val_loss: 2.2726\n",
            "Epoch 418/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0528 - val_loss: 2.2962\n",
            "Epoch 419/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0534 - val_loss: 2.2693\n",
            "Epoch 420/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0523 - val_loss: 2.3057\n",
            "Epoch 421/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0528 - val_loss: 2.3173\n",
            "Epoch 422/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0522 - val_loss: 2.3164\n",
            "Epoch 423/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0518 - val_loss: 2.3138\n",
            "Epoch 424/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0515 - val_loss: 2.3017\n",
            "Epoch 425/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0537 - val_loss: 2.2746\n",
            "Epoch 426/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0549 - val_loss: 2.2463\n",
            "Epoch 427/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0540 - val_loss: 2.2596\n",
            "Epoch 428/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0496 - val_loss: 2.2766\n",
            "Epoch 429/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0498 - val_loss: 2.2816\n",
            "Epoch 430/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0497 - val_loss: 2.2848\n",
            "Epoch 431/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0505 - val_loss: 2.2904\n",
            "Epoch 432/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0500 - val_loss: 2.2860\n",
            "Epoch 433/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0506 - val_loss: 2.3012\n",
            "Epoch 434/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0506 - val_loss: 2.3226\n",
            "Epoch 435/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0508 - val_loss: 2.3138\n",
            "Epoch 436/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0505 - val_loss: 2.3233\n",
            "Epoch 437/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0503 - val_loss: 2.3109\n",
            "Epoch 438/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0509 - val_loss: 2.2944\n",
            "Epoch 439/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0499 - val_loss: 2.3223\n",
            "Epoch 440/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0490 - val_loss: 2.3299\n",
            "Epoch 441/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0498 - val_loss: 2.3457\n",
            "Epoch 442/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0493 - val_loss: 2.3244\n",
            "Epoch 443/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0523 - val_loss: 2.3214\n",
            "Epoch 444/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0493 - val_loss: 2.2978\n",
            "Epoch 445/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0491 - val_loss: 2.3165\n",
            "Epoch 446/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0471 - val_loss: 2.2796\n",
            "Epoch 447/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0480 - val_loss: 2.2991\n",
            "Epoch 448/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0481 - val_loss: 2.3241\n",
            "Epoch 449/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0489 - val_loss: 2.2936\n",
            "Epoch 450/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0506 - val_loss: 2.2910\n",
            "Epoch 451/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0497 - val_loss: 2.2735\n",
            "Epoch 452/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0465 - val_loss: 2.2874\n",
            "Epoch 453/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0470 - val_loss: 2.3000\n",
            "Epoch 454/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0478 - val_loss: 2.3082\n",
            "Epoch 455/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0477 - val_loss: 2.3220\n",
            "Epoch 456/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0471 - val_loss: 2.3310\n",
            "Epoch 457/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0480 - val_loss: 2.3348\n",
            "Epoch 458/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0478 - val_loss: 2.3484\n",
            "Epoch 459/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0487 - val_loss: 2.3231\n",
            "Epoch 460/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0557 - val_loss: 2.3189\n",
            "Epoch 461/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0460 - val_loss: 2.2929\n",
            "Epoch 462/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0454 - val_loss: 2.2967\n",
            "Epoch 463/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0452 - val_loss: 2.3350\n",
            "Epoch 464/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0453 - val_loss: 2.3272\n",
            "Epoch 465/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0476 - val_loss: 2.3240\n",
            "Epoch 466/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0453 - val_loss: 2.3398\n",
            "Epoch 467/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0458 - val_loss: 2.3289\n",
            "Epoch 468/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0462 - val_loss: 2.3358\n",
            "Epoch 469/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0468 - val_loss: 2.3704\n",
            "Epoch 470/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0469 - val_loss: 2.3614\n",
            "Epoch 471/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0454 - val_loss: 2.3521\n",
            "Epoch 472/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0462 - val_loss: 2.3666\n",
            "Epoch 473/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0458 - val_loss: 2.3403\n",
            "Epoch 474/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0464 - val_loss: 2.3515\n",
            "Epoch 475/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0442 - val_loss: 2.3405\n",
            "Epoch 476/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0446 - val_loss: 2.3473\n",
            "Epoch 477/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0466 - val_loss: 2.3530\n",
            "Epoch 478/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0446 - val_loss: 2.3360\n",
            "Epoch 479/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0460 - val_loss: 2.3317\n",
            "Epoch 480/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0434 - val_loss: 2.3063\n",
            "Epoch 481/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0433 - val_loss: 2.3028\n",
            "Epoch 482/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0443 - val_loss: 2.3360\n",
            "Epoch 483/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0450 - val_loss: 2.3463\n",
            "Epoch 484/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0438 - val_loss: 2.3580\n",
            "Epoch 485/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0438 - val_loss: 2.3547\n",
            "Epoch 486/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0441 - val_loss: 2.3782\n",
            "Epoch 487/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0435 - val_loss: 2.3532\n",
            "Epoch 488/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0475 - val_loss: 2.3695\n",
            "Epoch 489/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0435 - val_loss: 2.3771\n",
            "Epoch 490/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0431 - val_loss: 2.3775\n",
            "Epoch 491/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0429 - val_loss: 2.3708\n",
            "Epoch 492/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0434 - val_loss: 2.3814\n",
            "Epoch 493/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0434 - val_loss: 2.3659\n",
            "Epoch 494/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0426 - val_loss: 2.3853\n",
            "Epoch 495/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0464 - val_loss: 2.3716\n",
            "Epoch 496/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0432 - val_loss: 2.3593\n",
            "Epoch 497/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0416 - val_loss: 2.3870\n",
            "Epoch 498/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0410 - val_loss: 2.3629\n",
            "Epoch 499/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0418 - val_loss: 2.3781\n",
            "Epoch 500/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0434 - val_loss: 2.3776\n",
            "Epoch 501/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0418 - val_loss: 2.3912\n",
            "Epoch 502/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0434 - val_loss: 2.3803\n",
            "Epoch 503/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0415 - val_loss: 2.4015\n",
            "Epoch 504/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0432 - val_loss: 2.4043\n",
            "Epoch 505/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0414 - val_loss: 2.4056\n",
            "Epoch 506/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0419 - val_loss: 2.3961\n",
            "Epoch 507/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0409 - val_loss: 2.4026\n",
            "Epoch 508/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0423 - val_loss: 2.4021\n",
            "Epoch 509/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0426 - val_loss: 2.4050\n",
            "Epoch 510/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0402 - val_loss: 2.4136\n",
            "Epoch 511/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0416 - val_loss: 2.4064\n",
            "Epoch 512/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0407 - val_loss: 2.4224\n",
            "Epoch 513/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0406 - val_loss: 2.4061\n",
            "Epoch 514/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0410 - val_loss: 2.4261\n",
            "Epoch 515/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0420 - val_loss: 2.4091\n",
            "Epoch 516/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0396 - val_loss: 2.4263\n",
            "Epoch 517/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0397 - val_loss: 2.4283\n",
            "Epoch 518/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0408 - val_loss: 2.4286\n",
            "Epoch 519/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0395 - val_loss: 2.4216\n",
            "Epoch 520/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0398 - val_loss: 2.4395\n",
            "Epoch 521/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0408 - val_loss: 2.4362\n",
            "Epoch 522/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0388 - val_loss: 2.4316\n",
            "Epoch 523/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0403 - val_loss: 2.3851\n",
            "Epoch 524/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0465 - val_loss: 2.3815\n",
            "Epoch 525/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0426 - val_loss: 2.3822\n",
            "Epoch 526/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0394 - val_loss: 2.3875\n",
            "Epoch 527/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0404 - val_loss: 2.3838\n",
            "Epoch 528/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0403 - val_loss: 2.3835\n",
            "Epoch 529/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0403 - val_loss: 2.3918\n",
            "Epoch 530/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0418 - val_loss: 2.4427\n",
            "Epoch 531/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0408 - val_loss: 2.4033\n",
            "Epoch 532/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0420 - val_loss: 2.4365\n",
            "Epoch 533/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0390 - val_loss: 2.4283\n",
            "Epoch 534/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0410 - val_loss: 2.4274\n",
            "Epoch 535/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0402 - val_loss: 2.4265\n",
            "Epoch 536/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0394 - val_loss: 2.4320\n",
            "Epoch 537/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0396 - val_loss: 2.4155\n",
            "Epoch 538/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0381 - val_loss: 2.4470\n",
            "Epoch 539/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0383 - val_loss: 2.4283\n",
            "Epoch 540/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0391 - val_loss: 2.4440\n",
            "Epoch 541/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0375 - val_loss: 2.4065\n",
            "Epoch 542/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0431 - val_loss: 2.4023\n",
            "Epoch 543/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0389 - val_loss: 2.3859\n",
            "Epoch 544/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0364 - val_loss: 2.4207\n",
            "Epoch 545/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0356 - val_loss: 2.3998\n",
            "Epoch 546/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0390 - val_loss: 2.4063\n",
            "Epoch 547/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0448 - val_loss: 2.4230\n",
            "Epoch 548/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0396 - val_loss: 2.4216\n",
            "Epoch 549/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0388 - val_loss: 2.4194\n",
            "Epoch 550/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0387 - val_loss: 2.4343\n",
            "Epoch 551/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0380 - val_loss: 2.4465\n",
            "Epoch 552/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0390 - val_loss: 2.4232\n",
            "Epoch 553/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0374 - val_loss: 2.4242\n",
            "Epoch 554/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0387 - val_loss: 2.4315\n",
            "Epoch 555/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0400 - val_loss: 2.4119\n",
            "Epoch 556/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0386 - val_loss: 2.4284\n",
            "Epoch 557/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0374 - val_loss: 2.4052\n",
            "Epoch 558/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0352 - val_loss: 2.4266\n",
            "Epoch 559/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0367 - val_loss: 2.4124\n",
            "Epoch 560/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0350 - val_loss: 2.4189\n",
            "Epoch 561/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0358 - val_loss: 2.4284\n",
            "Epoch 562/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0372 - val_loss: 2.4262\n",
            "Epoch 563/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0363 - val_loss: 2.4738\n",
            "Epoch 564/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0378 - val_loss: 2.4025\n",
            "Epoch 565/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0342 - val_loss: 2.4421\n",
            "Epoch 566/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0346 - val_loss: 2.3962\n",
            "Epoch 567/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0360 - val_loss: 2.4413\n",
            "Epoch 568/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0373 - val_loss: 2.4259\n",
            "Epoch 569/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0369 - val_loss: 2.3822\n",
            "Epoch 570/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0340 - val_loss: 2.4200\n",
            "Epoch 571/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0350 - val_loss: 2.4310\n",
            "Epoch 572/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0343 - val_loss: 2.4193\n",
            "Epoch 573/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0340 - val_loss: 2.4150\n",
            "Epoch 574/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0348 - val_loss: 2.4284\n",
            "Epoch 575/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0358 - val_loss: 2.4537\n",
            "Epoch 576/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0340 - val_loss: 2.4488\n",
            "Epoch 577/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0344 - val_loss: 2.4603\n",
            "Epoch 578/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0344 - val_loss: 2.4486\n",
            "Epoch 579/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0348 - val_loss: 2.4615\n",
            "Epoch 580/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0335 - val_loss: 2.4493\n",
            "Epoch 581/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0344 - val_loss: 2.4512\n",
            "Epoch 582/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0330 - val_loss: 2.4531\n",
            "Epoch 583/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0339 - val_loss: 2.4675\n",
            "Epoch 584/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0332 - val_loss: 2.4857\n",
            "Epoch 585/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0344 - val_loss: 2.4586\n",
            "Epoch 586/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0328 - val_loss: 2.4626\n",
            "Epoch 587/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0418 - val_loss: 2.4583\n",
            "Epoch 588/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0427 - val_loss: 2.4243\n",
            "Epoch 589/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0388 - val_loss: 2.4362\n",
            "Epoch 590/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0348 - val_loss: 2.4360\n",
            "Epoch 591/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0347 - val_loss: 2.4437\n",
            "Epoch 592/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0347 - val_loss: 2.4561\n",
            "Epoch 593/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0349 - val_loss: 2.4595\n",
            "Epoch 594/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0346 - val_loss: 2.4824\n",
            "Epoch 595/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0371 - val_loss: 2.4731\n",
            "Epoch 596/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0335 - val_loss: 2.4754\n",
            "Epoch 597/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0341 - val_loss: 2.4872\n",
            "Epoch 598/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0346 - val_loss: 2.4991\n",
            "Epoch 599/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0352 - val_loss: 2.4966\n",
            "Epoch 600/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0349 - val_loss: 2.5090\n",
            "Epoch 601/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0329 - val_loss: 2.4619\n",
            "Epoch 602/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0317 - val_loss: 2.4683\n",
            "Epoch 603/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0331 - val_loss: 2.4717\n",
            "Epoch 604/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0331 - val_loss: 2.4771\n",
            "Epoch 605/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0335 - val_loss: 2.4831\n",
            "Epoch 606/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0317 - val_loss: 2.4913\n",
            "Epoch 607/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0331 - val_loss: 2.4912\n",
