BERT_BiLSTM_CRF_NER.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "BERT-BiLSTM-CRF-NER.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [
        "Gg2abfV6MdC-"
      ],
      "toc_visible": true,
      "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/thingumajig/colab-experiments/blob/master/BERT_BiLSTM_CRF_NER.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cBYIXlQrLd4L",
        "colab_type": "text"
      },
      "source": [
        "#Setup"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6_plzSmULk1s",
        "colab_type": "text"
      },
      "source": [
        "## clone from git"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DLCUQARCnTvN",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "e040a70e-1400-43cf-c1fd-0608b7324c86"
      },
      "source": [
        "!pip install --upgrade tensorflow-gpu\n",
        "!pip install seqeval\n",
        "!pip install bert-tensorflow\n",
        "\n",
        "!pip list | grep tensorflow\n",
        "!ps ax | grep python\n"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting tensorflow-gpu\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/76/04/43153bfdfcf6c9a4c38ecdb971ca9a75b9a791bb69a764d652c359aca504/tensorflow_gpu-1.14.0-cp36-cp36m-manylinux1_x86_64.whl (377.0MB)\n",
            "\u001b[K     |████████████████████████████████| 377.0MB 49kB/s \n",
            "\u001b[?25hRequirement already satisfied, skipping upgrade: gast>=0.2.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (0.2.2)\n",
            "Requirement already satisfied, skipping upgrade: termcolor>=1.1.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.1.0)\n",
            "Requirement already satisfied, skipping upgrade: tensorboard<1.15.0,>=1.14.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.14.0)\n",
            "Requirement already satisfied, skipping upgrade: protobuf>=3.6.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (3.7.1)\n",
            "Requirement already satisfied, skipping upgrade: wrapt>=1.11.1 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.11.2)\n",
            "Requirement already satisfied, skipping upgrade: google-pasta>=0.1.6 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (0.1.7)\n",
            "Requirement already satisfied, skipping upgrade: tensorflow-estimator<1.15.0rc0,>=1.14.0rc0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.14.0)\n",
            "Requirement already satisfied, skipping upgrade: wheel>=0.26 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (0.33.4)\n",
            "Requirement already satisfied, skipping upgrade: numpy<2.0,>=1.14.5 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.16.4)\n",
            "Requirement already satisfied, skipping upgrade: astor>=0.6.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (0.8.0)\n",
            "Requirement already satisfied, skipping upgrade: grpcio>=1.8.6 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.15.0)\n",
            "Requirement already satisfied, skipping upgrade: absl-py>=0.7.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (0.7.1)\n",
            "Requirement already satisfied, skipping upgrade: keras-applications>=1.0.6 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.0.8)\n",
            "Requirement already satisfied, skipping upgrade: six>=1.10.0 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.12.0)\n",
            "Requirement already satisfied, skipping upgrade: keras-preprocessing>=1.0.5 in /usr/local/lib/python3.6/dist-packages (from tensorflow-gpu) (1.1.0)\n",
            "Requirement already satisfied, skipping upgrade: markdown>=2.6.8 in /usr/local/lib/python3.6/dist-packages (from tensorboard<1.15.0,>=1.14.0->tensorflow-gpu) (3.1.1)\n",
            "Requirement already satisfied, skipping upgrade: werkzeug>=0.11.15 in /usr/local/lib/python3.6/dist-packages (from tensorboard<1.15.0,>=1.14.0->tensorflow-gpu) (0.15.4)\n",
            "Requirement already satisfied, skipping upgrade: setuptools>=41.0.0 in /usr/local/lib/python3.6/dist-packages (from tensorboard<1.15.0,>=1.14.0->tensorflow-gpu) (41.0.1)\n",
            "Requirement already satisfied, skipping upgrade: h5py in /usr/local/lib/python3.6/dist-packages (from keras-applications>=1.0.6->tensorflow-gpu) (2.8.0)\n",
            "Installing collected packages: tensorflow-gpu\n",
            "Successfully installed tensorflow-gpu-1.14.0\n",
            "Collecting seqeval\n",
            "  Downloading https://files.pythonhosted.org/packages/34/91/068aca8d60ce56dd9ba4506850e876aba5e66a6f2f29aa223224b50df0de/seqeval-0.0.12.tar.gz\n",
            "Requirement already satisfied: numpy>=1.14.0 in /usr/local/lib/python3.6/dist-packages (from seqeval) (1.16.4)\n",
            "Requirement already satisfied: Keras>=2.2.4 in /usr/local/lib/python3.6/dist-packages (from seqeval) (2.2.4)\n",
            "Requirement already satisfied: keras-applications>=1.0.6 in /usr/local/lib/python3.6/dist-packages (from Keras>=2.2.4->seqeval) (1.0.8)\n",
            "Requirement already satisfied: h5py in /usr/local/lib/python3.6/dist-packages (from Keras>=2.2.4->seqeval) (2.8.0)\n",
            "Requirement already satisfied: keras-preprocessing>=1.0.5 in /usr/local/lib/python3.6/dist-packages (from Keras>=2.2.4->seqeval) (1.1.0)\n",
            "Requirement already satisfied: pyyaml in /usr/local/lib/python3.6/dist-packages (from Keras>=2.2.4->seqeval) (3.13)\n",
            "Requirement already satisfied: six>=1.9.0 in /usr/local/lib/python3.6/dist-packages (from Keras>=2.2.4->seqeval) (1.12.0)\n",
            "Requirement already satisfied: scipy>=0.14 in /usr/local/lib/python3.6/dist-packages (from Keras>=2.2.4->seqeval) (1.3.0)\n",
            "Building wheels for collected packages: seqeval\n",
            "  Building wheel for seqeval (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Stored in directory: /root/.cache/pip/wheels/4f/32/0a/df3b340a82583566975377d65e724895b3fad101a3fb729f68\n",
            "Successfully built seqeval\n",
            "Installing collected packages: seqeval\n",
            "Successfully installed seqeval-0.0.12\n",
            "Collecting bert-tensorflow\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/a6/66/7eb4e8b6ea35b7cc54c322c816f976167a43019750279a8473d355800a93/bert_tensorflow-1.0.1-py2.py3-none-any.whl (67kB)\n",
            "\u001b[K     |████████████████████████████████| 71kB 5.3MB/s \n",
            "\u001b[?25hRequirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from bert-tensorflow) (1.12.0)\n",
            "Installing collected packages: bert-tensorflow\n",
            "Successfully installed bert-tensorflow-1.0.1\n",
            "bert-tensorflow          1.0.1                \n",
            "mesh-tensorflow          0.0.5                \n",
            "tensorflow               1.14.0               \n",
            "tensorflow-estimator     1.14.0               \n",
            "tensorflow-gpu           1.14.0               \n",
            "tensorflow-hub           0.5.0                \n",
            "tensorflow-metadata      0.13.0               \n",
            "tensorflow-probability   0.7.0                \n",
            "     24 ?        Sl     0:02 /usr/bin/python2 /usr/local/bin/jupyter-notebook --ip=\"172.28.0.2\" --port=9000 --FileContentsManager.root_dir=\"/\" --MappingKernelManager.root_dir=\"/content\"\n",
            "    132 ?        Ssl    0:07 /usr/bin/python3 -m ipykernel_launcher -f /root/.local/share/jupyter/runtime/kernel-8bf53b51-970a-467e-9b32-0bace890dc52.json\n",
            "    226 ?        S      0:00 /bin/bash -c ps ax | grep python\n",
            "    228 ?        S      0:00 grep python\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JetNC4E8ymyy",
        "colab_type": "code",
        "outputId": "64312419-3e6a-4a7f-8f3d-4d8608dce2e8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 104
        }
      },
      "source": [
        "\n",
        "import sys\n",
        "\n",
        "# bert_path = 'bert'\n",
        "# !test -d $bert_path || git clone https://github.com/google-research/bert $bert_path\n",
        "# if not bert_path in sys.path:\n",
        "#   sys.path.insert(0, bert_path)\n",
        "\n",
        "\n",
        "\n",
        "bert_ner_path = 'bert-ner'  \n",
        "!test -d $bert_ner_path || git clone https://github.com/dsindex/BERT-BiLSTM-CRF-NER $bert_ner_path\n",
        "if not bert_ner_path in sys.path:\n",
        "  sys.path.insert(0, bert_ner_path)\n",
        "\n",
        "#!cp /content/BERT-NER/metrics.py /content/metrics.py\n",
