DeepAsr_CTC_Pipeline.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# DeepAsr (DeepAsrNetwork1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "S0FiiN9Y0FEs"
   },
   "outputs": [],
   "source": [
    "# !wget http://www.openslr.org/resources/12/train-clean-100.tar.gz"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "5JJMHx460FE4"
   },
   "outputs": [],
   "source": [
    "# !tar xzvf train-clean-100.tar.gz"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "SHLb6nDsUwkN"
   },
   "outputs": [],
   "source": [
    "# ! pip install tensorflow==2.1.0"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Vx9UdVs5384B"
   },
   "source": [
    "# 1. Prepare DataSet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "kw_18d180FFM"
   },
   "outputs": [],
   "source": [
    "import os\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import deepasr as asr\n",
    "import librosa"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "d29QeHTJVNOF",
    "outputId": "b79a2fa5-e783-4543-d5b1-125fb3a1bd92"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.1.0'"
      ]
     },
     "execution_count": 5,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "zQEGr0HfC5OF",
    "outputId": "88d851cc-ec17-42d8-c8a2-26cfda9506cb"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'0.0.9'"
      ]
     },
     "execution_count": 6,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "asr.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "NmOf6DzG0FFS"
   },
   "outputs": [],
   "source": [
    "# get audios and transcripts\n",
    "org_path = './LibriSpeech/train-clean-100/'\n",
    "count = 0\n",
    "inp = []\n",
    "k=0\n",
    "audio_name = []\n",
    "audio_trans = []\n",
    "for dir1 in os.listdir(org_path):\n",
    "  dir2_path = org_path+dir1+'/'\n",
    "  #print(dir2_path)\n",
    "  for dir2 in os.listdir(dir2_path):\n",
    "    dir3_path = dir2_path+dir2+'/'\n",
    "    \n",
    "    for audio in os.listdir(dir3_path):\n",
    "      if audio.endswith('.txt'):\n",
    "        k+=1\n",
    "        file_path = dir3_path + audio\n",
    "        with open(file_path) as f:\n",
    "          line = f.readlines()\n",
    "          for lines in line:\n",
    "            flac_path = dir3_path+lines.split()[0]+'.flac'\n",
    "            \n",
    "            audio_name.append(flac_path)\n",
    "\n",
    "            # print(cmd)\n",
    "            words2 = lines.split()[1:]\n",
    "            words4=' '.join(words2)\n",
    "            audio_trans.append(words4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "5E9POoGc0FFb"
   },
   "outputs": [],
   "source": [
    "# create dataset\n",
    "df = pd.DataFrame({\"path\":audio_name,\"transcripts\":audio_trans})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "TRuPCDxZrhJu",
    "outputId": "64b3dc3f-ec85-4fc4-cf4e-60204b6f719e"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(28539, 2)"
      ]
     },
     "execution_count": 9,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "g4bePqQvri5Q"
   },
   "outputs": [],
   "source": [
    "# filter transcript less than 100 charcters\n",
    "train_data = df[df['transcripts'].str.len() < 100]\n",
    "# train_df = df.sample(n = 3000) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "fiM94FU3rkh7",
    "outputId": "9c60db2a-8b85-47e9-a294-5e46d7c2c41e"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3194, 2)"
      ]
     },
     "execution_count": 11,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_data.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "EMDC5MYk4AyL"
   },
   "source": [
    "# 2. Prepare DeepAsr CTC Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "-youl7Mo0FFi"
   },
   "outputs": [],
   "source": [
    "# get CTCPipeline\n",
    "def get_config(feature_type: str = 'spectrogram', multi_gpu: bool = False):\n",
    "    # audio feature extractor\n",
    "    features_extractor = asr.features.preprocess(feature_type=feature_type, features_num=161,\n",
    "                                                 samplerate=16000,\n",
    "                                                 winlen=0.02,\n",
    "                                                 winstep=0.025,\n",
    "                                                 winfunc=np.hanning)\n",
    "    \n",
    "    # input label encoder\n",
    "    alphabet_en = asr.vocab.Alphabet(lang='en')\n",
    "    # training model\n",
    "    model = asr.model.get_deepasrnetwork1(\n",
    "        input_dim=161,\n",
    "        output_dim=29,\n",
    "        is_mixed_precision=True\n",
    "    )\n",
    "    # model optimizer\n",
    "    optimizer = tf.keras.optimizers.Adam(\n",
    "        lr=1e-4,\n",
    "        beta_1=0.9,\n",
    "        beta_2=0.999,\n",
    "        epsilon=1e-8\n",
    "    )\n",
    "    # output label deocder\n",
    "    decoder = asr.decoder.GreedyDecoder()\n",
    "    # CTCPipeline\n",
    "    pipeline = asr.pipeline.ctc_pipeline.CTCPipeline(\n",
    "        alphabet=alphabet_en, features_extractor=features_extractor, model=model, optimizer=optimizer, decoder=decoder,\n",
    "        sample_rate=16000, mono=True, multi_gpu=multi_gpu\n",
    "    )\n",
    "    return pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "MqdfySzuRtk5"
   },
   "outputs": [],
   "source": [
    "# CTCPiline for asr\n",
    "pipeline = get_config(feature_type = 'fbank', multi_gpu=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "WTG8iEwS4NKU"
   },
   "source": [
    "# 3. Model traning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "6QMxCI8T0qMK",
    "outputId": "a598acfb-1cce-41db-db3b-b2d5cdc062d9"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"DeepAsr\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "the_input (InputLayer)          [(None, None, 161)]  0                                            \n",
      "__________________________________________________________________________________________________\n",
      "BN_1 (BatchNormalization)       (None, None, 161)    644         the_input[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "Conv1D_1 (Conv1D)               (None, None, 220)    177320      BN_1[0][0]                       \n",
      "__________________________________________________________________________________________________\n",
      "CNBN_1 (BatchNormalization)     (None, None, 220)    880         Conv1D_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "Conv1D_2 (Conv1D)               (None, None, 220)    242220      CNBN_1[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "CNBN_2 (BatchNormalization)     (None, None, 220)    880         Conv1D_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "gru_1 (GRU)                     (None, None, 512)    1127424     CNBN_2[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "gru_2 (GRU)                     (None, None, 512)    1127424     CNBN_2[0][0]                     \n",
      "__________________________________________________________________________________________________\n",
      "concatenate (Concatenate)       (None, None, 1024)   0           gru_1[0][0]                      \n",
      "                                                                 gru_2[0][0]                      \n",
      "__________________________________________________________________________________________________\n",
      "BN_2 (BatchNormalization)       (None, None, 1024)   4096        concatenate[0][0]                \n",
      "__________________________________________________________________________________________________\n",
      "time_distributed (TimeDistribut (None, None, 30)     30750       BN_2[0][0]                       \n",
      "__________________________________________________________________________________________________\n",
      "the_output (TimeDistributed)    (None, None, 29)     899         time_distributed[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "the_labels (InputLayer)         [(None, None)]       0                                            \n",
      "__________________________________________________________________________________________________\n",
      "input_length (InputLayer)       [(None, 1)]          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "label_length (InputLayer)       [(None, 1)]          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "ctc (Lambda)                    (None, 1)            0           the_output[0][0]                 \n",
      "                                                                 the_labels[0][0]                 \n",
      "                                                                 input_length[0][0]               \n",
      "                                                                 label_length[0][0]               \n",
      "==================================================================================================\n",
      "Total params: 2,712,537\n",
      "Trainable params: 2,709,287\n",
      "Non-trainable params: 3,250\n",
      "__________________________________________________________________________________________________\n",
      "Feature Extraction in progress...\n",
      "Feature Extraction completed.\n",
      "input features:  (3194, 593, 161)\n",
      "input labels:  (3194, 99)\n",
      "Model training initiated...\n",
      "Train on 3194 samples\n",
      "Epoch 1/500\n",