            "Epoch 608/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0319 - val_loss: 2.5120\n",
            "Epoch 609/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0334 - val_loss: 2.5072\n",
            "Epoch 610/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0323 - val_loss: 2.4993\n",
            "Epoch 611/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0332 - val_loss: 2.5125\n",
            "Epoch 612/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0363 - val_loss: 2.5475\n",
            "Epoch 613/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0387 - val_loss: 2.4666\n",
            "Epoch 614/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0333 - val_loss: 2.4689\n",
            "Epoch 615/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0303 - val_loss: 2.4533\n",
            "Epoch 616/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0300 - val_loss: 2.4873\n",
            "Epoch 617/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0302 - val_loss: 2.4923\n",
            "Epoch 618/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0314 - val_loss: 2.4966\n",
            "Epoch 619/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0325 - val_loss: 2.4906\n",
            "Epoch 620/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0355 - val_loss: 2.4450\n",
            "Epoch 621/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0345 - val_loss: 2.4716\n",
            "Epoch 622/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0306 - val_loss: 2.4759\n",
            "Epoch 623/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0312 - val_loss: 2.4800\n",
            "Epoch 624/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0299 - val_loss: 2.4928\n",
            "Epoch 625/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0326 - val_loss: 2.4919\n",
            "Epoch 626/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0293 - val_loss: 2.5053\n",
            "Epoch 627/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0302 - val_loss: 2.5049\n",
            "Epoch 628/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0318 - val_loss: 2.4919\n",
            "Epoch 629/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0305 - val_loss: 2.5168\n",
            "Epoch 630/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0303 - val_loss: 2.5029\n",
            "Epoch 631/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0310 - val_loss: 2.5195\n",
            "Epoch 632/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0305 - val_loss: 2.5141\n",
            "Epoch 633/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0293 - val_loss: 2.5152\n",
            "Epoch 634/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0307 - val_loss: 2.5059\n",
            "Epoch 635/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0329 - val_loss: 2.5223\n",
            "Epoch 636/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0290 - val_loss: 2.5054\n",
            "Epoch 637/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0291 - val_loss: 2.5217\n",
            "Epoch 638/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0294 - val_loss: 2.5444\n",
            "Epoch 639/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0305 - val_loss: 2.5369\n",
            "Epoch 640/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0302 - val_loss: 2.5273\n",
            "Epoch 641/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0279 - val_loss: 2.5411\n",
            "Epoch 642/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0283 - val_loss: 2.5005\n",
            "Epoch 643/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0303 - val_loss: 2.5521\n",
            "Epoch 644/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0305 - val_loss: 2.5075\n",
            "Epoch 645/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0290 - val_loss: 2.5291\n",
            "Epoch 646/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0298 - val_loss: 2.4871\n",
            "Epoch 647/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0281 - val_loss: 2.5205\n",
            "Epoch 648/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0278 - val_loss: 2.5200\n",
            "Epoch 649/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0295 - val_loss: 2.5237\n",
            "Epoch 650/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0272 - val_loss: 2.5342\n",
            "Epoch 651/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0338 - val_loss: 2.4883\n",
            "Epoch 652/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0281 - val_loss: 2.4913\n",
            "Epoch 653/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0261 - val_loss: 2.4753\n",
            "Epoch 654/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0253 - val_loss: 2.5047\n",
            "Epoch 655/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0259 - val_loss: 2.4923\n",
            "Epoch 656/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0260 - val_loss: 2.5294\n",
            "Epoch 657/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0265 - val_loss: 2.5068\n",
            "Epoch 658/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0282 - val_loss: 2.5428\n",
            "Epoch 659/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0270 - val_loss: 2.5265\n",
            "Epoch 660/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0276 - val_loss: 2.5278\n",
            "Epoch 661/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0276 - val_loss: 2.5381\n",
            "Epoch 662/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0276 - val_loss: 2.5410\n",
            "Epoch 663/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0285 - val_loss: 2.5330\n",
            "Epoch 664/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0272 - val_loss: 2.5443\n",
            "Epoch 665/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0267 - val_loss: 2.5625\n",
            "Epoch 666/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0286 - val_loss: 2.5112\n",
            "Epoch 667/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0294 - val_loss: 2.5589\n",
            "Epoch 668/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0254 - val_loss: 2.5396\n",
            "Epoch 669/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0250 - val_loss: 2.5593\n",
            "Epoch 670/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0277 - val_loss: 2.5420\n",
            "Epoch 671/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0248 - val_loss: 2.5468\n",
            "Epoch 672/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0265 - val_loss: 2.5402\n",
            "Epoch 673/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0283 - val_loss: 2.5528\n",
            "Epoch 674/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0251 - val_loss: 2.5747\n",
            "Epoch 675/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0270 - val_loss: 2.5417\n",
            "Epoch 676/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0262 - val_loss: 2.5618\n",
            "Epoch 677/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0286 - val_loss: 2.5768\n",
            "Epoch 678/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0260 - val_loss: 2.5437\n",
            "Epoch 679/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0255 - val_loss: 2.5391\n",
            "Epoch 680/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0248 - val_loss: 2.5369\n",
            "Epoch 681/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0256 - val_loss: 2.5278\n",
            "Epoch 682/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0258 - val_loss: 2.5436\n",
            "Epoch 683/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0243 - val_loss: 2.5375\n",
            "Epoch 684/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0248 - val_loss: 2.5504\n",
            "Epoch 685/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0249 - val_loss: 2.5499\n",
            "Epoch 686/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0265 - val_loss: 2.5488\n",
            "Epoch 687/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0240 - val_loss: 2.5593\n",
            "Epoch 688/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0238 - val_loss: 2.5508\n",
            "Epoch 689/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0280 - val_loss: 2.5502\n",
            "Epoch 690/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0252 - val_loss: 2.5770\n",
            "Epoch 691/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0253 - val_loss: 2.5595\n",
            "Epoch 692/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0243 - val_loss: 2.5834\n",
            "Epoch 693/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0255 - val_loss: 2.5643\n",
            "Epoch 694/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0231 - val_loss: 2.5594\n",
            "Epoch 695/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0258 - val_loss: 2.5742\n",
            "Epoch 696/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0271 - val_loss: 2.5493\n",
            "Epoch 697/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0249 - val_loss: 2.5454\n",
            "Epoch 698/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0218 - val_loss: 2.5605\n",
            "Epoch 699/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0249 - val_loss: 2.5677\n",
            "Epoch 700/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0231 - val_loss: 2.5710\n",
            "Epoch 701/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0261 - val_loss: 2.5590\n",
            "Epoch 702/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0233 - val_loss: 2.5869\n",
            "Epoch 703/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0248 - val_loss: 2.5631\n",
            "Epoch 704/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0231 - val_loss: 2.5680\n",
            "Epoch 705/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0233 - val_loss: 2.5719\n",
            "Epoch 706/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0269 - val_loss: 2.5609\n",
            "Epoch 707/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0482 - val_loss: 2.5184\n",
            "Epoch 708/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0259 - val_loss: 2.5200\n",
            "Epoch 709/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0238 - val_loss: 2.5362\n",
            "Epoch 710/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0222 - val_loss: 2.5409\n",
            "Epoch 711/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0223 - val_loss: 2.5609\n",
            "Epoch 712/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0217 - val_loss: 2.5427\n",
            "Epoch 713/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0254 - val_loss: 2.5364\n",
            "Epoch 714/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0299 - val_loss: 2.5364\n",
            "Epoch 715/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0258 - val_loss: 2.5252\n",
            "Epoch 716/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0238 - val_loss: 2.5477\n",
            "Epoch 717/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0214 - val_loss: 2.5629\n",
            "Epoch 718/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0211 - val_loss: 2.5594\n",
            "Epoch 719/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0216 - val_loss: 2.5598\n",
            "Epoch 720/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0236 - val_loss: 2.5589\n",
            "Epoch 721/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0235 - val_loss: 2.5671\n",
            "Epoch 722/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0230 - val_loss: 2.5624\n",
            "Epoch 723/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0246 - val_loss: 2.5995\n",
            "Epoch 724/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0234 - val_loss: 2.5585\n",
            "Epoch 725/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0219 - val_loss: 2.5853\n",
            "Epoch 726/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0277 - val_loss: 2.5857\n",
            "Epoch 727/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0256 - val_loss: 2.5068\n",
            "Epoch 728/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0277 - val_loss: 2.5151\n",
            "Epoch 729/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0246 - val_loss: 2.4867\n",
            "Epoch 730/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0259 - val_loss: 2.5023\n",
            "Epoch 731/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0266 - val_loss: 2.5228\n",
            "Epoch 732/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0230 - val_loss: 2.5228\n",
            "Epoch 733/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0222 - val_loss: 2.5381\n",
            "Epoch 734/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0216 - val_loss: 2.5492\n",
            "Epoch 735/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0213 - val_loss: 2.5512\n",
            "Epoch 736/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0228 - val_loss: 2.5398\n",
            "Epoch 737/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0231 - val_loss: 2.5737\n",
            "Epoch 738/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0303 - val_loss: 2.5356\n",
            "Epoch 739/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0593 - val_loss: 2.4670\n",
            "Epoch 740/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0378 - val_loss: 2.4883\n",
            "Epoch 741/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0278 - val_loss: 2.5210\n",
            "Epoch 742/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0252 - val_loss: 2.5006\n",
            "Epoch 743/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0233 - val_loss: 2.5151\n",
            "Epoch 744/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0222 - val_loss: 2.5293\n",
            "Epoch 745/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0220 - val_loss: 2.5458\n",
            "Epoch 746/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0211 - val_loss: 2.5573\n",
            "Epoch 747/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0214 - val_loss: 2.5472\n",
            "Epoch 748/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0259 - val_loss: 2.5328\n",
            "Epoch 749/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0223 - val_loss: 2.5619\n",
            "Epoch 750/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0211 - val_loss: 2.5622\n",
            "Epoch 751/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0202 - val_loss: 2.5663\n",
            "Epoch 752/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0249 - val_loss: 2.5635\n",
            "Epoch 753/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0229 - val_loss: 2.5656\n",
            "Epoch 754/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0216 - val_loss: 2.5527\n",
            "Epoch 755/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0204 - val_loss: 2.5930\n",
            "Epoch 756/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0245 - val_loss: 2.5635\n",
            "Epoch 757/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0208 - val_loss: 2.5743\n",
            "Epoch 758/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0208 - val_loss: 2.5653\n",
            "Epoch 759/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0220 - val_loss: 2.5906\n",
            "Epoch 760/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0201 - val_loss: 2.5847\n",
            "Epoch 761/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0268 - val_loss: 2.5948\n",
            "Epoch 762/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0205 - val_loss: 2.6051\n",
            "Epoch 763/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0202 - val_loss: 2.5750\n",
            "Epoch 764/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0199 - val_loss: 2.5675\n",
            "Epoch 765/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0204 - val_loss: 2.5513\n",
            "Epoch 766/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0229 - val_loss: 2.5657\n",
            "Epoch 767/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0196 - val_loss: 2.5756\n",
            "Epoch 768/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0190 - val_loss: 2.5709\n",
            "Epoch 769/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0240 - val_loss: 2.5884\n",
            "Epoch 770/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0180 - val_loss: 2.6013\n",
            "Epoch 771/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0199 - val_loss: 2.5392\n",
            "Epoch 772/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0202 - val_loss: 2.6000\n",
            "Epoch 773/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0224 - val_loss: 2.6236\n",
            "Epoch 774/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0190 - val_loss: 2.5356\n",
            "Epoch 775/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0190 - val_loss: 2.5719\n",
            "Epoch 776/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0203 - val_loss: 2.5930\n",
            "Epoch 777/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0194 - val_loss: 2.5772\n",
            "Epoch 778/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0253 - val_loss: 2.5698\n",
            "Epoch 779/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0204 - val_loss: 2.5889\n",
            "Epoch 780/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0235 - val_loss: 2.5821\n",
            "Epoch 781/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0206 - val_loss: 2.5752\n",
            "Epoch 782/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0202 - val_loss: 2.5670\n",
            "Epoch 783/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0186 - val_loss: 2.5733\n",
            "Epoch 784/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0262 - val_loss: 2.5731\n",
            "Epoch 785/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0361 - val_loss: 2.5336\n",
            "Epoch 786/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0383 - val_loss: 2.5656\n",
            "Epoch 787/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0299 - val_loss: 2.5495\n",
            "Epoch 788/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0243 - val_loss: 2.5615\n",
            "Epoch 789/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0238 - val_loss: 2.5731\n",
            "Epoch 790/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0217 - val_loss: 2.5720\n",
            "Epoch 791/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0214 - val_loss: 2.5794\n",
            "Epoch 792/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0200 - val_loss: 2.5755\n",
            "Epoch 793/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0204 - val_loss: 2.5970\n",
            "Epoch 794/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0207 - val_loss: 2.5984\n",
            "Epoch 795/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0207 - val_loss: 2.5762\n",
            "Epoch 796/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0200 - val_loss: 2.5831\n",
            "Epoch 797/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0217 - val_loss: 2.6123\n",