        "\n"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Cloning into 'bert-ner'...\n",
            "remote: Enumerating objects: 286, done.\u001b[K\n",
            "remote: Total 286 (delta 0), reused 0 (delta 0), pack-reused 286\u001b[K\n",
            "Receiving objects: 100% (286/286), 3.62 MiB | 5.67 MiB/s, done.\n",
            "Resolving deltas: 100% (170/170), done.\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H93ZJ86sHeeF",
        "colab_type": "code",
        "outputId": "6c9a28f0-aa41-41ff-e939-1e4c3df350dd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 347
        }
      },
      "source": [
        "!rm -rfv $bert_ner_path/bert"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "removed 'bert-ner/bert/requirements.txt'\n",
            "removed 'bert-ner/bert/run_squad.py'\n",
            "removed 'bert-ner/bert/optimization.py'\n",
            "removed 'bert-ner/bert/extract_features.py'\n",
            "removed 'bert-ner/bert/run_pretraining.py'\n",
            "removed 'bert-ner/bert/CONTRIBUTING.md'\n",
            "removed 'bert-ner/bert/tokenization_test.py'\n",
            "removed 'bert-ner/bert/sample_text.txt'\n",
            "removed 'bert-ner/bert/multilingual.md'\n",
            "removed 'bert-ner/bert/README.md'\n",
            "removed 'bert-ner/bert/modeling.py'\n",
            "removed 'bert-ner/bert/optimization_test.py'\n",
            "removed 'bert-ner/bert/LICENSE'\n",
            "removed 'bert-ner/bert/__init__.py'\n",
            "removed 'bert-ner/bert/run_classifier.py'\n",
            "removed 'bert-ner/bert/tokenization.py'\n",
            "removed 'bert-ner/bert/modeling_test.py'\n",
            "removed 'bert-ner/bert/create_pretraining_data.py'\n",
            "removed directory 'bert-ner/bert'\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Nl1oHKK1LvuA",
        "colab_type": "text"
      },
      "source": [
        "##Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Z0o6FSSvy2yW",
        "colab_type": "code",
        "outputId": "3036169c-92a9-45cc-99e7-c6880e32296f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329
        }
      },
      "source": [
        "#@title Select model:\n",
        "import os\n",
        "\n",
        "bert_model = \"https://storage.googleapis.com/bert_models/2019_05_30/wwm_cased_L-24_H-1024_A-16.zip\" #@param [\"https://storage.googleapis.com/bert_models/2019_05_30/wwm_cased_L-24_H-1024_A-16.zip\", \"https://storage.googleapis.com/bert_models/2018_11_23/multi_cased_L-12_H-768_A-12.zip\"]\n",
        "bert_model_zip = os.path.basename(bert_model)\n",
        "bert_model_dir = os.path.splitext(bert_model_zip)[0]\n",
        "\n",
        "!test -d $bert_model_dir || wget $bert_model\n",
        "\n",
        "\n",
        "!test -d $bert_model_dir || unzip /content/$bert_model_zip"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--2019-07-11 01:00:58--  https://storage.googleapis.com/bert_models/2019_05_30/wwm_cased_L-24_H-1024_A-16.zip\n",
            "Resolving storage.googleapis.com (storage.googleapis.com)... 66.102.1.128, 2a00:1450:400c:c0a::80\n",
            "Connecting to storage.googleapis.com (storage.googleapis.com)|66.102.1.128|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 1242589256 (1.2G) [application/zip]\n",
            "Saving to: ‘wwm_cased_L-24_H-1024_A-16.zip’\n",
            "\n",
            "wwm_cased_L-24_H-10 100%[===================>]   1.16G  72.4MB/s    in 19s     \n",
            "\n",
            "2019-07-11 01:01:17 (63.1 MB/s) - ‘wwm_cased_L-24_H-1024_A-16.zip’ saved [1242589256/1242589256]\n",
            "\n",
            "Archive:  /content/wwm_cased_L-24_H-1024_A-16.zip\n",
            "   creating: wwm_cased_L-24_H-1024_A-16/\n",
            "  inflating: wwm_cased_L-24_H-1024_A-16/bert_model.ckpt.meta  \n",
            "  inflating: wwm_cased_L-24_H-1024_A-16/bert_model.ckpt.data-00000-of-00001  \n",
            "  inflating: wwm_cased_L-24_H-1024_A-16/vocab.txt  \n",
            "  inflating: wwm_cased_L-24_H-1024_A-16/bert_model.ckpt.index  \n",
            "  inflating: wwm_cased_L-24_H-1024_A-16/bert_config.json  \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "TSVMaPtiIO_t",
        "colab_type": "code",
        "outputId": "c34366d5-b09a-4054-a181-fdd91941ee5d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 124
        }
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&scope=email%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdocs.test%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive.photos.readonly%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fpeopleapi.readonly&response_type=code\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XCGDQ0tbLVGR",
        "colab_type": "text"
      },
      "source": [
        "##init flags"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9KCSUfYEIjOP",
        "colab_type": "code",
        "outputId": "7a98dad0-9d5d-4f4d-d3d7-f4ec5125d286",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        }
      },
      "source": [
        "from __future__ import absolute_import\n",
        "from __future__ import division\n",
        "from __future__ import print_function\n",
        "\n",
        "import collections\n",
        "import os\n",
        "import json\n",
        "from bert import modeling\n",
        "from bert import optimization\n",
        "from bert import tokenization\n",
        "import tensorflow as tf\n",
        "import codecs\n",
        "from tensorflow.contrib.layers.python.layers import initializers\n",
        "from tensorflow.contrib import crf\n",
        "from tensorflow.contrib import rnn, cudnn_rnn\n",
        "\n",
        "from tensorflow.contrib.tpu import TPUEstimator, TPUConfig\n",
        "from tensorflow.python.ops import math_ops\n",
        "\n",
        "import tf_metrics\n",
        "import pickle\n",
        "\n",
        "os.environ['CUDA_VISIBLE_DEVICES'] = '0'\n",
        "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'\n",
        "\n",
        "flags = tf.flags\n",
        "\n",
        "FLAGS = flags.FLAGS\n",
        "\n",
        "\n",
        "\n",
        "# print(FLAGS.__dict__['__wrapped'])\n",
        "def del_all_flags(FLAGS, r = []):\n",
        "    flags_dict = FLAGS._flags()\n",
        "    keys_list = [keys for keys in flags_dict]\n",
        "    for keys in keys_list:\n",
        "        if keys in r:\n",
        "          FLAGS.__delattr__(keys)\n",
        " \n",
        "# allows rerun cell\n",
        "del_all_flags(FLAGS, r = ['bert_config_file','data_dir',\n",
        "                         \"task_name\", \"output_dir\",\"init_checkpoint\",\n",
        "                         \"do_lower_case\", \"do_train\", \"use_tpu\", \"do_predict\",\n",
        "                         \"use_crf\", \"train_batch_size\", \"eval_batch_size\",\n",
        "                         \"learning_rate\", \"num_train_epochs\", \"warmup_proportion\",\n",
        "                         \"bert_dropout_rate\", \"bilstm_dropout_rate\",\n",
        "                          \"save_checkpoints_steps\", \"save_summary_steps\",\n",
        "                         'data_config_path', \"num_tpu_cores\", \"master\",\n",
        "                          \"vocab_file\", \"keep_checkpoint_max\",                          \n",
        "                          'lstm_size', 'max_seq_length',\n",
        "                          'predict_batch_size', 'iterations_per_loop',\n",
        "                          \n",
        "                         ])\n",
        "train_model_path = '/content/drive/My Drive/ner_checkpoints'\n",
        "if not os.path.exists(train_model_path):\n",
        "    os.makedirs(train_model_path)\n",
        "\n",
        "flags.DEFINE_string(\"data_dir\", f'/content/{bert_ner_path}/NERdata',\n",
        "    \"The input datadir. ex) 'NERdata'\")\n",
        "\n",
        "flags.DEFINE_string(\"bert_config_file\", os.path.join('/content',bert_model_dir,'bert_config.json'),\n",
        "    \"The config json file corresponding to the pre-trained BERT model. ex) 'bert_config.json'\")\n",
        "\n",
        "flags.DEFINE_string(\"task_name\", 'NER', \"The name of the task to train. ex) 'NER'\")\n",
        "\n",
        "flags.DEFINE_string(\"output_dir\", train_model_path,\n",
        "    \"The output directory where the model checkpoints will be written. ex) 'output'\")\n",
        "\n",
        "flags.DEFINE_string(\"init_checkpoint\", f\"/content/{bert_model_dir}/bert_model.ckpt\",\n",