      "3194/3194 [==============================] - 48s 15ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 2/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 3/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 4/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 5/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 6/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 7/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 8/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 9/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 10/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 11/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 12/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 13/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 14/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 15/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 16/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 17/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 18/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 19/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 20/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 21/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 22/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 23/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 24/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 25/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 26/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 27/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 28/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 29/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 30/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 31/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 32/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 33/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 34/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 35/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 36/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 37/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 38/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 39/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 40/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 41/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 42/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 43/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 44/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 45/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 46/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 47/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 48/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 49/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 50/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 51/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 52/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 53/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 54/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 55/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 56/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 57/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 58/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 59/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 60/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 61/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 62/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 63/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 64/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 65/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 66/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 67/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 68/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 69/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 70/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 71/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 72/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 73/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 74/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 75/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 76/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 77/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 78/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 79/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 80/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 81/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 82/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 83/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 84/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 85/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 86/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 87/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 88/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 89/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 90/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 91/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 92/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 93/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 94/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 95/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 96/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 97/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 98/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 99/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 100/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 101/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 102/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 103/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 104/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 105/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 106/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 107/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 108/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 109/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 110/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 111/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 112/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 113/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 114/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 115/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 116/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 117/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 118/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 119/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 120/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 121/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 122/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 123/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 124/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 125/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 126/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 127/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 128/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 129/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 130/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 131/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 132/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 133/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 134/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 135/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 136/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 137/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 138/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 139/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 140/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 141/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 142/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 143/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 144/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 145/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 146/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 147/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 148/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 149/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 150/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 151/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 152/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 153/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 154/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 155/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 156/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 157/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 158/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 159/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 160/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 161/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 162/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 163/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 164/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 165/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 166/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 167/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 168/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 169/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 170/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 171/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 172/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 173/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 174/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 175/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 176/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 177/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 178/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 179/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 180/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 181/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 182/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 183/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 184/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 185/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 186/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 187/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 188/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 189/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 190/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 191/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 192/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 193/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 194/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 195/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 196/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 197/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 198/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 199/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 200/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 201/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 202/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 203/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 204/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 205/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 206/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 207/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 208/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 209/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 