            "Epoch 798/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0222 - val_loss: 2.5891\n",
            "Epoch 799/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0182 - val_loss: 2.6193\n",
            "Epoch 800/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0197 - val_loss: 2.5985\n",
            "Epoch 801/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0206 - val_loss: 2.5884\n",
            "Epoch 802/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0180 - val_loss: 2.6029\n",
            "Epoch 803/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0190 - val_loss: 2.5759\n",
            "Epoch 804/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0181 - val_loss: 2.6142\n",
            "Epoch 805/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0203 - val_loss: 2.6098\n",
            "Epoch 806/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0192 - val_loss: 2.6084\n",
            "Epoch 807/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0181 - val_loss: 2.6117\n",
            "Epoch 808/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0201 - val_loss: 2.6448\n",
            "Epoch 809/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0184 - val_loss: 2.6550\n",
            "Epoch 810/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0163 - val_loss: 2.6228\n",
            "Epoch 811/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0211 - val_loss: 2.5982\n",
            "Epoch 812/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0197 - val_loss: 2.6430\n",
            "Epoch 813/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0188 - val_loss: 2.5945\n",
            "Epoch 814/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0159 - val_loss: 2.6171\n",
            "Epoch 815/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0186 - val_loss: 2.6018\n",
            "Epoch 816/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0197 - val_loss: 2.6060\n",
            "Epoch 817/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0176 - val_loss: 2.6013\n",
            "Epoch 818/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0169 - val_loss: 2.6108\n",
            "Epoch 819/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0171 - val_loss: 2.6070\n",
            "Epoch 820/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0168 - val_loss: 2.6285\n",
            "Epoch 821/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0182 - val_loss: 2.6107\n",
            "Epoch 822/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0176 - val_loss: 2.6254\n",
            "Epoch 823/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0193 - val_loss: 2.6149\n",
            "Epoch 824/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0150 - val_loss: 2.6145\n",
            "Epoch 825/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0180 - val_loss: 2.6244\n",
            "Epoch 826/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0175 - val_loss: 2.6432\n",
            "Epoch 827/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0177 - val_loss: 2.6300\n",
            "Epoch 828/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0159 - val_loss: 2.6423\n",
            "Epoch 829/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0175 - val_loss: 2.6105\n",
            "Epoch 830/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0158 - val_loss: 2.6245\n",
            "Epoch 831/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0168 - val_loss: 2.6130\n",
            "Epoch 832/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0183 - val_loss: 2.6458\n",
            "Epoch 833/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0177 - val_loss: 2.6388\n",
            "Epoch 834/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0181 - val_loss: 2.6446\n",
            "Epoch 835/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0151 - val_loss: 2.5905\n",
            "Epoch 836/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0152 - val_loss: 2.5774\n",
            "Epoch 837/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0170 - val_loss: 2.6309\n",
            "Epoch 838/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0165 - val_loss: 2.6206\n",
            "Epoch 839/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0158 - val_loss: 2.5959\n",
            "Epoch 840/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0165 - val_loss: 2.5897\n",
            "Epoch 841/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0166 - val_loss: 2.6164\n",
            "Epoch 842/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0139 - val_loss: 2.5824\n",
            "Epoch 843/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0185 - val_loss: 2.6263\n",
            "Epoch 844/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0151 - val_loss: 2.6349\n",
            "Epoch 845/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0168 - val_loss: 2.5905\n",
            "Epoch 846/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0158 - val_loss: 2.6292\n",
            "Epoch 847/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0167 - val_loss: 2.5969\n",
            "Epoch 848/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0150 - val_loss: 2.6241\n",
            "Epoch 849/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0144 - val_loss: 2.6278\n",
            "Epoch 850/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0194 - val_loss: 2.6269\n",
            "Epoch 851/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0153 - val_loss: 2.6171\n",
            "Epoch 852/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0149 - val_loss: 2.5868\n",
            "Epoch 853/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0192 - val_loss: 2.6280\n",
            "Epoch 854/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0157 - val_loss: 2.6374\n",
            "Epoch 855/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0129 - val_loss: 2.5939\n",
            "Epoch 856/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0156 - val_loss: 2.6572\n",
            "Epoch 857/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0146 - val_loss: 2.6385\n",
            "Epoch 858/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0166 - val_loss: 2.6350\n",
            "Epoch 859/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0144 - val_loss: 2.6075\n",
            "Epoch 860/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0160 - val_loss: 2.6206\n",
            "Epoch 861/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0156 - val_loss: 2.6222\n",
            "Epoch 862/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0137 - val_loss: 2.6106\n",
            "Epoch 863/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0136 - val_loss: 2.6204\n",
            "Epoch 864/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0177 - val_loss: 2.6425\n",
            "Epoch 865/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0153 - val_loss: 2.6325\n",
            "Epoch 866/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0155 - val_loss: 2.6229\n",
            "Epoch 867/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0158 - val_loss: 2.6202\n",
            "Epoch 868/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0178 - val_loss: 2.6299\n",
            "Epoch 869/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0137 - val_loss: 2.6449\n",
            "Epoch 870/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0126 - val_loss: 2.6328\n",
            "Epoch 871/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0145 - val_loss: 2.6383\n",
            "Epoch 872/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0144 - val_loss: 2.6343\n",
            "Epoch 873/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0171 - val_loss: 2.6264\n",
            "Epoch 874/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0136 - val_loss: 2.6750\n",
            "Epoch 875/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0150 - val_loss: 2.6592\n",
            "Epoch 876/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0145 - val_loss: 2.6541\n",
            "Epoch 877/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0155 - val_loss: 2.6484\n",
            "Epoch 878/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0132 - val_loss: 2.6567\n",
            "Epoch 879/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0146 - val_loss: 2.6380\n",
            "Epoch 880/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0149 - val_loss: 2.6316\n",
            "Epoch 881/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0141 - val_loss: 2.6693\n",
            "Epoch 882/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0142 - val_loss: 2.6392\n",
            "Epoch 883/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0141 - val_loss: 2.6457\n",
            "Epoch 884/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0146 - val_loss: 2.6506\n",
            "Epoch 885/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0129 - val_loss: 2.6563\n",
            "Epoch 886/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0138 - val_loss: 2.6789\n",
            "Epoch 887/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0144 - val_loss: 2.6537\n",
            "Epoch 888/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0142 - val_loss: 2.6716\n",
            "Epoch 889/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0154 - val_loss: 2.6748\n",
            "Epoch 890/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0135 - val_loss: 2.6720\n",
            "Epoch 891/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0132 - val_loss: 2.6714\n",
            "Epoch 892/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0122 - val_loss: 2.6791\n",
            "Epoch 893/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0163 - val_loss: 2.6510\n",
            "Epoch 894/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0158 - val_loss: 2.6274\n",
            "Epoch 895/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0143 - val_loss: 2.5973\n",
            "Epoch 896/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0125 - val_loss: 2.6220\n",
            "Epoch 897/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0123 - val_loss: 2.6417\n",
            "Epoch 898/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0109 - val_loss: 2.6461\n",
            "Epoch 899/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0156 - val_loss: 2.6215\n",
            "Epoch 900/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0136 - val_loss: 2.6517\n",
            "Epoch 901/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0121 - val_loss: 2.6555\n",
            "Epoch 902/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0131 - val_loss: 2.6264\n",
            "Epoch 903/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0139 - val_loss: 2.6567\n",
            "Epoch 904/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0119 - val_loss: 2.6528\n",
            "Epoch 905/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0115 - val_loss: 2.6682\n",
            "Epoch 906/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0176 - val_loss: 2.6537\n",
            "Epoch 907/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0120 - val_loss: 2.6676\n",
            "Epoch 908/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0115 - val_loss: 2.6629\n",
            "Epoch 909/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0164 - val_loss: 2.6647\n",
            "Epoch 910/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0118 - val_loss: 2.6763\n",
            "Epoch 911/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0128 - val_loss: 2.6867\n",
            "Epoch 912/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0140 - val_loss: 2.6649\n",
            "Epoch 913/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0113 - val_loss: 2.7204\n",
            "Epoch 914/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0117 - val_loss: 2.6851\n",
            "Epoch 915/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0142 - val_loss: 2.6648\n",
            "Epoch 916/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0118 - val_loss: 2.6867\n",
            "Epoch 917/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0133 - val_loss: 2.6602\n",
            "Epoch 918/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0138 - val_loss: 2.6979\n",
            "Epoch 919/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0111 - val_loss: 2.7017\n",
            "Epoch 920/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0129 - val_loss: 2.7123\n",
            "Epoch 921/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0131 - val_loss: 2.6960\n",
            "Epoch 922/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0130 - val_loss: 2.6904\n",
            "Epoch 923/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0117 - val_loss: 2.6790\n",
            "Epoch 924/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0123 - val_loss: 2.6870\n",
            "Epoch 925/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0129 - val_loss: 2.6805\n",
            "Epoch 926/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0129 - val_loss: 2.6938\n",
            "Epoch 927/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0117 - val_loss: 2.6619\n",
            "Epoch 928/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0111 - val_loss: 2.6765\n",
            "Epoch 929/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0119 - val_loss: 2.6992\n",
            "Epoch 930/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0134 - val_loss: 2.6522\n",
            "Epoch 931/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0121 - val_loss: 2.6711\n",
            "Epoch 932/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0122 - val_loss: 2.6611\n",
            "Epoch 933/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0112 - val_loss: 2.6769\n",
            "Epoch 934/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0113 - val_loss: 2.6557\n",
            "Epoch 935/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0110 - val_loss: 2.6743\n",
            "Epoch 936/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0124 - val_loss: 2.6855\n",
            "Epoch 937/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0138 - val_loss: 2.6759\n",
            "Epoch 938/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0107 - val_loss: 2.6845\n",
            "Epoch 939/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0121 - val_loss: 2.6899\n",
            "Epoch 940/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0110 - val_loss: 2.6753\n",
            "Epoch 941/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0114 - val_loss: 2.7133\n",
            "Epoch 942/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0134 - val_loss: 2.6735\n",
            "Epoch 943/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0114 - val_loss: 2.6774\n",
            "Epoch 944/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0100 - val_loss: 2.6923\n",
            "Epoch 945/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0115 - val_loss: 2.6997\n",
            "Epoch 946/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0119 - val_loss: 2.6718\n",
            "Epoch 947/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0108 - val_loss: 2.7036\n",
            "Epoch 948/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0096 - val_loss: 2.6972\n",
            "Epoch 949/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0138 - val_loss: 2.6971\n",
            "Epoch 950/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0124 - val_loss: 2.7054\n",
            "Epoch 951/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0124 - val_loss: 2.6968\n",
            "Epoch 952/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0117 - val_loss: 2.6637\n",
            "Epoch 953/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0097 - val_loss: 2.6822\n",
            "Epoch 954/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0110 - val_loss: 2.6881\n",
            "Epoch 955/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0104 - val_loss: 2.6676\n",
            "Epoch 956/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0102 - val_loss: 2.6774\n",
            "Epoch 957/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0108 - val_loss: 2.6891\n",
            "Epoch 958/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0121 - val_loss: 2.6616\n",
            "Epoch 959/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0150 - val_loss: 2.6452\n",
            "Epoch 960/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0110 - val_loss: 2.6480\n",
            "Epoch 961/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0093 - val_loss: 2.6614\n",
            "Epoch 962/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0121 - val_loss: 2.6470\n",
            "Epoch 963/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0093 - val_loss: 2.6583\n",
            "Epoch 964/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0098 - val_loss: 2.6720\n",
            "Epoch 965/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0105 - val_loss: 2.6698\n",
            "Epoch 966/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0105 - val_loss: 2.6805\n",
            "Epoch 967/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0123 - val_loss: 2.6856\n",
            "Epoch 968/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0109 - val_loss: 2.6938\n",
            "Epoch 969/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0099 - val_loss: 2.7026\n",
            "Epoch 970/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0124 - val_loss: 2.6863\n",
            "Epoch 971/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0114 - val_loss: 2.6686\n",
            "Epoch 972/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0117 - val_loss: 2.6941\n",
            "Epoch 973/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0101 - val_loss: 2.6911\n",
            "Epoch 974/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0123 - val_loss: 2.6920\n",
            "Epoch 975/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0112 - val_loss: 2.6955\n",
            "Epoch 976/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0093 - val_loss: 2.7122\n",
            "Epoch 977/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0104 - val_loss: 2.6950\n",
            "Epoch 978/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0101 - val_loss: 2.7023\n",
            "Epoch 979/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0112 - val_loss: 2.6989\n",
            "Epoch 980/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0097 - val_loss: 2.7077\n",
            "Epoch 981/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0130 - val_loss: 2.6844\n",
            "Epoch 982/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0107 - val_loss: 2.6968\n",
            "Epoch 983/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0117 - val_loss: 2.6805\n",
            "Epoch 984/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0105 - val_loss: 2.6949\n",
            "Epoch 985/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0114 - val_loss: 2.6801\n",
            "Epoch 986/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0097 - val_loss: 2.7057\n",
            "Epoch 987/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0080 - val_loss: 2.7043\n",