        "    \"Initial checkpoint (usually from a pre-trained BERT model).\")\n",
        "\n",
        "flags.DEFINE_bool(\"do_lower_case\", False,\n",
        "    \"Whether to lower case the input text.\")\n",
        "\n",
        "flags.DEFINE_integer(\"max_seq_length\", 128,\n",
        "    \"The maximum total input sequence length after WordPiece tokenization.\")\n",
        "\n",
        "flags.DEFINE_bool(\"do_train\", True,\n",
        "    \"Whether to run training.\")\n",
        "\n",
        "flags.DEFINE_bool(\"use_tpu\", False, \"Whether to use TPU or GPU/CPU.\")\n",
        "\n",
        "flags.DEFINE_bool(\"do_predict\", True, \"Whether to run the model in inference mode on the test set.\")\n",
        "\n",
        "flags.DEFINE_bool(\"use_crf\", True, \"Whether to use CRF decoding.\")\n",
        "\n",
        "flags.DEFINE_integer(\"train_batch_size\", 16, \"Total batch size for training.\")\n",
        "\n",
        "flags.DEFINE_integer(\"eval_batch_size\", 8, \"Total batch size for eval.\")\n",
        "\n",
        "flags.DEFINE_integer(\"predict_batch_size\", 8, \"Total batch size for predict.\")\n",
        "\n",
        "flags.DEFINE_float(\"learning_rate\", 5e-5, \"The initial learning rate for Adam.\")\n",
        "\n",
        "flags.DEFINE_float(\"num_train_epochs\", 3.0, \"Total number of training epochs to perform.\")\n",
        "\n",
        "flags.DEFINE_float(\n",
        "    \"warmup_proportion\", 0.1,\n",
        "    \"Proportion of training to perform linear learning rate warmup for. \"\n",
        "    \"E.g., 0.1 = 10% of training.\")\n",
        "\n",
        "flags.DEFINE_float(\"bert_dropout_rate\", 0.2,\n",
        "                   \"Proportion of dropout for bert embedding.\")\n",
        "\n",
        "flags.DEFINE_float(\"bilstm_dropout_rate\", 0.2,\n",
        "                   \"Proportion of dropout for bilstm.\")\n",
        "\n",
        "flags.DEFINE_integer(\"save_checkpoints_steps\", 1000,\n",
        "                     \"How often to save the model checkpoint.\")\n",
        "\n",
        "flags.DEFINE_integer(\"save_summary_steps\", 100,\n",
        "                     \"Save summaries every this many steps\")\n",
        "\n",
        "flags.DEFINE_integer(\"keep_checkpoint_max\", 20,\n",
        "                  \"The maximum number of recent checkpoint files to keep\")\n",
        "\n",
        "flags.DEFINE_integer(\"iterations_per_loop\", 1000,\n",
        "                     \"How many steps to make in each estimator call.\")\n",
        "\n",
        "flags.DEFINE_string(\"vocab_file\", f'/content/{bert_model_dir}/vocab.txt',\n",
        "                    \"The vocabulary file that the BERT model was trained on. ex) 'vocab.txt'\")\n",
        "\n",
        "flags.DEFINE_string(\"master\", None, \"[Optional] TensorFlow master URL.\")\n",
        "\n",
        "flags.DEFINE_integer(\n",
        "    \"num_tpu_cores\", 8,\n",
        "    \"Only used if `use_tpu` is True. Total number of TPU cores to use.\")\n",
        "\n",
        "flags.DEFINE_string('data_config_path', 'data.conf',\n",
        "                    'data config file, which save train and dev config')\n",
        "\n",
        "flags.DEFINE_integer('lstm_size', 128, 'size of lstm units')\n",
        "\n",
        "print(FLAGS.__dict__['__wrapped'])\n"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING: Logging before flag parsing goes to stderr.\n",
            "W0711 01:02:26.499118 140338823006080 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/bert/optimization.py:87: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
            "\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:\n",
            "  --bert_config_file: The config json file corresponding to the pre-trained BERT\n",
            "    model. ex) 'bert_config.json'\n",
            "    (default: '/content/wwm_cased_L-24_H-1024_A-16/bert_config.json')\n",
            "  --bert_dropout_rate: Proportion of dropout for bert embedding.\n",
            "    (default: '0.2')\n",
            "    (a number)\n",
            "  --bilstm_dropout_rate: Proportion of dropout for bilstm.\n",
            "    (default: '0.2')\n",
            "    (a number)\n",
            "  --data_config_path: data config file, which save train and dev config\n",
            "    (default: 'data.conf')\n",
            "  --data_dir: The input datadir. ex) 'NERdata'\n",
            "    (default: '/content/bert-ner/NERdata')\n",
            "  --[no]do_lower_case: Whether to lower case the input text.\n",
            "    (default: 'false')\n",
            "  --[no]do_predict: Whether to run the model in inference mode on the test set.\n",
            "    (default: 'true')\n",
            "  --[no]do_train: Whether to run training.\n",
            "    (default: 'true')\n",
            "  --eval_batch_size: Total batch size for eval.\n",
            "    (default: '8')\n",
            "    (an integer)\n",
            "  --init_checkpoint: Initial checkpoint (usually from a pre-trained BERT model).\n",
            "    (default: '/content/wwm_cased_L-24_H-1024_A-16/bert_model.ckpt')\n",
            "  --iterations_per_loop: How many steps to make in each estimator call.\n",
            "    (default: '1000')\n",
            "    (an integer)\n",
            "  --keep_checkpoint_max: The maximum number of recent checkpoint files to keep\n",
            "    (default: '20')\n",
            "    (an integer)\n",
            "  --learning_rate: The initial learning rate for Adam.\n",
            "    (default: '5e-05')\n",
            "    (a number)\n",
            "  --lstm_size: size of lstm units\n",
            "    (default: '128')\n",
            "    (an integer)\n",
            "  --master: [Optional] TensorFlow master URL.\n",
            "  --max_seq_length: The maximum total input sequence length after WordPiece\n",
            "    tokenization.\n",
            "    (default: '128')\n",
            "    (an integer)\n",
            "  --num_tpu_cores: Only used if `use_tpu` is True. Total number of TPU cores to\n",
            "    use.\n",
            "    (default: '8')\n",
            "    (an integer)\n",
            "  --num_train_epochs: Total number of training epochs to perform.\n",
            "    (default: '3.0')\n",
            "    (a number)\n",
            "  --output_dir: The output directory where the model checkpoints will be\n",
            "    written. ex) 'output'\n",
            "    (default: '/content/drive/My Drive/ner_checkpoints')\n",
            "  --predict_batch_size: Total batch size for predict.\n",
            "    (default: '8')\n",
            "    (an integer)\n",
            "  --save_checkpoints_steps: How often to save the model checkpoint.\n",
            "    (default: '1000')\n",
            "    (an integer)\n",
            "  --save_summary_steps: Save summaries every this many steps\n",
            "    (default: '100')\n",
            "    (an integer)\n",
            "  --task_name: The name of the task to train. ex) 'NER'\n",
            "    (default: 'NER')\n",
            "  --train_batch_size: Total batch size for training.\n",
            "    (default: '16')\n",
            "    (an integer)\n",
            "  --[no]use_crf: Whether to use CRF decoding.\n",
            "    (default: 'true')\n",
            "  --[no]use_tpu: Whether to use TPU or GPU/CPU.\n",
            "    (default: 'false')\n",
            "  --vocab_file: The vocabulary file that the BERT model was trained on. ex)\n",
            "    'vocab.txt'\n",
            "    (default: '/content/wwm_cased_L-24_H-1024_A-16/vocab.txt')\n",
            "  --warmup_proportion: Proportion of training to perform linear learning rate\n",
            "    warmup for. E.g., 0.1 = 10% of training.\n",
            "    (default: '0.1')\n",
            "    (a number)\n",
            "\n",
            "absl.app:\n",
            "  --[no]only_check_args: Set to true to validate args and exit.\n",
            "    (default: 'false')\n",
            "  --[no]pdb_post_mortem: Set to true to handle uncaught exceptions with PDB post\n",
            "    mortem.\n",
            "    (default: 'false')\n",
            "  --profile_file: Dump profile information to a file (for python -m pstats).\n",
            "    Implies --run_with_profiling.\n",
            "  --[no]run_with_pdb: Set to true for PDB debug mode\n",
            "    (default: 'false')\n",