210/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 211/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 212/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 213/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 214/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 215/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 216/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 217/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 218/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 219/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 220/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 221/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 222/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 223/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 224/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 225/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 226/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 227/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 228/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 229/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 230/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 231/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 232/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 233/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 234/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 235/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 236/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 237/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 238/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 239/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 240/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 241/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 242/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 243/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 244/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 245/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 246/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 247/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 248/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 249/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 250/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 251/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 252/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 253/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 254/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 255/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 256/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 257/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 258/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 259/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 260/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 261/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 262/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 263/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 264/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 265/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 266/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 267/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 268/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 269/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 270/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 271/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 272/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 273/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 274/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 275/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 276/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 277/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 278/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 279/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 280/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 281/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 282/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 283/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 284/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 285/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 286/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 287/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 288/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 289/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 290/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 291/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 292/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 293/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 294/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 295/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 296/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 297/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 298/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 299/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 300/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 301/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 302/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 303/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 304/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 305/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 306/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 307/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 308/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 309/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 310/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 311/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 312/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 313/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 314/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 315/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 316/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 317/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 318/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 319/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 320/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 321/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 322/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 323/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 324/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 325/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 326/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 327/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 328/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 329/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 330/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 331/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 332/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 333/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 334/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 335/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 336/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 337/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 338/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 339/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 340/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 341/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 342/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 343/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 344/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 345/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 346/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 347/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 348/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 349/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 350/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 351/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 352/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 353/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 354/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 355/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 356/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 357/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 358/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 359/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 360/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 361/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 362/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 363/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 364/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 365/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 366/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 367/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 368/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 369/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 370/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 371/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 372/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 373/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 374/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 375/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 376/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 377/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 378/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 379/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 380/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 381/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 382/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 383/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 384/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 385/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 386/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 387/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 388/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 389/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 390/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 391/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 392/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 393/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 394/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 395/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 396/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 397/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 398/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 399/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 400/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 401/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 402/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 403/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 404/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 405/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 406/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 407/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 408/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 409/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 410/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 411/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 412/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 413/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 414/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 415/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 416/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 417/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 418/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 9.3926e-04\n",