            "Epoch 988/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0098 - val_loss: 2.7101\n",
            "Epoch 989/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0121 - val_loss: 2.7180\n",
            "Epoch 990/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0113 - val_loss: 2.7244\n",
            "Epoch 991/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0086 - val_loss: 2.7051\n",
            "Epoch 992/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0121 - val_loss: 2.7261\n",
            "Epoch 993/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0102 - val_loss: 2.7144\n",
            "Epoch 994/1000\n",
            "484/484 [==============================] - 4s 8ms/sample - loss: 0.0091 - val_loss: 2.7100\n",
            "Epoch 995/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0097 - val_loss: 2.6988\n",
            "Epoch 996/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0125 - val_loss: 2.7192\n",
            "Epoch 997/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0102 - val_loss: 2.6895\n",
            "Epoch 998/1000\n",
            "484/484 [==============================] - 3s 7ms/sample - loss: 0.0102 - val_loss: 2.6812\n",
            "Epoch 999/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0102 - val_loss: 2.6645\n",
            "Epoch 1000/1000\n",
            "484/484 [==============================] - 4s 7ms/sample - loss: 0.0090 - val_loss: 2.6643\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "j9Uy6wzk8pQi",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "Next step: inference mode (sampling):\n",
        "+ encode input and retrieve initial decoder state\n",
        "+ run one step of decoder with this initial state and a \"*start of sequence*\" token as target. Output will be the next target token.\n",
        "+ Repeat with the current target token and current states"
      ]
    },
    {
      "metadata": {
        "id": "0bJVGZFq8N4L",
        "colab_type": "code",
        "outputId": "021073c3-48c4-4c9a-bb86-bcd889f03eb8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 6874
        }
      },
      "cell_type": "code",
      "source": [
        "# Define sampling models\n",
        "encoder_model = Model(encoder_inputs, encoder_states)\n",
        "\n",
        "decoder_state_input_h = Input(shape=(latent_dim,))\n",
        "decoder_state_input_c = Input(shape=(latent_dim,))\n",
        "decoder_states_inputs = [decoder_state_input_h, decoder_state_input_c]\n",
        "decoder_outputs, state_h, state_c = decoder_lstm(\n",
        "    decoder_inputs, initial_state=decoder_states_inputs)\n",
        "decoder_states = [state_h, state_c]\n",
        "decoder_outputs = decoder_dense(decoder_outputs)\n",
        "decoder_model = Model(\n",
        "    [decoder_inputs] + decoder_states_inputs,\n",
        "    [decoder_outputs] + decoder_states)\n",
        "\n",
        "# Reverse-lookup token index to decode sequences back to\n",
        "# something readable.\n",
        "reverse_input_char_index = dict(\n",
        "    (i, char) for char, i in input_token_index.items())\n",
        "reverse_target_char_index = dict(\n",
        "    (i, char) for char, i in target_token_index.items())\n",
        "\n",
        "\n",
        "def decode_sequence(input_seq):\n",
        "    # Encode the input as state vectors.\n",
        "    states_value = encoder_model.predict(input_seq)\n",
        "\n",
        "    # Generate empty target sequence of length 1.\n",
        "    target_seq = np.zeros((1, 1, num_decoder_tokens))\n",
        "    # Populate the first character of target sequence with the start character.\n",
        "    target_seq[0, 0, target_token_index['\\t']] = 1.\n",
        "\n",
        "    # Sampling loop for a batch of sequences\n",
        "    # (to simplify, here we assume a batch of size 1).\n",
        "    stop_condition = False\n",
        "    decoded_sentence = ''\n",
        "    while not stop_condition:\n",
        "        output_tokens, h, c = decoder_model.predict(\n",
        "            [target_seq] + states_value)\n",
        "\n",
        "        # Sample a token\n",
        "        sampled_token_index = np.argmax(output_tokens[0, -1, :])\n",
        "        sampled_char = reverse_target_char_index[sampled_token_index]\n",
        "        decoded_sentence += sampled_char\n",
        "\n",
        "        # Exit condition: either hit max length\n",
        "        # or find stop character.\n",
        "        if (sampled_char == '\\n' or\n",
        "           len(decoded_sentence) > max_decoder_seq_length):\n",
        "            stop_condition = True\n",
        "\n",
        "        # Update the target sequence (of length 1).\n",
        "        target_seq = np.zeros((1, 1, num_decoder_tokens))\n",
        "        target_seq[0, 0, sampled_token_index] = 1.\n",
        "\n",
        "        # Update states\n",
        "        states_value = [h, c]\n",
        "\n",
        "    return decoded_sentence\n",
        "\n",
        "\n",
        "for seq_index in range(100):\n",
        "    # Take one sequence (part of the training set)\n",
        "    # for trying out decoding.\n",
        "    input_seq = encoder_input_data[seq_index: seq_index + 1]\n",
        "    decoded_sentence = decode_sequence(input_seq)\n",
        "    print('-')\n",
        "    print('Input sentence:', input_texts[seq_index])\n",
        "    print('Decoded sentence:', decoded_sentence)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "-\n",
            "Input sentence: Wow!\n",
            "Decoded sentence: Carai!\n",
            "\n",
            "-\n",
            "Input sentence: Really?\n",
            "Decoded sentence: De veritat?\n",
            "\n",
            "-\n",
            "Input sentence: Thanks.\n",
            "Decoded sentence: Mercès per tot.\n",
            "\n",
            "-\n",
            "Input sentence: Goodbye!\n",
            "Decoded sentence: Adéu!\n",
            "\n",
            "-\n",
            "Input sentence: Hurry up.\n",
            "Decoded sentence: Afanya't.\n",
            "\n",
            "-\n",
            "Input sentence: Too late.\n",
            "Decoded sentence: Massa tard.\n",
            "\n",
            "-\n",
            "Input sentence: Thank you.\n",
            "Decoded sentence: Gràcies!\n",
            "\n",
            "-\n",
            "Input sentence: Can I help?\n",
            "Decoded sentence: Puc ajudar?\n",
            "\n",
            "-\n",
            "Input sentence: I envy him.\n",
            "Decoded sentence: L'envejo.\n",
            "\n",
            "-\n",
            "Input sentence: Time flies.\n",
            "Decoded sentence: El temps vola.\n",
            "\n",
            "-\n",
            "Input sentence: I'm 17, too.\n",
            "Decoded sentence: Jo també tinc 17 anys.\n",
            "\n",
            "-\n",
            "Input sentence: I'm at home.\n",
            "Decoded sentence: Estic a casa.\n",
            "\n",
            "-\n",
            "Input sentence: Make a wish.\n",
            "Decoded sentence: Demana un desig\n",
            "\n",
            "-\n",
            "Input sentence: Money talks.\n",
            "Decoded sentence: Qui paga, mana.\n",
            "\n",
            "-\n",
            "Input sentence: We love you.\n",
            "Decoded sentence: T'estimem.\n",
            "\n",
            "-\n",
            "Input sentence: We love you.\n",
            "Decoded sentence: T'estimem.\n",
            "\n",
            "-\n",
            "Input sentence: Who are you?\n",
            "Decoded sentence: Qui sou?\n",
            "\n",
            "-\n",
            "Input sentence: Who are you?\n",
            "Decoded sentence: Qui sou?\n",
            "\n",
            "-\n",
            "Input sentence: Who are you?\n",
            "Decoded sentence: Qui sou?\n",
            "\n",
            "-\n",
            "Input sentence: Who are you?\n",
            "Decoded sentence: Qui sou?\n",
            "\n",
            "-\n",
            "Input sentence: He has a dog.\n",
            "Decoded sentence: Ell té un gos.\n",
            "\n",
            "-\n",
            "Input sentence: She stood up.\n",
            "Decoded sentence: Ella es va aixecar.\n",
            "\n",
            "-\n",
            "Input sentence: Hi, everybody.\n",
            "Decoded sentence: Hola a tots.\n",
            "\n",
            "-\n",
            "Input sentence: I'm desperate.\n",
            "Decoded sentence: Estic desesperat.\n",
            "\n",
            "-\n",
            "Input sentence: Let me try it.\n",
            "Decoded sentence: Deixa'm intentar-ho.\n",
            "\n",
            "-\n",
            "Input sentence: You look good.\n",
            "Decoded sentence: Tens bona cara.\n",
            "\n",
            "-\n",
            "Input sentence: You look good.\n",
            "Decoded sentence: Tens bona cara.\n",
            "\n",
            "-\n",
            "Input sentence: You look good.\n",
            "Decoded sentence: Tens bona cara.\n",
            "\n",
            "-\n",
            "Input sentence: Are you insane?\n",
            "Decoded sentence: Estàs boig?\n",
            "\n",
            "-\n",
            "Input sentence: Can I help you?\n",
            "Decoded sentence: Puc ajudar?\n",
            "\n",
            "-\n",
            "Input sentence: Happy New Year!\n",
            "Decoded sentence: Bon any nou!\n",
            "\n",
            "-\n",
            "Input sentence: I need a stamp.\n",
            "Decoded sentence: Necessito un segell.\n",
            "\n",
            "-\n",
            "Input sentence: I saw him jump.\n",
            "Decoded sentence: El vaig veure estripar la carta.\n",
            "\n",
            "-\n",
            "Input sentence: Leave me alone!\n",
            "Decoded sentence: Deixa'm en pau!\n",
            "\n",
            "-\n",
            "Input sentence: Who painted it?\n",
            "Decoded sentence: Qui ho ha pintat?\n",
            "\n",
            "-\n",
            "Input sentence: Her book is red.\n",
            "Decoded sentence: El seu llibre és roig.\n",
            "\n",
            "-\n",
            "Input sentence: I didn't say it.\n",
            "Decoded sentence: No ho he dit pas.\n",
            "\n",
            "-\n",
            "Input sentence: I felt the same.\n",
            "Decoded sentence: Em sentia igual.\n",
            "\n",
            "-\n",
            "Input sentence: I have two cats.\n",
            "Decoded sentence: Tinc dues filles.\n",
            "\n",
            "-\n",
            "Input sentence: I speak Swedish.\n",
            "Decoded sentence: Hi havia deu ous en total.\n",
            "\n",
            "-\n",
            "Input sentence: It's cold today!\n",
            "Decoded sentence: Avui fa fred!\n",
            "\n",
            "-\n",
            "Input sentence: It's your fault.\n",
            "Decoded sentence: És culpa teva.\n",
            "\n",
            "-\n",
            "Input sentence: Who are you all?\n",
            "Decoded sentence: Qui sou?\n",
            "\n",
            "-\n",
            "Input sentence: Who are you all?\n",
            "Decoded sentence: Qui sou?\n",
            "\n",
            "-\n",
            "Input sentence: Here is your bag.\n",
            "Decoded sentence: Ací està la teua bossa.\n",
            "\n",
            "-\n",
            "Input sentence: Here is your bag.\n",
            "Decoded sentence: Ací està la teua bossa.\n",
            "\n",
            "-\n",
            "Input sentence: Here is your bag.\n",
            "Decoded sentence: Ací està la teua bossa.\n",
            "\n",
            "-\n",
            "Input sentence: I am now on duty.\n",
            "Decoded sentence: Crec que l'elecció està clara.\n",
            "\n",
            "-\n",
            "Input sentence: I ate the cheese.\n",
            "Decoded sentence: Em vaig menjar el formatge.\n",
            "\n",
            "-\n",
            "Input sentence: I have a problem.\n",
            "Decoded sentence: Tinc un problema.\n",
            "\n",
            "-\n",
            "Input sentence: I have a problem.\n",
            "Decoded sentence: Tinc un problema.\n",
            "\n",
            "-\n",
            "Input sentence: I have no family.\n",
            "Decoded sentence: Cada dia menjo galetes per esmorzar.\n",
            "\n",
            "-\n",
            "Input sentence: I work in a bank.\n",
            "Decoded sentence: Jo treballo a un banc.\n",
            "\n",
            "-\n",
            "Input sentence: I wrote a letter.\n",
            "Decoded sentence: Vaig escriure una carta.\n",
            "\n",
            "-\n",
            "Input sentence: I'm already late.\n",
            "Decoded sentence: Ja faig tard.\n",
            "\n",
            "-\n",
            "Input sentence: I'm not a doctor.\n",
            "Decoded sentence: Jo no sóc metge.\n",
            "\n",
            "-\n",
            "Input sentence: Let go of my arm.\n",
            "Decoded sentence: Deixa'm anar el braç.\n",
            "\n",
            "-\n",
            "Input sentence: She lives nearby.\n",
            "Decoded sentence: Viu aquí prop.\n",
            "\n",
            "-\n",
            "Input sentence: They're my books.\n",
            "Decoded sentence: Són els meus llibres.\n",
            "\n",
            "-\n",
            "Input sentence: This is your dog.\n",
            "Decoded sentence: Aquest és el teu gos.\n",
            "\n",
            "-\n",
            "Input sentence: Tom isn't hungry.\n",
            "Decoded sentence: Tom és massa jove per portar un cotxe.\n",
            "\n",
            "-\n",
            "Input sentence: Tom walked alone.\n",
            "Decoded sentence: En Tom caminava sol.\n",
            "\n",
            "-\n",
            "Input sentence: What is going on?\n",
            "Decoded sentence: Què hi ha?\n",
            "\n",
            "-\n",
            "Input sentence: What is going on?\n",
            "Decoded sentence: Què hi ha?\n",
            "\n",
            "-\n",
            "Input sentence: Who are you with?\n",
            "Decoded sentence: Qui sou?\n",
            "\n",
            "-\n",
            "Input sentence: Who are you with?\n",
            "Decoded sentence: Qui sou?\n",
            "\n",
            "-\n",
            "Input sentence: Answer in English.\n",
            "Decoded sentence: Contesta en anglès!\n",
            "\n",
            "-\n",
            "Input sentence: He went back home.\n",
            "Decoded sentence: Ell va tornar a casa.\n",
            "\n",
            "-\n",
            "Input sentence: I have an earache.\n",
            "Decoded sentence: Tinc otitis.\n",
            "\n",
            "-\n",
            "Input sentence: I have black eyes.\n",
            "Decoded sentence: Tinc els ulls negres.\n",
            "\n",
            "-\n",
            "Input sentence: I think he did it.\n",
            "Decoded sentence: Crec que ho va fer ell.\n",
            "\n",
            "-\n",
            "Input sentence: I'm a salesperson.\n",
            "Decoded sentence: Sóc venedor.\n",
            "\n",
            "-\n",
            "Input sentence: Let me have a try.\n",
            "Decoded sentence: Deixa'm intentar-ho.\n",
            "\n",
            "-\n",
            "Input sentence: Nobody is perfect.\n",
            "Decoded sentence: Ningú és perfecte.\n",
            "\n",
            "-\n",
            "Input sentence: She has long hair.\n",
            "Decoded sentence: Ella té un gat blanc.\n",
            "\n",
            "-\n",
            "Input sentence: Tom doesn't drink.\n",
            "Decoded sentence: Tom ja no va a l'escola.\n",
            "\n",
            "-\n",
            "Input sentence: Why is snow white?\n",
            "Decoded sentence: Per què és blanca la neu?\n",
            "\n",
            "-\n",
            "Input sentence: You need to hurry.\n",
            "Decoded sentence: T'has d'afanyar.\n",
            "\n",
            "-\n",
            "Input sentence: Are you hungry now?\n",
            "Decoded sentence: Teniu fam, ara?\n",
            "\n",
            "-\n",
            "Input sentence: Are you hungry now?\n",
            "Decoded sentence: Teniu fam, ara?\n",
            "\n",
            "-\n",
            "Input sentence: Are you hungry now?\n",
            "Decoded sentence: Teniu fam, ara?\n",
            "\n",
            "-\n",
            "Input sentence: Come along with me.\n",
            "Decoded sentence: Acompanya'm.\n",
            "\n",
            "-\n",
            "Input sentence: Do it by all means.\n",
            "Decoded sentence: Fes-ho com sigui.\n",
            "\n",
            "-\n",
            "Input sentence: Do it by all means.\n",
            "Decoded sentence: Fes-ho com sigui.\n",
            "\n",
            "-\n",
            "Input sentence: Everyone loves him.\n",
            "Decoded sentence: Tots l'estimen.\n",
            "\n",
            "-\n",
            "Input sentence: He speaks fluently.\n",
            "Decoded sentence: Ell parla cinc idiomes.\n",
            "\n",
            "-\n",
            "Input sentence: I don't have a cat.\n",
            "Decoded sentence: Crec que no podem córrer eixe risc.\n",
            "\n",
            "-\n",
            "Input sentence: I don't want sugar.\n",
            "Decoded sentence: No vull sucre.\n",
            "\n",
            "-\n",
            "Input sentence: I have lost my key.\n",
            "Decoded sentence: He perdut la meva clau.\n",
            "\n",
            "-\n",
            "Input sentence: It was a nightmare.\n",
            "Decoded sentence: Va ser un malson.\n",
            "\n",
            "-\n",
            "Input sentence: Shut up and listen!\n",
            "Decoded sentence: Calla i escolta!\n",
            "\n",
            "-\n",
            "Input sentence: Shut up and listen.\n",
            "Decoded sentence: Calla i escolta!\n",
            "\n",
            "-\n",
            "Input sentence: Stay out of my way.\n",
            "Decoded sentence: Fora del meu camí!\n",
            "\n",
            "-\n",
            "Input sentence: This coat fits you.\n",
            "Decoded sentence: Aquest abric et queda bé.\n",
            "\n",
            "-\n",
            "Input sentence: Tom is a silly man.\n",
            "Decoded sentence: En Tom és un professor d'història.\n",
            "\n",
            "-\n",
            "Input sentence: What's the problem?\n",
            "Decoded sentence: Quin problema hi ha?\n",
            "\n",
            "-\n",
            "Input sentence: What's your secret?\n",
            "Decoded sentence: Quin és el teu secret?\n",
            "\n",
            "-\n",
            "Input sentence: You are a good boy.\n",
            "Decoded sentence: Ets un bon noi.\n",
            "\n",
            "-\n",
            "Input sentence: Europe is in crisis.\n",
            "Decoded sentence: Europa està en crisi.\n",
            "\n",
            "-\n",
            "Input sentence: Everybody loves him.\n",
            "Decoded sentence: Tots l'estimen.\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "LOb_TKFeB8wu",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Exercise\n",
        "\n",
        "Define a `tf` recurrent model for classifying MNIST digits when these digits are represented as sequence of image columns."