            "  --[no]run_with_profiling: Set to true for profiling the script. Execution will\n",
            "    be slower, and the output format might change over time.\n",
            "    (default: 'false')\n",
            "  --[no]use_cprofile_for_profiling: Use cProfile instead of the profile module\n",
            "    for profiling. This has no effect unless --run_with_profiling is set.\n",
            "    (default: 'true')\n",
            "\n",
            "absl.logging:\n",
            "  --[no]alsologtostderr: also log to stderr?\n",
            "    (default: 'false')\n",
            "  --log_dir: directory to write logfiles into\n",
            "    (default: '')\n",
            "  --[no]logtostderr: Should only log to stderr?\n",
            "    (default: 'false')\n",
            "  --[no]showprefixforinfo: If False, do not prepend prefix to info messages when\n",
            "    it's logged to stderr, --verbosity is set to INFO level, and python logging\n",
            "    is used.\n",
            "    (default: 'true')\n",
            "  --stderrthreshold: log messages at this level, or more severe, to stderr in\n",
            "    addition to the logfile.  Possible values are 'debug', 'info', 'warning',\n",
            "    'error', and 'fatal'.  Obsoletes --alsologtostderr. Using --alsologtostderr\n",
            "    cancels the effect of this flag. Please also note that this flag is subject\n",
            "    to --verbosity and requires logfile not be stderr.\n",
            "    (default: 'fatal')\n",
            "  -v,--verbosity: Logging verbosity level. Messages logged at this level or\n",
            "    lower will be included. Set to 1 for debug logging. If the flag was not set\n",
            "    or supplied, the value will be changed from the default of -1 (warning) to 0\n",
            "    (info) after flags are parsed.\n",
            "    (default: '-1')\n",
            "    (an integer)\n",
            "\n",
            "absl.testing.absltest:\n",
            "  --test_random_seed: Random seed for testing. Some test frameworks may change\n",
            "    the default value of this flag between runs, so it is not appropriate for\n",
            "    seeding probabilistic tests.\n",
            "    (default: '301')\n",
            "    (an integer)\n",
            "  --test_randomize_ordering_seed: If positive, use this as a seed to randomize\n",
            "    the execution order for test cases. If \"random\", pick a random seed to use.\n",
            "    If 0 or not set, do not randomize test case execution order. This flag also\n",
            "    overrides the TEST_RANDOMIZE_ORDERING_SEED environment variable.\n",
            "  --test_srcdir: Root of directory tree where source files live\n",
            "    (default: '')\n",
            "  --test_tmpdir: Directory for temporary testing files\n",
            "    (default: '/tmp/absl_testing')\n",
            "  --xml_output_file: File to store XML test results\n",
            "    (default: '')\n",
            "\n",
            "tensorflow.python.ops.parallel_for.pfor:\n",
            "  --[no]op_conversion_fallback_to_while_loop: If true, falls back to using a\n",
            "    while loop for ops for which a converter is not defined.\n",
            "    (default: 'false')\n",
            "\n",
            "absl.flags:\n",
            "  --flagfile: Insert flag definitions from the given file into the command line.\n",
            "    (default: '')\n",
            "  --undefok: comma-separated list of flag names that it is okay to specify on\n",
            "    the command line even if the program does not define a flag with that name.\n",
            "    IMPORTANT: flags in this list that have arguments MUST use the --flag=value\n",
            "    format.\n",
            "    (default: '')\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Wh8zo1UXnP6f",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-fsoHJgHL5Cc",
        "colab_type": "text"
      },
      "source": [
        "## input example"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AmFo2JDdL2T1",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class InputExample(object):\n",
        "    \"\"\"A single training/test example for simple sequence classification.\"\"\"\n",
        "\n",
        "    def __init__(self, guid, text, pos, chunk, label=None):\n",
        "        \"\"\"Constructs a InputExample.\n",
        "\n",
        "        Args:\n",
        "          guid: Unique id for the example.\n",
        "          text: string. The untokenized text of the first sequence. For single\n",
        "            sequence tasks, only this sequence must be specified.\n",
        "          pos: string. The pos(part of speech) tag of the example.\n",
        "          chunk: string. The chunk tag of the example.\n",
        "          label: (Optional) string. The label of the example. This should be\n",
        "            specified for train and dev examples, but not for test examples.\n",
        "        \"\"\"\n",
        "        self.guid  = guid\n",
        "        self.text  = text\n",
        "        self.pos   = pos\n",
        "        self.chunk = chunk\n",
        "        self.label = label\n",
        "\n",
        "        # untokenized data\n",
        "        self.tokens = text.split()\n",
        "        self.poss   = pos.split()\n",
        "        self.chunks = chunk.split()\n",
        "        self.labels = []\n",
        "        if label: self.labels = label.split()\n",
        "\n",
        "        # tokenized data\n",
        "        self.tokenized_tokens = []\n",
        "        self.tokenized_poss   = []\n",
        "        self.tokenized_chunks = []\n",
        "        self.tokenized_labels = []\n",
        "\n",
        "\n",
        "class InputFeatures(object):\n",
        "    \"\"\"A single set of features of data.\"\"\"\n",
        "\n",
        "    def __init__(self, input_ids, input_mask, segment_ids, label_ids, ):\n",
        "        self.input_ids = input_ids\n",
        "        self.input_mask = input_mask\n",
        "        self.segment_ids = segment_ids\n",
        "        self.label_ids = label_ids\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_Yl8f7uXMIyN",
        "colab_type": "text"
      },
      "source": [
        "## NER  Processor"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6tf4RIleMNyw",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class NerProcessor(object):\n",
        "    \"\"\"A class for data converters for sequence classification data sets.\"\"\"\n",
        "\n",
        "    def get_train_examples(self, data_dir):\n",
        "        \"\"\"Gets a collection of `InputExample`s for the train set.\"\"\"\n",
        "        return self._create_example(\n",
        "            self._read_data(os.path.join(data_dir, \"train.txt\")), \"train\"\n",
        "        )\n",
        "\n",
        "    def get_dev_examples(self, data_dir):\n",
        "        \"\"\"Gets a collection of `InputExample`s for the dev set.\"\"\"\n",
        "        return self._create_example(\n",
        "            self._read_data(os.path.join(data_dir, \"dev.txt\")), \"dev\"\n",
        "        )\n",
        "\n",
        "    def get_test_examples(self, data_dir):\n",
        "        \"\"\"Gets a collection of `InputExample`s for the test set.\"\"\"\n",
        "        return self._create_example(\n",
        "            self._read_data(os.path.join(data_dir, \"test.txt\")), \"test\")\n",
        "\n",
        "    def get_labels(self):\n",
        "        \"\"\"Gets the list of labels for this data set.\"\"\"\n",
        "        '''\n",
        "        return [\"O\", \"B-PER\", \"I-PER\", \"B-ORG\", \"I-ORG\", \"B-LOC\", \"I-LOC\", \"B-MISC\", \"I-MISC\", \"X\", \"[CLS]\", \"[SEP]\"]\n",
        "        '''\n",
        "        return [\"O\", \"B-PER\", \"I-PER\", \"B-ORG\", \"I-ORG\", \"B-LOC\", \"I-LOC\", \"B-MISC\", \"I-MISC\", \"X\"]\n",
        "\n",
        "    def _create_example(self, lines, set_type):\n",
        "        examples = []\n",
        "        for (i, line) in enumerate(lines): # line = (w, p, c, l)\n",
        "            guid  = \"%s-%s\" % (set_type, i)\n",
        "            text  = tokenization.convert_to_unicode(line[0])\n",
        "            pos   = tokenization.convert_to_unicode(line[1])\n",
        "            chunk = tokenization.convert_to_unicode(line[2])\n",
        "            label = tokenization.convert_to_unicode(line[3])\n",