      "Epoch 419/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 420/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 6.2617e-04\n",
      "Epoch 421/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 422/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 423/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 424/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 9.3926e-04\n",
      "Epoch 425/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 426/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 9.3926e-04\n",
      "Epoch 427/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 428/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 429/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 430/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 431/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0000e+00\n",
      "Epoch 432/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 433/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 6.2617e-04\n",
      "Epoch 434/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 6.2617e-04\n",
      "Epoch 435/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 6.2617e-04\n",
      "Epoch 436/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 437/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 438/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 439/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 6.2617e-04\n",
      "Epoch 440/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 441/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 3.1309e-04\n",
      "Epoch 442/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 6.2617e-04\n",
      "Epoch 443/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0016\n",
      "Epoch 444/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0013\n",
      "Epoch 445/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0013\n",
      "Epoch 446/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0013\n",
      "Epoch 447/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 9.3926e-04\n",
      "Epoch 448/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0016\n",
      "Epoch 449/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0013\n",
      "Epoch 450/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0019\n",
      "Epoch 451/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 9.3926e-04\n",
      "Epoch 452/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0013\n",
      "Epoch 453/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0041\n",
      "Epoch 454/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0028\n",
      "Epoch 455/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0034\n",
      "Epoch 456/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0019\n",
      "Epoch 457/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0025\n",
      "Epoch 458/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0013\n",
      "Epoch 459/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0028\n",
      "Epoch 460/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0025\n",
      "Epoch 461/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0022\n",
      "Epoch 462/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0031\n",
      "Epoch 463/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0044\n",
      "Epoch 464/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0028\n",
      "Epoch 465/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0031\n",
      "Epoch 466/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 9.3926e-04\n",
      "Epoch 467/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 9.3926e-04\n",
      "Epoch 468/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0013\n",
      "Epoch 469/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0047\n",
      "Epoch 470/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0044\n",
      "Epoch 471/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0085\n",
      "Epoch 472/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0100\n",
      "Epoch 473/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0163\n",
      "Epoch 474/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0157\n",
      "Epoch 475/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0059\n",
      "Epoch 476/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0053\n",
      "Epoch 477/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0066\n",
      "Epoch 478/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0078\n",
      "Epoch 479/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0091\n",
      "Epoch 480/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0138\n",
      "Epoch 481/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0066\n",
      "Epoch 482/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0081\n",
      "Epoch 483/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0119\n",
      "Epoch 484/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0185\n",
      "Epoch 485/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0100\n",
      "Epoch 486/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0138\n",
      "Epoch 487/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0238\n",
      "Epoch 488/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0122\n",
      "Epoch 489/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0110\n",
      "Epoch 490/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0150\n",
      "Epoch 491/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0041\n",
      "Epoch 492/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 9.3926e-04\n",
      "Epoch 493/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0022\n",
      "Epoch 494/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0050\n",
      "Epoch 495/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0053\n",
      "Epoch 496/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0028\n",
      "Epoch 497/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0047\n",
      "Epoch 498/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0128\n",
      "Epoch 499/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0185\n",
      "Epoch 500/500\n",
      "3194/3194 [==============================] - 39s 12ms/sample - loss: inf - accuracy: 0.0222\n"
     ]
    }
   ],
   "source": [
    "# train asr model\n",
    "history = pipeline.fit(train_dataset = train_data, batch_size=128, epochs=500)\n",
    "\n",
    "# history = pipeline.fit_iter(train_dataset = train_data, batch_size=32, epochs=3,iter_num=500,checkpoint=project_path+'checkpoints')\n",
    "# history = pipeline.fit_generator(train_dataset = train_data, batch_size=32, epochs=500)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "5WbeF-OWwhZB"
   },
   "outputs": [],
   "source": [
    "# save deepasr ctc pipeline\n",
    "pipeline.save(project_path+'checkpoints')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "o_psolNH4XFl"
   },
   "source": [
    "# 4. Model testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "EEgiUEkVc07E",
    "outputId": "fb6184a0-c0d9-4fe9-f445-37477d4661ff"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:No training configuration found in save file: the model was *not* compiled. Compile it manually.\n"
     ]
    }
   ],
   "source": [
    "# load saved ctc pipeline\n",
    "pipeline1 = asr.pipeline.load(project_path+'checkpoints')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 67
    },
    "colab_type": "code",
    "id": "xkPo_3SMtzHp",
    "outputId": "bcdc7fa7-6852-4aa8-99e3-6db3d8508ada"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Audio File: ./LibriSpeech/train-clean-100/27/124992/27-124992-0063.flac\n",
      "Audio Transcription: WENT THROUGH THE PLAINS BUT WHEN THEY CAME NEAR THE MOUNTAINS\n",
      "Trancript length: 61\n"
     ]
    }
   ],
   "source": [
    "# get testing audio and transcript from dataset\n",
    "index = np.random.randint(train_data.shape[0])\n",
    "data = train_data.iloc[index]\n",
    "test_file = data[0]\n",
    "test_transcript = data[1]\n",
    "# Audio file\n",
    "print(\"Audio File:\",test_file)\n",
    "# ground truth\n",
    "print(\"Audio Transcription:\", test_transcript)\n",
    "print(\"Transcript length:\",len(test_transcript))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 87
    },
    "colab_type": "code",
    "id": "moqXWTQVvdxC",
    "outputId": "0645ce28-1da9-447e-cc9d-93a4f57096c8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/keras/backend.py:5811: sparse_to_dense (from tensorflow.python.ops.sparse_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Create a `tf.sparse.SparseTensor` and use `tf.sparse.to_dense` instead.\n"
     ]
    }
   ],
   "source": [
    "# predict labels\n",
    "pred= pipeline1.predict(test_file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "oNvRyWq8weZs",
    "outputId": "88732c3b-5412-4a84-bb25-d517d4251a8c"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'WENT THROUGH THE PLAINS BUT WHEN THEY CAME NEAR THE MOUNTAINS'"
      ]
     },
     "execution_count": 15,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred[0].upper()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "AFMoK13mtR6V"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "machine_shape": "hm",
   "name": "DeepAsr-CTC_Pipeline.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}