      ]
    },
    {
      "metadata": {
        "id": "YrrAWJgvI_LF",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Bidirectional LSTM\n",
        "\n",
        "So  far,  we  have  focused  on  RNNs  that  look  into  the  past    to predict future values in the sequence, but not to  to make predictions based on future values by reading throught the sequence backwards. Why?\n",
        "\n",
        "**Bi-directional  deep  neural networs**,  at each time-step, $t$,  maintain two hidden layers, one for the left-to-right propagation and another for the right-to-left  propagation (hence, consuming twice  as  much  memory  space).   \n",
        "\n",
        "The final classification result, $\\hat{y}$, is generated through combining the score results produced by both RNN hidden layers.\n",
        "\n",
        "<img src = \"./images/t9.png\"  width = \"600\">"
      ]
    },
    {
      "metadata": {
        "id": "-4nHl7n5I_LG",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "The equations are (arrows are for designing left-to-right and right-to-left tensors):\n",
        "\n",
        "$$\n",
        "\t\\overrightarrow{h}_t = f(\\overrightarrow{W} x_t + \\overrightarrow{V} \\overrightarrow{h}_{t-1} + \\overrightarrow{b})\n",
        "$$\n",
        "$$\n",
        "\t\\overleftarrow{h}_t = f(\\overleftarrow{W} x_t + \\overleftarrow{V} \\overleftarrow{h}_{t+1} + \\overleftarrow{b})\n",
        "$$\n",
        "$$\n",
        "\t\\hat{y}_t = g(U h_t + c) = g(U [\\overrightarrow{h}_t; \\overleftarrow{h}_t] + c)\n",
        "$$\n",
        "\n",
        "$[\\overrightarrow{h}_t; \\overleftarrow{h}_t]$ summarizes the past and future of a single element of the sequence.\n",
        "\n",
        "\n",
        "Bidirectional RNNs can be stacked as usual!"
      ]
    },
    {
      "metadata": {
        "id": "jZeEUPn7AZC3",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "### New `keras` layer \n",
        "\n",
        "We will use an `embedding` layer.\n",
        "\n",
        "This layer turns positive integers (indexes) into dense vectors of fixed size (to be learned during training). eg.` [[4], [20]] -> [[0.25, 0.1], [0.6, -0.2]]`\n",
        "\n",
        "This layer can only be used as the first layer in a model.\n",
        "\n",
        "```\n",
        "keras.layers.Embedding(input_dim, output_dim, embeddings_initializer='uniform', embeddings_regularizer=None, activity_regularizer=None, embeddings_constraint=None, mask_zero=False, input_length=None)```"
      ]
    },
    {
      "metadata": {
        "id": "Hh-P5fUPEvq0",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Example\n",
        "\n",
        "The task is IMDB Movie reviews sentiment classification.\n",
        "\n",
        "IMDB dataset contains 25,000 movies reviews from IMDB, labeled by sentiment (positive/negative). Reviews have been preprocessed, and each review is encoded as a sequence of word indexes (integers). For convenience, words are indexed by overall frequency in the dataset, so that for instance the integer `3` encodes the 3rd most frequent word in the data. \n",
        "\n",
        "As a convention, `0` does not stand for a specific word, but instead is used to encode any unknown word.\n",
        "\n",
        "`num_words` parameter denotes the number of most frequent words to consider. Any less frequent word will appear as `0` \n",
        "value in the sequence data."
      ]
    },
    {
      "metadata": {
        "id": "ZDKLg5Pk_iDL",
        "colab_type": "code",
        "outputId": "07d16702-d301-45fb-a1d9-930df2fa2738",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 466
        }
      },
      "cell_type": "code",
      "source": [
        "import numpy as np\n",
        "\n",
        "from tensorflow.keras.preprocessing import sequence\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.layers import Dense, Dropout, Embedding, LSTM, Bidirectional\n",
        "from tensorflow.keras.datasets import imdb\n",
        "\n",
        "max_features = 20000\n",
        "# cut texts after this number of words\n",
        "# (among top max_features most common words)\n",
        "maxlen = 100\n",
        "batch_size = 32\n",
        "\n",
        "print('Loading data...')\n",
        "(x_train, y_train), (x_test, y_test) = imdb.load_data(num_words=max_features)\n",
        "print(len(x_train), 'train sequences')\n",
        "print(len(x_test), 'test sequences')\n",
        "\n",
        "print('Pad sequences (samples x time)')\n",
        "x_train = sequence.pad_sequences(x_train, maxlen=maxlen)\n",
        "x_test = sequence.pad_sequences(x_test, maxlen=maxlen)\n",
        "print('x_train shape:', x_train.shape)\n",
        "print('x_test shape:', x_test.shape)\n",
        "y_train = np.array(y_train)\n",
        "y_test = np.array(y_test)\n",
        "\n",
        "model = Sequential()\n",
        "model.add(Embedding(max_features, 128, input_length=maxlen))\n",
        "model.add(Bidirectional(LSTM(64)))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(Dense(1, activation='sigmoid'))\n",
        "\n",
        "# try using different optimizers and different optimizer configs\n",
        "model.compile('adam', 'binary_crossentropy', metrics=['accuracy'])\n",
        "\n",
        "print('Train...')\n",
        "model.fit(x_train, y_train,\n",
        "          batch_size=batch_size,\n",
        "          epochs=4,\n",
        "          validation_data=[x_test, y_test])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Loading data...\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/imdb.npz\n",
            "17465344/17464789 [==============================] - 0s 0us/step\n",
            "25000 train sequences\n",
            "25000 test sequences\n",
            "Pad sequences (samples x time)\n",
            "x_train shape: (25000, 100)\n",
            "x_test shape: (25000, 100)\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/layers/core.py:143: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
            "Train...\n",
            "Train on 25000 samples, validate on 25000 samples\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.cast instead.\n",
            "Epoch 1/4\n",
            "25000/25000 [==============================] - 682s 27ms/sample - loss: 0.4088 - acc: 0.8111 - val_loss: 0.4375 - val_acc: 0.8019\n",
            "Epoch 2/4\n",
            "25000/25000 [==============================] - 681s 27ms/sample - loss: 0.2225 - acc: 0.9155 - val_loss: 0.3700 - val_acc: 0.8406\n",
            "Epoch 3/4\n",
            "25000/25000 [==============================] - 686s 27ms/sample - loss: 0.1272 - acc: 0.9551 - val_loss: 0.4827 - val_acc: 0.8412\n",
            "Epoch 4/4\n",
            "25000/25000 [==============================] - 680s 27ms/sample - loss: 0.0715 - acc: 0.9762 - val_loss: 0.5749 - val_acc: 0.8366\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f3403c61080>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "metadata": {
        "id": "ZCNa8f38I_LR",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Name generation with LSTM\n",
        "\n",
        "We are going to train RNN \"character-level\" language models. \n",
        "\n",
        "That is, we’ll give the RNN a huge chunk of text and ask it to model the probability distribution of the next character in the sequence given a sequence of previous characters. This will then allow us to generate new text one character at a time.\n",
        "\n",
        "We will encode each character into a vector using ``1-of-k`` encoding (i.e. all zero except for a single one at the index of the character in the vocabulary), and feed them into the RNN one at a time. \n",
        "\n",
        "At test time, we will feed a sequence of characters into the RNN and get a distribution over what characters are likely to come next. We sample from this distribution, and feed it right back in to get the next letter. Repeat this process and you’re sampling text!\n",
        "\n",
        "We can also play with the temperature of the Softmax during sampling. Decreasing the temperature from 1 to some lower number (e.g. 0.5) makes the RNN more confident, but also more conservative in its samples. Conversely, higher temperatures will give more diversity but at cost of more mistakes.\n",
        "\n",
        "> Remider: **Softmax**\n",
        ">$$\n",
        "\tP(y = j | \\mathbf{x}) = \\frac{\\exp(\\mathbf{x}^T \\mathbf{w}_j)/\\tau}{\\sum_{k=1}^K \\exp(\\mathbf{x}^T \\mathbf{w}_k)/\\tau}\n",
        "$$\n",
        "> $\\tau$ is called *temperature*. For high temperatures ( $ \\tau \\to \\infty $ ), all $y$ have nearly the same probability. For a low temperature ( $ \\tau \\to 0^{+}$), the probability of the most probable $y$ tends to 1."
      ]
    },
    {
      "metadata": {
        "id": "oK1YXgquI_LT",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "In order to process sequences of symbols with RNN we need to represent these symbols by numbers.\n",
        "\n",
        "Let's suppose we have $|V|$ different symbols. The most simple representation is the **one-hot vector**: Represent every word as an $\\mathbb{R}^{|V|\\times1}$ vector with all $0$s and one $1$ at the index of that word in the sorted english language. Symbol vectors in this type of encoding would appear as the following:\n",
        "\n",
        "$$w^{s_1} = \\left[ \\begin{array}{c} 1 \\\\ 0 \\\\ 0 \\\\ \\vdots \\\\ 0 \\end{array} \\right], w^{s_2} = \\left[ \\begin{array}{c} 0 \\\\ 1 \\\\ 0 \\\\ \\vdots \\\\ 0 \\end{array} \\right], w^{s_3} = \\left[ \\begin{array}{c} 0 \\\\ 0 \\\\ 1 \\\\ \\vdots \\\\ 0 \\end{array} \\right], \\cdots \n",
        "w^{s_{|V|}} = \\left[ \\begin{array}{c} 0 \\\\ 0 \\\\ 0 \\\\ \\vdots \\\\ 1 \\end{array} \\right] $$\n",
        "\n",
        "We represent each symbol as a completely independent entity. This symbol representation does not give us directly any notion of similarity."