        "            examples.append(InputExample(guid=guid, text=text, pos=pos, chunk=chunk, label=label))\n",
        "        return examples\n",
        "\n",
        "    @classmethod\n",
        "    def _read_data(cls, input_file):\n",
        "        \"\"\"Reads a BIO data.\"\"\"\n",
        "        with codecs.open(input_file, 'r', encoding='utf-8') as f:\n",
        "            lines  = []\n",
        "            words  = []\n",
        "            poss   = []\n",
        "            chunks = []\n",
        "            labels = []\n",
        "            for line in f:\n",
        "                contents = line.strip()\n",
        "                if contents.startswith(\"-DOCSTART-\"):\n",
        "                    continue\n",
        "                if len(contents) == 0: # newline\n",
        "                    if len(words) == 0: continue\n",
        "                    assert(len(words) == len(poss))\n",
        "                    assert(len(poss) == len(chunks))\n",
        "                    assert(len(chunks) == len(labels))\n",
        "                    w = ' '.join(words)\n",
        "                    p = ' '.join(poss)\n",
        "                    c = ' '.join(chunks)\n",
        "                    l = ' '.join(labels)\n",
        "                    lines.append([w, p, c, l])\n",
        "                    words  = []\n",
        "                    poss   = []\n",
        "                    chunks = []\n",
        "                    labels = []\n",
        "                    continue\n",
        "                tokens = line.strip().split(' ')\n",
        "                assert(len(tokens) == 4)\n",
        "                word  = tokens[0]\n",
        "                pos   = tokens[1]\n",
        "                chunk = tokens[2]\n",
        "                label = tokens[-1]\n",
        "                words.append(word)\n",
        "                poss.append(pos)\n",
        "                chunks.append(chunk)\n",
        "                labels.append(label)\n",
        "            return lines\n",
        "\n",
        "    #@classmethod\n",
        "    #def _read_text_data(cls, input_file):\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Gg2abfV6MdC-",
        "colab_type": "text"
      },
      "source": [
        "## Create model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l4sfG1-aMhRH",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def convert_single_example_to_feature(ex_index, example, label_map, max_seq_length, tokenizer, mode):\n",
        "    textlist  = example.tokens\n",
        "    poslist   = example.poss\n",
        "    chunklist = example.chunks\n",
        "    labellist = example.labels\n",
        "    tokens = []\n",
        "    poss   = []\n",
        "    chunks = []\n",
        "    labels = []\n",
        "    for i, word in enumerate(textlist):\n",
        "        token = tokenizer.tokenize(word)\n",
        "        tokens.extend(token)\n",
        "        pos_1   = poslist[i]\n",
        "        chunk_1 = chunklist[i]\n",
        "        label_1 = labellist[i]\n",
        "        for m in range(len(token)):\n",
        "            if m == 0:\n",
        "                poss.append(pos_1)\n",
        "                chunks.append(chunk_1)\n",
        "                labels.append(label_1)\n",
        "            else:\n",
        "                poss.append(\"X\")\n",
        "                chunks.append(\"X\")\n",
        "                labels.append(\"X\")\n",
        "    # tokens = tokenizer.tokenize(example.text)\n",
        "    if len(tokens) >= max_seq_length - 1:\n",
        "        tokens = tokens[0:(max_seq_length - 2)]\n",
        "        poss   = poss[0:(max_seq_length - 2)]\n",
        "        chunks = chunks[0:(max_seq_length - 2)]\n",
        "        labels = labels[0:(max_seq_length - 2)]\n",
        "\n",
        "    # save tokens, poss, chunks, labels back to example\n",
        "    example.tokenized_tokens = tokens\n",
        "    example.tokenized_poss   = poss\n",
        "    example.tokenized_chunks = chunks\n",
        "    example.tokenized_labels = labels\n",
        "\n",
        "    ntokens = []\n",
        "    segment_ids = []\n",
        "    label_ids = []\n",
        "    ntokens.append(\"[CLS]\")\n",
        "    segment_ids.append(0)\n",
        "    #label_ids.append(label_map[\"[CLS]\"])\n",
        "    label_ids.append(0)\n",
        "    for i, token in enumerate(tokens):\n",
        "        ntokens.append(token)\n",
        "        segment_ids.append(0)\n",
        "        label_ids.append(label_map[labels[i]])\n",
        "    ntokens.append(\"[SEP]\")\n",
        "    segment_ids.append(0)\n",
        "    #label_ids.append(label_map[\"[SEP]\"])\n",
        "    label_ids.append(0)\n",
        "    input_ids = tokenizer.convert_tokens_to_ids(ntokens)\n",
        "    input_mask = [1] * len(input_ids)\n",
        "    # padding\n",
        "    while len(input_ids) < max_seq_length:\n",
        "        input_ids.append(0)\n",
        "        input_mask.append(0)\n",
        "        segment_ids.append(0)\n",
        "        # we don't concerned about it!\n",
        "        label_ids.append(0)\n",
        "        ntokens.append(\"**NULL**\")\n",
        "    assert len(input_ids) == max_seq_length\n",
        "    assert len(input_mask) == max_seq_length\n",
        "    assert len(segment_ids) == max_seq_length\n",
        "    assert len(label_ids) == max_seq_length\n",
        "\n",
        "    if ex_index < 5:\n",
        "        tf.logging.info(\"*** Example ***\")\n",
        "        tf.logging.info(\"guid: %s\" % (example.guid))\n",
        "        tf.logging.info(\"tokens: %s\" % \" \".join(\n",
        "            [tokenization.printable_text(x) for x in tokens]))\n",
        "        tf.logging.info(\"input_ids: %s\" % \" \".join([str(x) for x in input_ids]))\n",
        "        tf.logging.info(\"input_mask: %s\" % \" \".join([str(x) for x in input_mask]))\n",
        "        tf.logging.info(\"segment_ids: %s\" % \" \".join([str(x) for x in segment_ids]))\n",
        "        tf.logging.info(\"label_ids: %s\" % \" \".join([str(x) for x in label_ids]))\n",
        "\n",
        "    feature = InputFeatures(\n",
        "        input_ids=input_ids,\n",
        "        input_mask=input_mask,\n",
        "        segment_ids=segment_ids,\n",
        "        label_ids=label_ids,\n",
        "    )\n",
        "    return feature\n",
        "\n",
        "\n",
        "def convert_feature_to_tf_example(feature):\n",
        "    def create_int_feature(values):\n",
        "        f = tf.train.Feature(int64_list=tf.train.Int64List(value=list(values)))\n",
        "        return f\n",
        "\n",
        "    features = collections.OrderedDict()\n",
        "    features[\"input_ids\"] = create_int_feature(feature.input_ids)\n",
        "    features[\"input_mask\"] = create_int_feature(feature.input_mask)\n",
        "    features[\"segment_ids\"] = create_int_feature(feature.segment_ids)\n",
        "    features[\"label_ids\"] = create_int_feature(feature.label_ids)\n",
        "    tf_example = tf.train.Example(features=tf.train.Features(feature=features))\n",
        "    return tf_example\n",
        "\n",
        "\n",
        "def filed_based_convert_examples_to_features(examples, label_list, max_seq_length, tokenizer, output_file, mode=None):\n",
        "    # build labe2id.pkl\n",
        "    label_map = {}\n",
        "    for (i, label) in enumerate(label_list, 1): # 0 index for '0' padding\n",
        "        label_map[label] = i\n",
        "    with codecs.open(os.path.join(FLAGS.output_dir,'label2id.pkl'), 'wb') as w:\n",
        "        pickle.dump(label_map, w)\n",
        "\n",
        "    # convert examples => features => tf_examples(tf.train.Example) => tf records(TFRecord, file)\n",
        "    writer = tf.python_io.TFRecordWriter(output_file)\n",
        "    for (ex_index, example) in enumerate(examples):\n",
        "        if ex_index % 5000 == 0:\n",
        "            tf.logging.info(\"Writing example %d of %d\" % (ex_index, len(examples)))\n",
        "        feature = convert_single_example_to_feature(ex_index, example, label_map, max_seq_length, tokenizer, mode)\n",
        "        tf_example = convert_feature_to_tf_example(feature)\n",
        "        writer.write(tf_example.SerializeToString())\n",
        "\n",
        "\n",
        "def file_based_input_fn_builder(input_file, seq_length, is_training, drop_remainder):\n",
        "    name_to_features = {\n",
        "        \"input_ids\": tf.FixedLenFeature([seq_length], tf.int64),\n",
        "        \"input_mask\": tf.FixedLenFeature([seq_length], tf.int64),\n",
        "        \"segment_ids\": tf.FixedLenFeature([seq_length], tf.int64),\n",
        "        \"label_ids\": tf.FixedLenFeature([seq_length], tf.int64),\n",
        "    }\n",
        "\n",
        "    def _decode_record(record, name_to_features):\n",
        "        example = tf.parse_single_example(record, name_to_features)\n",
        "        for name in list(example.keys()):\n",
        "            t = example[name]\n",
        "            if t.dtype == tf.int64:\n",
        "                t = tf.to_int32(t)\n",
        "            example[name] = t\n",
        "        return example\n",
        "\n",
        "    def input_fn(params):\n",
        "        batch_size = params[\"batch_size\"]\n",
        "        d = tf.data.TFRecordDataset(input_file)\n",
        "        if is_training:\n",
        "            d = d.repeat()\n",
        "            d = d.shuffle(buffer_size=100)\n",
        "        d = d.apply(tf.contrib.data.map_and_batch(\n",
        "            lambda record: _decode_record(record, name_to_features),\n",
        "            batch_size=batch_size,\n",
        "            drop_remainder=drop_remainder\n",
        "        ))\n",
        "        return d\n",
        "\n",
        "    return input_fn\n",
        "\n",
        "\n",
        "def create_model(bert_config, is_training, input_ids, input_mask,\n",
        "                 segment_ids, labels, num_labels, use_one_hot_embeddings):\n",
        "\n",
        "    model = modeling.BertModel(\n",
        "        config=bert_config,\n",
        "        is_training=is_training,\n",
        "        input_ids=input_ids,\n",
        "        input_mask=input_mask,\n",
        "        token_type_ids=segment_ids,\n",
        "        use_one_hot_embeddings=use_one_hot_embeddings\n",
        "    )\n",
        "    embedding = model.get_sequence_output() # (batch_size, seq_length, embedding_size)\n",
        "    '''\n",
        "    embedding_1 = model.get_all_encoder_layers()[-2]\n",
        "    embedding_2 = model.get_all_encoder_layers()[-1]\n",
        "    embedding = tf.concat([embedding_1, embedding_2], axis=-1)\n",
        "    '''\n",
        "    if is_training:\n",
        "        # dropout embedding\n",
        "        embedding = tf.layers.dropout(embedding, rate=FLAGS.bert_dropout_rate, training=is_training)\n",
        "    embedding_size = embedding.shape[-1].value # embedding_size\n",
        "    seq_length = embedding.shape[1].value\n",
        "\n",
        "    used = tf.sign(tf.abs(input_ids))\n",
        "    lengths = tf.reduce_sum(used, reduction_indices=1)  # (batch_size)\n",
        "    print('seq_length', seq_length)\n",
        "    print('lengths', lengths)\n",
        "\n",
        "    def bi_lstm_fused(inputs, lengths, rnn_size, is_training, dropout_rate=0.5, scope='bi-lstm-fused'):\n",
        "        with tf.variable_scope(scope):\n",
        "            t = tf.transpose(inputs, perm=[1, 0, 2])  # Need time-major\n",
        "            lstm_cell_fw = tf.contrib.rnn.LSTMBlockFusedCell(rnn_size)\n",
        "            lstm_cell_bw = tf.contrib.rnn.LSTMBlockFusedCell(rnn_size)\n",
        "            lstm_cell_bw = tf.contrib.rnn.TimeReversedFusedRNN(lstm_cell_bw)\n",
        "            output_fw, _ = lstm_cell_fw(t, dtype=tf.float32, sequence_length=lengths)\n",
        "            output_bw, _ = lstm_cell_bw(t, dtype=tf.float32, sequence_length=lengths)\n",
        "            outputs = tf.concat([output_fw, output_bw], axis=-1)\n",
        "            outputs = tf.transpose(outputs, perm=[1, 0, 2])\n",
        "            return tf.layers.dropout(outputs, rate=dropout_rate, training=is_training)\n",
        "\n",
        "    def lstm_layer(inputs, lengths, is_training):\n",
        "        rnn_output = tf.identity(inputs)\n",
        "        for i in range(2):\n",
        "            scope = 'bi-lstm-fused-%s' % i\n",
        "            rnn_output = bi_lstm_fused(rnn_output,\n",
        "                                       lengths,\n",
        "                                       rnn_size=FLAGS.lstm_size,\n",
        "                                       is_training=is_training,\n",
        "                                       dropout_rate=FLAGS.bilstm_dropout_rate,\n",
        "                                       scope=scope)  # (batch_size, seq_length, 2*rnn_size)\n",
        "        return rnn_output\n",
        "\n",
        "    def project_layer(inputs, out_dim, seq_length, scope='project'):\n",
        "        with tf.variable_scope(scope):\n",
        "            in_dim = inputs.get_shape().as_list()[-1]\n",
        "            weight = tf.get_variable('W', shape=[in_dim, out_dim],\n",
        "                                     dtype=tf.float32, initializer=initializers.xavier_initializer())\n",
        "            bias = tf.get_variable('b', shape=[out_dim], dtype=tf.float32,\n",
        "                                    initializer=tf.zeros_initializer())\n",
        "            t_output = tf.reshape(inputs, [-1, in_dim])            # (batch_size*seq_length, in_dim)\n",
        "            output = tf.matmul(t_output, weight) + bias            # (batch_size*seq_length, out_dim)\n",
        "            output = tf.reshape(output, [-1, seq_length, out_dim]) # (batch_size, seq_length, out_dim)\n",
        "            return output\n",
        "\n",
        "    def loss_layer(logits, labels, num_labels, lengths, input_mask):\n",
        "        trans = tf.get_variable(\n",
        "            \"transitions\",\n",
        "            shape=[num_labels, num_labels],\n",
        "            initializer=initializers.xavier_initializer())\n",
        "        if FLAGS.use_crf:\n",
        "            with tf.variable_scope(\"crf-loss\"):\n",
        "                log_likelihood, trans = tf.contrib.crf.crf_log_likelihood(\n",
        "                    inputs=logits,\n",
        "                    tag_indices=labels,\n",
        "                    transition_params=trans,\n",
        "                    sequence_lengths=lengths)\n",
        "                per_example_loss = -log_likelihood\n",
        "                loss = tf.reduce_mean(per_example_loss)\n",
        "                return loss, per_example_loss, trans\n",
        "        else:\n",
        "            labels_one_hot = tf.one_hot(labels, num_labels)\n",
        "            cross_entropy = labels_one_hot * tf.log(tf.nn.softmax(logits))\n",
        "            cross_entropy = -tf.reduce_sum(cross_entropy, reduction_indices=2)\n",
        "            cross_entropy *= tf.to_float(input_mask)\n",
        "            cross_entropy = tf.reduce_sum(cross_entropy, reduction_indices=1)\n",
        "            cross_entropy /= tf.cast(lengths, tf.float32)\n",
        "            per_example_loss = cross_entropy\n",
        "            loss = tf.reduce_mean(per_example_loss)\n",
        "            return loss, per_example_loss, trans\n",
        "    '''    \n",
        "    # 1\n",
        "    logits = project_layer(embedding, num_labels, seq_length, scope='project')\n",
        "    '''\n",
        "    '''\n",
        "    # 2\n",
        "    lstm_outputs = lstm_layer(embedding, lengths, is_training)\n",
        "    p1 = project_layer(lstm_outputs, FLAGS.lstm_size, seq_length, scope='project-1')\n",
        "    p2 = project_layer(p1, num_labels, seq_length, scope='project-2')\n",
        "    logits = p2\n",
        "    '''\n",
        "    # 3\n",
        "    lstm_outputs = lstm_layer(embedding, lengths, is_training)\n",
        "    logits = project_layer(lstm_outputs, num_labels, seq_length, scope='project')\n",
        "    loss, per_example_loss, trans = loss_layer(logits, labels, num_labels, lengths, input_mask)\n",
        "    if FLAGS.use_crf:\n",
        "        pred_ids, _ = crf.crf_decode(potentials=logits, transition_params=trans, sequence_length=lengths)\n",
        "    else:\n",
        "        probabilities = tf.nn.softmax(logits, axis=-1)\n",
        "        pred_ids = tf.argmax(probabilities,axis=-1)\n",
        "\n",
        "    # masking for confirmation\n",
        "    pred_ids *= input_mask\n",
        "\n",