      ]
    },
    {
      "metadata": {
        "id": "3TyVSdivI_LU",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "To train our model we need text to learn from a large dataset of names. Fortunately we don’t need any labels to train a language model, just raw text.\n",
        "\n",
        "Places names: You can download 52,700 Catalan names from a dataset available on http://territori.gencat.cat/ca/01_departament/11_normativa_i_documentacio/03_documentacio/02_territori_i_mobilitat/cartografia/nomenclator_oficial_de_toponimia_de_catalunya/"
      ]
    },
    {
      "metadata": {
        "id": "B2NhClt8I_LV",
        "colab_type": "code",
        "outputId": "edab2bd0-a4e9-4d1a-8fe0-31a23de1449c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 360
        }
      },
      "cell_type": "code",
      "source": [
        "from __future__ import print_function\n",
        "from tensorflow.keras.models import Sequential\n",
        "from tensorflow.keras.layers import Dense, Dropout, Activation\n",
        "from tensorflow.keras.layers import LSTM\n",
        "from tensorflow.keras.optimizers import RMSprop\n",
        "import numpy as np\n",
        "import random\n",
        "import sys\n",
        "import requests\n",
        "\n",
        "data_path = 'https://raw.githubusercontent.com/DataScienceUB/DeepLearningMaster2019/master/data/NombresMujerBarcelona.txt'\n",
        "\n",
        "\n",
        "response = requests.get(data_path)\n",
        "data = response.text\n",
        "text = data.lower()\n",
        "\n",
        "print(text[0:100],'\\n')\n",
        "\n",
        "# text = text.replace(\"\\n\", \" \")\n",
        "    \n",
        "print('corpus length:', len(text))\n",
        "\n",
        "chars = sorted(list(set(text)))\n",
        "print('total chars:', len(chars))\n",
        "char_indices = dict((c, i) for i, c in enumerate(chars))\n",
        "indices_char = dict((i, c) for i, c in enumerate(chars))\n",
        "\n",
        "# cut the text in semi-redundant sequences of maxlen characters\n",
        "maxlen = 20\n",
        "step = 3\n",
        "sentences = []\n",
        "next_chars = []\n",
        "for i in range(0, len(text) - maxlen, step):\n",
        "    sentences.append(text[i: i + maxlen])\n",
        "    next_chars.append(text[i + maxlen])\n",
        "print('nb sequences:', len(sentences))\n",
        "\n",
        "print('Vectorization...')\n",
        "X = np.zeros((len(sentences), maxlen, len(chars)), dtype=np.bool)\n",
        "y = np.zeros((len(sentences), len(chars)), dtype=np.bool)\n",
        "for i, sentence in enumerate(sentences):\n",
        "    for t, char in enumerate(sentence):\n",
        "        X[i, t, char_indices[char]] = 1\n",
        "    y[i, char_indices[next_chars[i]]] = 1"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "abelina\n",
            "abida\n",
            "abigail\n",
            "abilia\n",
            "abir\n",
            "abrar\n",
            "abril\n",
            "ada\n",
            "adaia\n",
            "adalgisa\n",
            "adama\n",
            "adara\n",
            "adaya\n",
            "adela\n",
            "adela maria \n",
            "\n",
            "corpus length: 47527\n",
            "total chars: 30\n",
            "nb sequences: 15836\n",
            "Vectorization...\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "v_NlPLq1I_La",
        "colab_type": "code",
        "outputId": "268262f2-4f3e-4174-f5d0-2ae538fdf53c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "# build the model\n",
        "\n",
        "print('Build model...')\n",
        "model = Sequential()\n",
        "model.add(LSTM(64, \n",
        "#               return_sequences=True,\n",
        "               recurrent_dropout=0.2, \n",
        "               dropout=0.2, \n",
        "               input_shape=(maxlen, len(chars))))\n",
        "#model.add(LSTM(64, \n",
        "#               recurrent_dropout=0.2, \n",
        "#               dropout=0.2))\n",
        "model.add(Dense(len(chars)))\n",
        "model.add(Activation('relu'))\n",
        "model.add(Dropout(0.2))\n",
        "model.add(Activation('softmax'))\n",
        "\n",
        "optimizer = RMSprop(lr=0.01)\n",
        "model.compile(loss='categorical_crossentropy', optimizer=optimizer)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Build model...\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "oXLZiChCI_Le",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "The simplest way to use the Keras LSTM model to make predictions is to first start off with a seed sequence as input, generate the next character then update the seed sequence to add the generated character on the end and trim off the first character. \n",
        "\n",
        "This process is repeated for as long as we want to predict new characters"
      ]
    },
    {
      "metadata": {
        "id": "opkgEV7TI_Le",
        "colab_type": "code",
        "outputId": "e2fc3408-01c7-4ef8-802b-1ff4a3dba4d7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 22423
        }
      },
      "cell_type": "code",
      "source": [
        "def sample(preds, temperature=1.0):\n",
        "    # helper function to sample an index from a probability array\n",
        "    preds = np.asarray(preds).astype('float64')\n",
        "    preds = np.log(preds) / temperature\n",
        "    exp_preds = np.exp(preds)\n",
        "    preds = exp_preds / np.sum(exp_preds)\n",
        "    probas = np.random.multinomial(1, preds, 1)\n",
        "    return np.argmax(probas)\n",
        "\n",
        "# train the model, output generated text after each iteration\n",
        "for iteration in range(1, 60):\n",
        "    print()\n",
        "    print('-' * 50)\n",
        "    print('Iteration', iteration)\n",
        "    model.fit(X, y, batch_size=256, epochs=1)\n",
        "\n",
        "    start_index = random.randint(0, len(text) - maxlen - 1)\n",
        "    generated = ''\n",
        "    sentence = text[start_index: start_index + maxlen]\n",
        "    generated += sentence\n",
        "    print('----- Generating with seed: \"' + sentence.replace(\"\\n\", \" \") + '\"')\n",
        "        \n",
        "    for diversity in [0.5, 1.0]:\n",
        "        print()\n",
        "        print('----- diversity:', diversity)\n",
        "        for i in range(50):\n",
        "            \n",
        "            x = np.zeros((1, maxlen, len(chars)))\n",
        "            for t, char in enumerate(sentence):\n",
        "                x[0, t, char_indices[char]] = 1.\n",
        "\n",
        "            preds = model.predict(x, verbose=0)[0]\n",
        "            next_index = sample(preds, diversity)\n",
        "            next_char = indices_char[next_index]\n",
        "\n",
        "            generated += next_char\n",
        "            sentence = sentence[1:] + next_char\n",
        "\n",
        "            sys.stdout.write(next_char)\n",
        "            sys.stdout.flush()\n",
        "        print()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "--------------------------------------------------\n",
            "Iteration 1\n",
            "15836/15836 [==============================] - 5s 328us/sample - loss: 3.0374\n",
            "----- Generating with seed: \"salwa samah samanta \"\n",
            "\n",
            "----- diversity: 0.5\n",
            "hvfianwa\n",
            "tarii\n",
            "carna\n",
            "aanma\n",
            "a\n",
            "afela\n",
            "\n",
            "ariena\n",
            "aaradna\n",
            "\n",
            "----- diversity: 1.0\n",
            "j\n",
            "txnee\n",
            "maxfekin\n",
            "nt lina\n",
            "fa cbwetmjxçleblen\n",
            "muhxie\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 2\n",
            "15836/15836 [==============================] - 4s 259us/sample - loss: 2.7421\n",
            "----- Generating with seed: \"lla orosia otilia ou\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "lia\n",
            "ançia\n",
            "ariba\n",
            "cibena\n",
            "jelina\n",
            "iaraa\n",
            "faria\n",
            "aelra\n",
            "in\n",
            "\n",
            "----- diversity: 1.0\n",
            "aeçfa besçpallri\n",
            "vzinani\n",
            "ñrhaelen\n",
            "xiaevñod\n",
            "wnhe\n",
            "ju\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 3\n",
            "15836/15836 [==============================] - 4s 263us/sample - loss: 2.6775\n",
            "----- Generating with seed: \"en elena elena carme\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "na\n",
            "çariena\n",
            "taria alena\n",
            "laria elanza\n",
            "anina\n",
            "ananike\n",
            "\n",
            "\n",
            "----- diversity: 1.0\n",
            "coqle\n",
            "stvela \n",
            "glaiiokatvopy\n",
            "ali\n",
            "smarwa\n",
            "nçijes gpra\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 4\n",
            "15836/15836 [==============================] - 4s 260us/sample - loss: 2.6106\n",
            "----- Generating with seed: \"manda arminda armine\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "\n",
            "daruña\n",
            "taria\n",
            "naria\n",
            "fera\n",
            "belind\n",
            "gariça\n",
            "cariqgerh\n",
            "q\n",
            "\n",
            "----- diversity: 1.0\n",
            "vawts\n",
            "pulñu\n",
            "alekcbel\n",
            "amrhaç\n",
            "ormi\n",
            "ddri açizcizaaobf\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 5\n",
            "15836/15836 [==============================] - 4s 258us/sample - loss: 2.5682\n",
            "----- Generating with seed: \"lingling lioba lipin\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "a\n",
            "ylan\n",
            "ara janie\n",
            "ena wa bra\n",
            "aria ena\n",
            "saura çaria\n",
            "a\n",
            "\n",
            "----- diversity: 1.0\n",
            "lenugwrtepwañisadb añi viñjañzotsggñgia\n",
            "hcwenxa\n",
            "yç\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 6\n",
            "15836/15836 [==============================] - 4s 259us/sample - loss: 2.5115\n",
            "----- Generating with seed: \"tou neelam neha neid\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "ena\n",
            "eliya\n",
            "sina\n",
            "iela\n",
            "liriça\n",
            "fimera\n",
            "dira\n",
            "tina\n",
            "ilia\n",
            "s\n",
            "\n",
            "----- diversity: 1.0\n",
            "iaey\n",
            "gia\n",
            "iawbiqyq\n",
            "ñiñcia çladelbidbr çñ\n",
            "gianyn zik\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 7\n",
            "15836/15836 [==============================] - 4s 260us/sample - loss: 2.4840\n",
            "----- Generating with seed: \"aria martina maria m\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "aria\n",
            "inaa\n",
            "iawañta\n",
            "rana ila\n",
            "\n",
            "aria ana\n",
            "albela\n",
            "ana la\n",
            "\n",
            "----- diversity: 1.0\n",
            "emila\n",
            "\n",
            "aosa vmari\n",
            "yaelxçaha\n",
            "ale\n",
            "irejenorqzuwkrssli\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 8\n",
            "15836/15836 [==============================] - 4s 261us/sample - loss: 2.4482\n",
            "----- Generating with seed: \"aria salome maria sa\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "rtana\n",
            "anuhe \n",
            "avy\n",
            "hana\n",
            "anene\n",
            "uania\n",
            "anarina\n",
            "ana\n",
            "alin\n",
            "\n",
            "----- diversity: 1.0\n",
            "ada\n",
            "aprznaçanarzwça\n",
            "ueueia\n",
            "npzaon\n",
            "\n",
            "linatngrnwjamçd\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 9\n",
            "15836/15836 [==============================] - 4s 259us/sample - loss: 2.4251\n",
            "----- Generating with seed: \"donia celerina celes\"\n",
            "\n",
            "----- diversity: 0.5\n",
            " elita\n",
            "pari lexan\n",
            "ele\n",
            "alewa\n",
            "ani anvina\n",
            "ana kaxiana\n",
            "\n",
            "----- diversity: 1.0\n",
            "\n",
            "anayasfa\n",
            " papq\n",
            "qijanwen\n",
            "\n",
            "ariwia amarba\n",
            "da uacia\n",
            "i\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 10\n",
            "15836/15836 [==============================] - 4s 258us/sample - loss: 2.4071\n",
            "----- Generating with seed: \"ilica milka milouda \"\n",
            "\n",
            "----- diversity: 0.5\n",
            "kigasra\n",
            "liaa\n",
            "litaqia\n",
            "bliqia\n",
            "\n",
            "ilana\n",
            "lifia elena\n",
            "xiu\n",
            "\n",
            "----- diversity: 1.0\n",
            "feapiç\n",
            "driena\n",
            "rozawar\n",
            "ñçguegda\n",
            "pmxcneta\n",
            "ñt maq\n",
            "pat\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 11\n",
            "15836/15836 [==============================] - 4s 259us/sample - loss: 2.3549\n",
            "----- Generating with seed: \" diana rocio diana v\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "inte\n",
            "ilana ilana\n",
            "pirela elina\n",
            "ina virena\n",
            "ariqina s\n",
            "\n",
            "----- diversity: 1.0\n",
            "aret exinaviranfvinba\n",
            "uricankeogqiiu\n",
            "dfuwcaxzhlam\n",
            "\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 12\n",
            "15836/15836 [==============================] - 4s 266us/sample - loss: 2.3676\n",
            "----- Generating with seed: \"ar karen karen alexa\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "nina\n",
            "taria delena\n",
            "antrea deneben\n",
            "ana ilia\n",
            "ana eliz\n",
            "\n",
            "----- diversity: 1.0\n",
            "a\n",
            "abriagtbñçda\n",
            "a\n",
            "ançrdja paariu\n",
            "cakypi\n",
            "ñazfela\n",
            "uah\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 13\n",
            "15836/15836 [==============================] - 4s 269us/sample - loss: 2.3276\n",
            "----- Generating with seed: \"fatma fatna fatou fa\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "tka\n",
            "atela\n",
            "oyaja\n",
            "algelemat\n",
            "ale\n",
            "alie\n",
            "anfel\n",
            "alela aks\n",
            "\n",
            "----- diversity: 1.0\n",
            "erña\n",
            "zaria nizbia\n",
            "zalfa wñvagraxacaley\n",
            "nafis intiç\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 14\n",
            "15836/15836 [==============================] - 4s 270us/sample - loss: 2.3144\n",
            "----- Generating with seed: \"lamyaa lamyae lana l\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "aina\n",
            "laia\n",
            "leadina\n",
            "laki alena\n",
            "laria aleu\n",
            "haria liha\n",
            "\n",
            "----- diversity: 1.0\n",
            "m\n",
            "maria u ilolia\n",
            "çarq erina\n",
            "juany xaewa\n",
            "xhaurben\n",
            "w\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 15\n",
            "15836/15836 [==============================] - 4s 269us/sample - loss: 2.3068\n",
            "----- Generating with seed: \" daina daira daisy d\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "aija\n",
            "takya\n",
            "nata\n",
            "natipa\n",
            "zana\n",
            "nastina\n",
            "zani\n",
            "alina\n",
            "ani\n",
            "\n",
            "----- diversity: 1.0\n",
            "gpa\n",
            "uai\n",
            "aug\n",
            "\n",
            "waaq\n",
            "apooginra\n",
            "aza lkdramara\n",
            "vaxianna\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 16\n",
            "15836/15836 [==============================] - 4s 264us/sample - loss: 2.2855\n",
            "----- Generating with seed: \"zabeth mirian raquel\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "za\n",
            "ñdria rebel\n",
            "dcaria çaria\n",
            "zjria alixanfra\n",
            "olaza \n",
            "\n",
            "----- diversity: 1.0\n",
            "gdrwerviy\n",
            "çdoria iyakvxgaaha\n",
            "opa izatwan\n",
            "oñaçaçpeg\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 17\n",
            "15836/15836 [==============================] - 4s 260us/sample - loss: 2.2630\n",
            "----- Generating with seed: \"bigail abilia abir a\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "maria\n",
            "aarunka\n",
            "alacia\n",
            "alena barvinta\n",
            "ana zaria\n",
            "ala \n",
            "\n",
            "----- diversity: 1.0\n",
            "jttkpfeckana yaeika\n",
            "kanna lviñs\n",
            "ana  ksaduta\n",
            "karna\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 18\n",
            "15836/15836 [==============================] - 4s 258us/sample - loss: 2.2711\n",
            "----- Generating with seed: \"af intissar ioana io\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "naola\n",
            "winela\n",
            "silana  maria\n",
            "gijiana\n",
            "çilia\n",
            "nivia\n",
            "ni\n",
            "\n",
            "\n",
            "----- diversity: 1.0\n",
            "n\n",
            "xixweqeerboshucifyniña\n",
            "dinmhsjengszikia\n",
            "kizmmafj\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 19\n",
            "15836/15836 [==============================] - 4s 257us/sample - loss: 2.2460\n",
            "----- Generating with seed: \"a judit maria judith\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "er\n",
            "karia qiqena\n",
            "alia antenia\n",
            "ana laria\n",
            "ana dinta\n",
            "a\n",
            "\n",
            "----- diversity: 1.0\n",
            "ncgle eltiniazfagina behzxpataç\n",
            "nara dfles\n",
            "dpamtnq\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 20\n",
            "15836/15836 [==============================] - 4s 259us/sample - loss: 2.2313\n",
            "----- Generating with seed: \"bah sabela sabiha sa\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "lah\n",
            "catça\n",
            "mataria\n",
            "zatala\n",
            "ñatara\n",
            "zatala\n",
            "ban\n",
            "ñat car\n",
            "\n",
            "----- diversity: 1.0\n",
            "tsçna\n",
            "karna aljka\n",
            "haria dkaoz\n",
            "takhca natelda\n",
            "çatvi\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 21\n",
            "15836/15836 [==============================] - 4s 261us/sample - loss: 2.2320\n",
            "----- Generating with seed: \"ieves maria noel mar\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "ia tarouita\n",
            "ñaria ganpre\n",
            "maria qilena\n",
            "qidaria alet\n",
            "\n",
            "----- diversity: 1.0\n",
            "ana\n",
            "criqia maria\n",
            "miopkkah \n",
            "di geworia\n",
            "rphela\n",
            "çjjan\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 22\n",
            "15836/15836 [==============================] - 4s 265us/sample - loss: 2.2407\n",
            "----- Generating with seed: \"abriela andrea isabe\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "l\n",
            "antia anlee\n",
            "ana telela\n",
            "ana haria\n",
            "ana aleena\n",
            "ana \n",
            "\n",
            "----- diversity: 1.0\n",
            "viwena\n",
            "anayady\n",
            "anoñple\n",
            "anlah fxejpnñma\n",
            "anu\n",
            "euoliag\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 23\n",
            "15836/15836 [==============================] - 4s 259us/sample - loss: 2.2112\n",
            "----- Generating with seed: \"jacoba jacqueline ja\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "min\n",
            "ñahia\n",
            "haxina\n",
            "danima\n",
            " ahre\n",
            "nacita\n",
            "najia\n",
            "najia\n",
            "n\n",
            "\n",
            "----- diversity: 1.0\n",
            "a ñzwa\n",
            "maf\n",
            "qaxçana\n",
            "casulao carzqi\n",
            "raxxa açcvia\n",
            "uda\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 24\n",
            "15836/15836 [==============================] - 4s 257us/sample - loss: 2.2160\n",
            "----- Generating with seed: \"ina ana catalina ana\"\n",
            "\n",
            "----- diversity: 0.5\n",
            " lasel\n",
            "ana  areza\n",
            "ana liana\n",
            "ana  arpanbela\n",
            "ala xil\n",
            "\n",
            "----- diversity: 1.0\n",
            "oria alqeefaxulza\n",
            "xjana lrepça\n",
            "vacienxpagjpj\n",
            "kaçan\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 25\n",
            "15836/15836 [==============================] - 4s 262us/sample - loss: 2.2037\n",
            "----- Generating with seed: \"enia yesica yesika y\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "ehina\n",
            "çelia\n",
            "ewira\n",
            "ocelia\n",
            "jensi\n",
            "maria elena\n",
            "granien\n",
            "\n",
            "----- diversity: 1.0\n",
            "\n",
            "\n",
            "rarxa\n",
            "ely qiharia\n",
            "zalñassqa\n",
            "iagvça\n",
            "\n",
            "queoklia\n",
            "riç\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 26\n",
            "15836/15836 [==============================] - 4s 260us/sample - loss: 2.1954\n",
            "----- Generating with seed: \" ceferina celedonia \"\n",
            "\n",
            "----- diversity: 0.5\n",
            "xelgiña\n",
            "elena ñajan\n",
            "ñeta ilar\n",
            "nataria aleka\n",
            "ana la\n",
            "\n",
            "----- diversity: 1.0\n",
            "citza\n",
            "nawig patnixypy\n",
            "qajanrabñtg\n",
            "haviply\n",
            "hanz\n",
            "mar\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 27\n",
            "15836/15836 [==============================] - 4s 258us/sample - loss: 2.1901\n",
            "----- Generating with seed: \" haidee haifen haiya\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "\n",
            "ñaija\n",
            "iana\n",
            "naia\n",
            "naila\n",
            "naia\n",
            "naiba\n",
            "iana\n",
            "iena\n",
            "inat\n",
            "i\n",
            "\n",
            "----- diversity: 1.0\n",
            "fmepvan\n",
            "iaofua\n",
            "iwafir\n",
            "eilhe\n",
            "iania\n",
            "jiaktuba\n",
            "xiaksge\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 28\n",
            "15836/15836 [==============================] - 4s 255us/sample - loss: 2.1806\n",
            "----- Generating with seed: \"liana juliana andrea\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "\n",
            "alzia qilina\n",
            "alivia\n",
            "alina\n",
            "alziwa\n",
            "alñqia\n",
            "alelen\n",
            "al\n",
            "\n",
            "----- diversity: 1.0\n",
            "epel\n",
            "alela attren\n",
            "avviela oleda\n",
            "aliça  vgrja\n",
            "aliga\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 29\n",
            "15836/15836 [==============================] - 4s 260us/sample - loss: 2.1732\n",
            "----- Generating with seed: \"pcion maria concha c\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "vintina\n",
            "j\n",
            "xina\n",
            "jxlina\n",
            " torina\n",
            "qjenzina\n",
            "edielina\n",
            "ex\n",
            "\n",
            "----- diversity: 1.0\n",
            "xiawele\n",
            "çka xñaa\n",
            "ñaocira ya\n",
            "\n",
            "iñuta\n",
            "wik elred\n",
            "çiiaw\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 30\n",
            "15836/15836 [==============================] - 4s 258us/sample - loss: 2.1767\n",
            "----- Generating with seed: \"sandra maria alena a\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "lena  erena\n",
            "alelina geryania\n",
            "aleiana imaria\n",
            "aleka \n",
            "\n",
            "----- diversity: 1.0\n",
            "alekarkastewqyaiñla\n",
            "mpivwelmm\n",
            "tentc jgha\n",
            "tenma fel\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 31\n",
            "15836/15836 [==============================] - 4s 256us/sample - loss: 2.1856\n",
            "----- Generating with seed: \"agmar daiana daina d\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "aira\n",
            "xiatixan\n",
            "jaigdina\n",
            "zariaa\n",
            "ialina\n",
            "iahina\n",
            "iaqola\n",
            "\n",
            "----- diversity: 1.0\n",
            "\n",
            "iacna\n",
            "idnna\n",
            "izako ñaira\n",
            "hiugca qilyouna\n",
            "bñcfelfe\n",
            "\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 32\n",
            "15836/15836 [==============================] - 4s 256us/sample - loss: 2.1581\n",
            "----- Generating with seed: \"a jose francisca jos\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "ejanlwanka\n",
            "maria maria\n",
            "maria maria\n",
            "alarstela\n",
            "allia\n",
            "\n",
            "----- diversity: 1.0\n",
            "valiha\n",
            "alkpvioh\n",
            "algkagzka\n",
            "apcilajmakerbetqdebñnru\n",
            "\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 33\n",
            "15836/15836 [==============================] - 4s 260us/sample - loss: 2.1333\n",
            "----- Generating with seed: \"ila silke silvana si\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "lkme\n",
            "çiliya\n",
            "mildia\n",
            "liqia\n",
            "lida\n",
            "lista\n",
            "liela\n",
            "liqia\n",
            "li\n",
            "\n",
            "----- diversity: 1.0\n",
            "dka\n",
            "lihermwil çahbea\n",
            "liujsaqilz\n",
            "lizbkly\n",
            "liondjma\n",
            "l\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 34\n",
            "15836/15836 [==============================] - 4s 258us/sample - loss: 2.1354\n",
            "----- Generating with seed: \"alia thelma theresa \"\n",
            "\n",
            "----- diversity: 0.5\n",
            "deçena\n",
            "merecia\n",
            "yeliiana\n",
            "meri\n",
            "\n",
            "ereniar\n",
            "maria mereza\n",
            "\n",
            "----- diversity: 1.0\n",
            "ftaña\n",
            "maria ethis\n",
            "maria zmxiu\n",
            "maria qñgwiwa\n",
            "rmarm\n",
            "\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 35\n",
            "15836/15836 [==============================] - 4s 260us/sample - loss: 2.1097\n",
            "----- Generating with seed: \"alicia isabel alicia\"\n",
            "\n",
            "----- diversity: 0.5\n",
            " ila\n",
            "anila aleka\n",
            "aniela ihela\n",
            "anelia elena\n",
            "anna lh\n",
            "\n",
            "----- diversity: 1.0\n",
            "aria\n",
            "ana filiba\n",
            "aowa mtpika\n",
            "abrianhiz\n",
            "algqvitu\n",
            "aqy\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 36\n",
            "15836/15836 [==============================] - 4s 260us/sample - loss: 2.1117\n",
            "----- Generating with seed: \" maria gabriele gabr\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "iela\n",
            "qafi maria\n",
            "maria meria\n",
            "maria igane\n",
            "maria elid\n",
            "\n",
            "----- diversity: 1.0\n",
            "xewa\n",
            "maria awdrea\n",
            "maria allena\n",
            "vañtixyn maria\n",
            "xnxt\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 37\n",
            "15836/15836 [==============================] - 4s 263us/sample - loss: 2.1035\n",
            "----- Generating with seed: \"onarda leoncia leone\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "ña\n",
            "legenia\n",
            "lelita\n",
            "leeli\n",
            "leina\n",
            "leiça\n",
            "leina\n",
            "leia\n",
            "lei\n",
            "\n",
            "----- diversity: 1.0\n",
            "a\n",
            "llxenxalenhelq zpbilretja\n",
            "csakfzrgabetyciquerxas\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 38\n",
            "15836/15836 [==============================] - 4s 259us/sample - loss: 2.0981\n",
            "----- Generating with seed: \"ndra maria alfonsa m\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "aria maria\n",
            "maria lirina\n",
            "maria zaqmel\n",
            "maria \n",
            "maria \n",
            "\n",
            "----- diversity: 1.0\n",
            "pakruq\n",
            "raojkata\n",
            "roviuwel\n",
            "bcvaluwe\n",
            "ñgqwelia\n",
            " çt\n",
            "\n",
            "wn\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 39\n",
            "15836/15836 [==============================] - 4s 261us/sample - loss: 2.0974\n",
            "----- Generating with seed: \"nna katarzyna anna m\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "aria\n",
            "anna maria\n",