        "    print('#' * 20)\n",
        "    print('shape of output_layer:', embedding.shape)\n",
        "    print('embedding_size:%d' % embedding_size)\n",
        "    print('seq_length:%d' % seq_length)\n",
        "    print('shape of logit', logits.shape)\n",
        "    print('shape of loss', loss.shape)\n",
        "    print('shape of per_example_loss', per_example_loss.shape)\n",
        "    print('num labels:%d' % num_labels)\n",
        "    print('#' * 20)\n",
        "    return (loss, per_example_loss, logits, trans, pred_ids)\n",
        "\n",
        "\n",
        "def model_fn_builder(bert_config, num_labels, init_checkpoint, learning_rate,\n",
        "                     num_train_steps, num_warmup_steps, use_tpu,\n",
        "                     use_one_hot_embeddings):\n",
        "\n",
        "    def model_fn(features, labels, mode, params):\n",
        "        tf.logging.info(\"*** Features ***\")\n",
        "        for name in sorted(features.keys()):\n",
        "            tf.logging.info(\"  name = %s, shape = %s\" % (name, features[name].shape))\n",
        "        input_ids = features[\"input_ids\"]\n",
        "        input_mask = features[\"input_mask\"]\n",
        "        segment_ids = features[\"segment_ids\"]\n",
        "        label_ids = features[\"label_ids\"]\n",
        "\n",
        "        print('shape of input_ids', input_ids.shape)\n",
        "        print('shape of label_ids', label_ids.shape)\n",
        "        # label_mask = features[\"label_mask\"]\n",
        "        is_training = (mode == tf.estimator.ModeKeys.TRAIN)\n",
        "\n",
        "        (total_loss, per_example_loss, logits, trans, pred_ids) = create_model(\n",
        "            bert_config, is_training, input_ids, input_mask, segment_ids, label_ids,\n",
        "            num_labels, use_one_hot_embeddings)\n",
        "\n",
        "        print('shape of pred_ids', pred_ids.shape)\n",
        "\n",
        "        global_step = tf.train.get_or_create_global_step()\n",
        "        # add summary\n",
        "        tf.summary.scalar('loss', total_loss)\n",
        "\n",
        "        tvars = tf.trainable_variables()\n",
        "        scaffold_fn = None\n",
        "        if init_checkpoint and is_training:\n",
        "            (assignment_map, initialized_variable_names) = modeling.get_assignment_map_from_checkpoint(tvars,\n",
        "                                                                                                       init_checkpoint)\n",
        "            if use_tpu:\n",
        "                def tpu_scaffold():\n",
        "                    tf.train.init_from_checkpoint(init_checkpoint, assignment_map)\n",
        "                    return tf.train.Scaffold()\n",
        "                scaffold_fn = tpu_scaffold\n",
        "            else:\n",
        "                tf.train.init_from_checkpoint(init_checkpoint, assignment_map)\n",
        "            tf.logging.info(\"**** Trainable Variables ****\")\n",
        "            for var in tvars:\n",
        "                init_string = \"\"\n",
        "                if var.name in initialized_variable_names:\n",
        "                    init_string = \", *INIT_FROM_CKPT*\"\n",
        "                tf.logging.info(\"  name = %s, shape = %s%s\", var.name, var.shape,\n",
        "                                init_string)\n",
        "\n",
        "        output_spec = None\n",
        "        if mode == tf.estimator.ModeKeys.PREDICT:\n",
        "            output_spec = tf.contrib.tpu.TPUEstimatorSpec(\n",
        "                mode=mode,\n",
        "                predictions=pred_ids,\n",
        "                scaffold_fn=scaffold_fn\n",
        "            )\n",
        "        else:\n",
        "            if mode == tf.estimator.ModeKeys.TRAIN:\n",
        "                '''\n",
        "                train_op = optimization.create_optimizer(\n",
        "                    total_loss, learning_rate, num_train_steps, num_warmup_steps, use_tpu)\n",
        "                '''\n",
        "                lr = tf.train.exponential_decay(learning_rate, global_step, 5000, 0.9, staircase=True)\n",
        "                optimizer = tf.train.AdamOptimizer(lr)\n",
        "                grads, _ = tf.clip_by_global_norm(tf.gradients(total_loss, tvars), 1.5)\n",
        "                train_op = optimizer.apply_gradients(zip(grads, tvars), global_step=global_step)\n",
        "                logging_hook = tf.train.LoggingTensorHook({\"batch_loss\" : total_loss}, every_n_iter=10)\n",
        "                output_spec = tf.contrib.tpu.TPUEstimatorSpec(\n",
        "                    mode=mode,\n",
        "                    loss=total_loss,\n",
        "                    train_op=train_op,\n",
        "                    training_hooks = [logging_hook],\n",
        "                    scaffold_fn=scaffold_fn)\n",
        "            else: # mode == tf.estimator.ModeKeys.EVAL:\n",
        "                def metric_fn(label_ids, pred_ids, per_example_loss, input_mask):\n",
        "                    # ['<pad>'] + [\"O\", \"B-PER\", \"I-PER\", \"B-ORG\", \"I-ORG\", \"B-LOC\", \"I-LOC\", \"B-MISC\", \"I-MISC\", \"X\"]\n",
        "                    indices = [2, 3, 4, 5, 6, 7, 8, 9]\n",
        "                    precision = tf_metrics.precision(label_ids, pred_ids, num_labels, indices, input_mask)\n",
        "                    recall = tf_metrics.recall(label_ids, pred_ids, num_labels, indices, input_mask)\n",
        "                    f = tf_metrics.f1(label_ids, pred_ids, num_labels, indices, input_mask)\n",
        "                    accuracy = tf.metrics.accuracy(label_ids, pred_ids, input_mask)\n",
        "                    loss = tf.metrics.mean(per_example_loss)\n",
        "                    return {\n",
        "                        'eval_precision': precision,\n",
        "                        'eval_recall': recall,\n",
        "                        'eval_f': f,\n",
        "                        'eval_accuracy': accuracy,\n",
        "                        'eval_loss': loss,\n",
        "                    }\n",
        "                eval_metrics = (metric_fn, [label_ids, pred_ids, per_example_loss, input_mask])\n",
        "                output_spec = tf.contrib.tpu.TPUEstimatorSpec(\n",
        "                    mode=mode,\n",
        "                    loss=total_loss,\n",
        "                    eval_metrics=eval_metrics,\n",
        "                    scaffold_fn=scaffold_fn)\n",
        "        return output_spec\n",
        "\n",
        "    return model_fn\n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "S1V4SkheEeXw",
        "colab_type": "text"
      },
      "source": [
        "## main"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IjBDEgHOEc7s",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def main(_):\n",
        "    tf.logging.set_verbosity(tf.logging.INFO)\n",
        "    processors = {\n",
        "        \"ner\": NerProcessor\n",
        "    }\n",
        "\n",
        "    bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)\n",
        "\n",
        "    if FLAGS.max_seq_length > bert_config.max_position_embeddings:\n",
        "        raise ValueError(\n",
        "            \"Cannot use sequence length %d because the BERT model \"\n",
        "            \"was only trained up to sequence length %d\" %\n",
        "            (FLAGS.max_seq_length, bert_config.max_position_embeddings))\n",
        "\n",
        "    task_name = FLAGS.task_name.lower()\n",
        "    if task_name not in processors:\n",
        "        raise ValueError(\"Task not found: %s\" % (task_name))\n",
        "    processor = processors[task_name]()\n",
        "\n",
        "    label_list = processor.get_labels()\n",
        "\n",
        "    tokenizer = tokenization.FullTokenizer(\n",
        "        vocab_file=FLAGS.vocab_file, do_lower_case=FLAGS.do_lower_case)\n",
        "    tpu_cluster_resolver = None\n",
        "    if FLAGS.use_tpu and FLAGS.tpu_name:\n",
        "        tpu_cluster_resolver = tf.contrib.cluster_resolver.TPUClusterResolver(\n",
        "            FLAGS.tpu_name, zone=FLAGS.tpu_zone, project=FLAGS.gcp_project)\n",
        "\n",
        "    is_per_host = tf.contrib.tpu.InputPipelineConfig.PER_HOST_V2\n",
        "\n",
        "    run_config = tf.contrib.tpu.RunConfig(\n",
        "        cluster=tpu_cluster_resolver,\n",