            "anna alecia\n",
            "anna zarina\n",
            "ana maria\n",
            "\n",
            "\n",
            "----- diversity: 1.0\n",
            "ara cliyana\n",
            "ana uvoneka\n",
            "ana marolin\n",
            "ana qiaq\n",
            "inats\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 40\n",
            "15836/15836 [==============================] - 4s 262us/sample - loss: 2.0840\n",
            "----- Generating with seed: \"geles rosa angelica \"\n",
            "\n",
            "----- diversity: 0.5\n",
            "rweha maria\n",
            "rewina\n",
            "rewina\n",
            "reima\n",
            "heria\n",
            "nersi\n",
            "cermen\n",
            "\n",
            "----- diversity: 1.0\n",
            "iwe\n",
            "leg\n",
            "le\n",
            "ljimqee\n",
            "lefd beatrir\n",
            "lipla aliyar\n",
            "liae\n",
            "\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 41\n",
            "15836/15836 [==============================] - 4s 263us/sample - loss: 2.0973\n",
            "----- Generating with seed: \"har kayla keiko keil\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "a\n",
            "nelena\n",
            "nelina\n",
            "nelina\n",
            "nelina\n",
            "nelina\n",
            "nela\n",
            "nelena\n",
            "n\n",
            "\n",
            "----- diversity: 1.0\n",
            "eli elekah\n",
            "nyhbih\n",
            "npikukag\n",
            "naeliwat\n",
            "naçitabma\n",
            "naze\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 42\n",
            "15836/15836 [==============================] - 4s 263us/sample - loss: 2.0895\n",
            "----- Generating with seed: \"abel adriana lucia a\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "ra maria\n",
            "ara maria\n",
            "ara maria\n",
            "ara ntiça\n",
            "ara salia\n",
            "a\n",
            "\n",
            "----- diversity: 1.0\n",
            "qrakqinejanhazaiwalfa\n",
            "ona ñafrawkfsaa\n",
            "zfohacpu\n",
            "lfñ\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 43\n",
            "15836/15836 [==============================] - 4s 261us/sample - loss: 2.0999\n",
            "----- Generating with seed: \"ydia lyubov lyudmila\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "\n",
            "l alia\n",
            "lqia\n",
            "lia\n",
            "liela\n",
            "licia\n",
            "lika\n",
            "lila\n",
            "lipig\n",
            "lihel\n",
            "\n",
            "----- diversity: 1.0\n",
            "wer\n",
            "\n",
            "ijo\n",
            "xivçevle\n",
            "hsi\n",
            "btñeiñe\n",
            "xianoaño\n",
            "piutjaku\n",
            "mi\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 44\n",
            "15836/15836 [==============================] - 4s 264us/sample - loss: 2.0716\n",
            "----- Generating with seed: \"ta maria alcazar mar\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "ia elena\n",
            "maria elena\n",
            "maria delena\n",
            "maria jeania\n",
            "mar\n",
            "\n",
            "----- diversity: 1.0\n",
            "ia i ttenxeljmariaza\n",
            "dracieña\n",
            "rcgzag\n",
            "rilena\n",
            "hidlao\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 45\n",
            "15836/15836 [==============================] - 4s 261us/sample - loss: 2.0626\n",
            "----- Generating with seed: \"ria robledo maria ro\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "da\n",
            "maria qvanela\n",
            "maria roure\n",
            "maria orena\n",
            "maria ain\n",
            "\n",
            "----- diversity: 1.0\n",
            "er\n",
            "maria beatricv\n",
            "maria ggganxanuka\n",
            "marba q\n",
            "marta\n",
            "\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 46\n",
            "15836/15836 [==============================] - 4s 261us/sample - loss: 2.0687\n",
            "----- Generating with seed: \" paola alexandra pao\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "la elita\n",
            "uania elena\n",
            "ana maria\n",
            "ana lorena\n",
            "ana raer\n",
            "\n",
            "----- diversity: 1.0\n",
            "ita\n",
            "anaeuzuenta\n",
            "anliajangzcbeta\n",
            "aniifa\n",
            "anixane\n",
            "anu\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 47\n",
            "15836/15836 [==============================] - 4s 257us/sample - loss: 2.0694\n",
            "----- Generating with seed: \"a heliodora henar he\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "nva\n",
            " eça\n",
            "eeya\n",
            "peza\n",
            "sefha\n",
            "emiana\n",
            "erelina\n",
            "erita\n",
            "erih\n",
            "\n",
            "----- diversity: 1.0\n",
            "o\n",
            "eridkexa\n",
            "eriydke\n",
            "ereciia tjexanla\n",
            "\n",
            "ergafeñqghae\n",
            "\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 48\n",
            "15836/15836 [==============================] - 4s 262us/sample - loss: 2.0664\n",
            "----- Generating with seed: \"alentyna valeria val\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "\n",
            "zalga\n",
            "natja\n",
            "qadina\n",
            "banetcia\n",
            "zanisg\n",
            "banija\n",
            "narina\n",
            "\n",
            "\n",
            "----- diversity: 1.0\n",
            "\n",
            "çakefne\n",
            "kafawin\n",
            " nanññ\n",
            "ñkamia\n",
            "lua ñbea\n",
            "lhay\n",
            "lmaur\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 49\n",
            "15836/15836 [==============================] - 4s 261us/sample - loss: 2.0566\n",
            "----- Generating with seed: \"arolina wendy elizab\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "etl\n",
            "wene\n",
            "tene\n",
            "nelena\n",
            "nela\n",
            "nele\n",
            "nele\n",
            "nele\n",
            "neli\n",
            "neli\n",
            "\n",
            "----- diversity: 1.0\n",
            "td bez\n",
            "tennzaiza\n",
            "tecuo\n",
            "tetlmsma\n",
            "\n",
            "eibpg\n",
            "cexinafyfza\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 50\n",
            "15836/15836 [==============================] - 4s 264us/sample - loss: 2.0770\n",
            "----- Generating with seed: \" elisea elisenda eli\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "za\n",
            "elica\n",
            "elina\n",
            "elenia\n",
            "elione\n",
            "elenia\n",
            "elica\n",
            "eliña\n",
            "el\n",
            "\n",
            "----- diversity: 1.0\n",
            "etqive\n",
            "eleça\n",
            "eleva\n",
            "ellfna\n",
            "eltgerçanu\n",
            "elela fjijanc\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 51\n",
            "15836/15836 [==============================] - 4s 262us/sample - loss: 2.0555\n",
            "----- Generating with seed: \"anda mireia mireia m\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "aria\n",
            "mir\n",
            "hirano\n",
            " ineta\n",
            "firna\n",
            "finñe\n",
            "mirfe\n",
            "mirena\n",
            "mi\n",
            "\n",
            "----- diversity: 1.0\n",
            "rtajn\n",
            "minma\n",
            "mina\n",
            "minjmakrtzw\n",
            "miz kzabma\n",
            "milkajenti\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 52\n",
            "15836/15836 [==============================] - 4s 263us/sample - loss: 2.0636\n",
            "----- Generating with seed: \"xiliadora ava avelin\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "a\n",
            "alone\n",
            "alenna\n",
            "alenna\n",
            "aleria\n",
            "alenxania\n",
            "aleuvia\n",
            "ale\n",
            "\n",
            "----- diversity: 1.0\n",
            "aycekaq\n",
            "alca beatria\n",
            "anggoh gucikla\n",
            "ane tacçrexa\n",
            "a\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 53\n",
            "15836/15836 [==============================] - 4s 269us/sample - loss: 2.0566\n",
            "----- Generating with seed: \"vidad maria nazaret \"\n",
            "\n",
            "----- diversity: 0.5\n",
            "maria ançela\n",
            "maria antvilia\n",
            "maria jene\n",
            "maria anher\n",
            "\n",
            "----- diversity: 1.0\n",
            "u\n",
            "maria benena\n",
            "maria wança\n",
            "maria çynmxeldena\n",
            "maria\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 54\n",
            "15836/15836 [==============================] - 4s 271us/sample - loss: 2.0387\n",
            "----- Generating with seed: \"a paola ginebra gine\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "la\n",
            "ñiqi\n",
            "livia uñaça\n",
            "liñvia\n",
            "lipa\n",
            "liqi\n",
            "lilça\n",
            "lipilia\n",
            "\n",
            "----- diversity: 1.0\n",
            "\n",
            "lilxencwañylil\n",
            "lifa\n",
            "lizahviyça\n",
            "licvicia\n",
            "lieun\n",
            "lig\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 55\n",
            "15836/15836 [==============================] - 4s 272us/sample - loss: 2.0544\n",
            "----- Generating with seed: \"ahira chahrazad chai\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "ga\n",
            "sjuana\n",
            "ha maria\n",
            "iaona\n",
            "vanira\n",
            "erania\n",
            "maria elena\n",
            "\n",
            "----- diversity: 1.0\n",
            "\n",
            "malia smaria\n",
            "maria jenet\n",
            "maria tezra\n",
            "maria g\n",
            "mari\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 56\n",
            "15836/15836 [==============================] - 4s 267us/sample - loss: 2.0495\n",
            "----- Generating with seed: \"ea gador gaelle gaga\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "le\n",
            "zaka\n",
            "daama\n",
            "vamina\n",
            "uat eliubat\n",
            "etata\n",
            "yatqia\n",
            "kats\n",
            "\n",
            "----- diversity: 1.0\n",
            "ie\n",
            "hatka\n",
            "yatiujmaxal\n",
            "çatudi\n",
            "çavegki\n",
            "nayba\n",
            " axñwas\n",
            "\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 57\n",
            "15836/15836 [==============================] - 4s 266us/sample - loss: 2.0448\n",
            "----- Generating with seed: \"a cesca chadia chafi\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "la\n",
            "dcana\n",
            "xene\n",
            "rea\n",
            "reia\n",
            "regizah\n",
            "reiana\n",
            "reia\n",
            "reivia\n",
            "\n",
            "\n",
            "----- diversity: 1.0\n",
            "peleljinceka\n",
            "maaica\n",
            "mayofcana\n",
            "mawa\n",
            "madtermy\n",
            "matha\n",
            "\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 58\n",
            "15836/15836 [==============================] - 4s 258us/sample - loss: 2.0333\n",
            "----- Generating with seed: \"me diarra mamma mana\"\n",
            "\n",
            "----- diversity: 0.5\n",
            "\n",
            "mana\n",
            "mana\n",
            "manna\n",
            "manja\n",
            "manarina\n",
            "mania\n",
            "manar\n",
            "manbel\n",
            "\n",
            "----- diversity: 1.0\n",
            " elena\n",
            "marsaceliamça\n",
            "martelfa maññmça\n",
            "machqgerem\n",
            "m\n",
            "\n",
            "--------------------------------------------------\n",
            "Iteration 59\n",
            "15836/15836 [==============================] - 4s 263us/sample - loss: 2.0380\n",
            "----- Generating with seed: \"arla patricia carla \"\n",
            "\n",
            "----- diversity: 0.5\n",
            "maria\n",
            "çatarina\n",
            "al\n",
            "ala\n",
            "ala\n",
            "aliana\n",
            "alalina\n",
            "aleda\n",
            "ale\n",
            "\n",
            "----- diversity: 1.0\n",
            "qjsansa\n",
            "alffmwins\n",
            "alina\n",
            "alyke\n",
            "awulelska\n",
            "azapna\n",
            "aaw\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "fyPOcNhtI_Lh",
        "colab_type": "code",
        "outputId": "df465c8a-e203-4d8c-f34b-ac0ffd937a05",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "'aliana' in text"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "False"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "metadata": {
        "collapsed": true,
        "id": "2RnFynMYI_Lk",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "In the case of places, after several hours you can generate names such as:\n",
        "\n",
        "+ Alzinetes, torrent de les\n",
        "+ Alzinetes, vall de les\n",
        "+ **Alzinó, Mas d'**\n",
        "+ Alzinosa, collada de l'\n",
        "+ Alzinosa, font de l'\n",
        "\n",
        "-\n",
        "\n",
        "+ Benavent, roc de\n",
        "+ Benaviure, Cal\n",
        "+ **Benca**\n",
        "+ Bendiners, pla de\n",
        "+ Benedi, roc del\n",
        "\n",
        "-\n",
        "\n",
        "+ Fiola, la\n",
        "+ Fiola, puig de la\n",
        "+ **Fiper, Granja del**\n",
        "+ Firassa, Finca\n",
        "+ Firell\n",
        "\n",
        "-\n",
        "\n",
        "+ Regueret, lo\n",
        "+ Regueret, lo\n",
        "+ **Regueró**\n",
        "+ Reguerols, els\n",
        "+ Reguerons, els\n",
        "\n",
        "-\n",
        "\n",
        "+ Vallverdú, Mas de\n",
        "+ Vallverdú, serrat de\n",
        "+ **Vallvicamanyà**\n",
        "+ Vallvidrera\n",
        "+ Vallvidrera, riera de\n",
        "\n",
        "-\n",
        "\n",
        "+ Terraubella, Corral de\n",
        "+ Terraubes\n",
        "+ **Terravanca**\n",
        "+ Terrer Nou, Can\n",
        "+ Terrer Roig, lo\n",
        "\n",
        "where names in **bold** are generated and other names are the nearest neighbours (in the training dataset) of the generated name."
      ]
    },
    {
      "metadata": {
        "id": "K8RRMNP7awH9",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "## Recommended Readings\n",
        "\n",
        "+ [Neural Machine Translation With Attention Mechanism](https://machinetalk.org/2019/03/29/neural-machine-translation-with-attention-mechanism/)"
      ]
    }
  ]
}