        "        master=FLAGS.master,\n",
        "        model_dir=FLAGS.output_dir,\n",
        "        save_checkpoints_steps=FLAGS.save_checkpoints_steps,\n",
        "        keep_checkpoint_max=FLAGS.keep_checkpoint_max,\n",
        "        save_summary_steps=FLAGS.save_summary_steps,\n",
        "        tpu_config=tf.contrib.tpu.TPUConfig(\n",
        "            iterations_per_loop=FLAGS.iterations_per_loop,\n",
        "            num_shards=FLAGS.num_tpu_cores,\n",
        "            per_host_input_for_training=is_per_host))\n",
        "\n",
        "    train_examples = None\n",
        "    num_train_steps = None\n",
        "    num_warmup_steps = None\n",
        "\n",
        "    if os.path.exists(FLAGS.data_config_path):\n",
        "        with codecs.open(FLAGS.data_config_path) as fd:\n",
        "            print(f'=========== load existed config:{FLAGS.data_config_path}')\n",
        "            data_config = json.load(fd)\n",
        "    else:\n",
        "        data_config = {}\n",
        "\n",
        "    if FLAGS.do_train:\n",
        "        if len(data_config) == 0:\n",
        "            train_examples = processor.get_train_examples(FLAGS.data_dir)\n",
        "            num_train_steps = int((len(train_examples) / FLAGS.train_batch_size) * FLAGS.num_train_epochs)\n",
        "            num_warmup_steps = int(num_train_steps * FLAGS.warmup_proportion)\n",
        "            data_config['num_train_steps'] = num_train_steps\n",
        "            data_config['num_warmup_steps'] = num_warmup_steps\n",
        "            data_config['num_train_size'] = len(train_examples)\n",
        "        else:\n",
        "            num_train_steps = int(data_config['num_train_steps'])\n",
        "            num_warmup_steps = int(data_config['num_warmup_steps'])\n",
        "\n",
        "    model_fn = model_fn_builder(\n",
        "        bert_config=bert_config,\n",
        "        num_labels=len(label_list) + 1, # 1 for '0' padding\n",
        "        init_checkpoint=FLAGS.init_checkpoint,\n",
        "        learning_rate=FLAGS.learning_rate,\n",
        "        num_train_steps=num_train_steps,\n",
        "        num_warmup_steps=num_warmup_steps,\n",
        "        use_tpu=FLAGS.use_tpu,\n",
        "        use_one_hot_embeddings=FLAGS.use_tpu)\n",
        "\n",
        "    if os.path.exists(FLAGS.output_dir):\n",
        "      print(f'================== use WarmStartSettings from {FLAGS.output_dir}')\n",
        "      ws = tf.estimator.WarmStartSettings(ckpt_to_initialize_from=FLAGS.output_dir)\n",
        "    else:\n",
        "      ws = None\n",
        "    estimator = tf.contrib.tpu.TPUEstimator(\n",
        "        use_tpu=FLAGS.use_tpu,\n",
        "        model_fn=model_fn,\n",
        "        config=run_config,\n",
        "        train_batch_size=FLAGS.train_batch_size,\n",
        "        eval_batch_size=FLAGS.eval_batch_size,\n",
        "        predict_batch_size=FLAGS.predict_batch_size,\n",
        "        warm_start_from = ws\n",
        "    )\n",
        "\n",
        "    if FLAGS.do_train:\n",
        "        # prepare train_input_fn\n",
        "        if data_config.get('train.tf_record_path', '') == '':\n",
        "            train_file = os.path.join(FLAGS.output_dir, \"train.tf_record\")\n",
        "            filed_based_convert_examples_to_features(\n",
        "                train_examples, label_list, FLAGS.max_seq_length, tokenizer, train_file)\n",
        "        else:\n",
        "            train_file = data_config.get('train.tf_record_path')\n",
        "        print(f'================== train file: {train_file}')\n",
        "        num_train_size = num_train_size = int(data_config['num_train_size'])\n",
        "        print(\"***** Running training *****\")\n",
        "        print(f\"  Num examples = {num_train_size}\" )\n",
        "        print(f\"  Batch size = {FLAGS.train_batch_size}\")\n",
        "        print(f\"  Num steps = {num_train_steps}\")\n",
        "        train_input_fn = file_based_input_fn_builder(\n",
        "            input_file=train_file,\n",
        "            seq_length=FLAGS.max_seq_length,\n",
        "            is_training=True,\n",
        "            drop_remainder=True)\n",
        "        # prepare eval_input_fn\n",
        "        if data_config.get('eval.tf_record_path', '') == '':\n",
        "            eval_examples = processor.get_dev_examples(FLAGS.data_dir)\n",
        "            eval_file = os.path.join(FLAGS.output_dir, \"eval.tf_record\")\n",
        "            filed_based_convert_examples_to_features(\n",
        "                eval_examples, label_list, FLAGS.max_seq_length, tokenizer, eval_file)\n",
        "            data_config['eval.tf_record_path'] = eval_file\n",
        "            data_config['num_eval_size'] = len(eval_examples)\n",
        "        else:\n",
        "            eval_file = data_config['eval.tf_record_path']\n",
        "        num_eval_size = data_config.get('num_eval_size', 0)\n",
        "        print(\"***** Running evaluation *****\")\n",
        "        print(f\"  Num examples = {num_eval_size}\" )\n",
        "        print(f\"  Batch size = {FLAGS.eval_batch_size}\" )\n",
        "        eval_steps = None\n",
        "        if FLAGS.use_tpu:\n",
        "            eval_steps = int(num_eval_size / FLAGS.eval_batch_size)\n",
        "        eval_drop_remainder = True if FLAGS.use_tpu else False\n",
        "        eval_input_fn = file_based_input_fn_builder(\n",
        "            input_file=eval_file,\n",
        "            seq_length=FLAGS.max_seq_length,\n",
        "            is_training=False,\n",
        "            drop_remainder=eval_drop_remainder)\n",
        "        # train and evaluate tf.estimator.experimental.\n",
        "        hook = tf.estimator.experimental.stop_if_no_decrease_hook(\n",
        "            estimator, 'eval_f', 3000, min_steps=30000, run_every_secs=120)\n",
        "        train_spec = tf.estimator.TrainSpec(input_fn=train_input_fn, max_steps=num_train_steps, hooks=[hook])\n",
        "        eval_spec = tf.estimator.EvalSpec(input_fn=eval_input_fn, throttle_secs=120)\n",
        "        tp = tf.estimator.train_and_evaluate(estimator, train_spec, eval_spec)\n",
        "        result = tp[0]\n",
        "        \n",
        "        output_eval_file = os.path.join(FLAGS.output_dir, \"eval_results.txt\")\n",
        "        with codecs.open(output_eval_file, \"w\", encoding='utf-8') as writer:\n",
        "            print(\"***** Eval results *****\")\n",
        "            for key in sorted(result.keys()):\n",
        "                print(f\"  {key} = {str(result[key])}\")\n",
        "                writer.write(f\"{key} = {str(result[key])}\\n\")\n",
        "\n",
        "    if not os.path.exists(FLAGS.data_config_path):\n",
        "        with codecs.open(FLAGS.data_config_path, 'a', encoding='utf-8') as fd:\n",
        "            json.dump(data_config, fd)\n",
        "\n",
        "    if FLAGS.do_predict:\n",
        "        # prepare predict_input_fn\n",
        "        predict_examples = processor.get_test_examples(FLAGS.data_dir)\n",
        "        predict_file = os.path.join(FLAGS.output_dir, \"predict.tf_record\")\n",
        "        filed_based_convert_examples_to_features(predict_examples, label_list,\n",
        "                                                 FLAGS.max_seq_length, tokenizer,\n",
        "                                                 predict_file, mode=\"test\")\n",
        "\n",
        "        print(\"***** Running prediction*****\")\n",
        "        print(f\"  Num examples = {len(predict_examples)}\" )\n",
        "        print(f\"  Batch size = {FLAGS.predict_batch_size}\" )\n",
        "        if FLAGS.use_tpu:\n",
        "            # Warning: According to tpu_estimator.py Prediction on TPU is an\n",
        "            # experimental feature and hence not supported here\n",
        "            raise ValueError(\"Prediction in TPU not supported\")\n",
        "        predict_drop_remainder = True if FLAGS.use_tpu else False\n",
        "        predict_input_fn = file_based_input_fn_builder(\n",
        "            input_file=predict_file,\n",
        "            seq_length=FLAGS.max_seq_length,\n",
        "            is_training=False,\n",
        "            drop_remainder=predict_drop_remainder)\n",
        "\n",
        "        # predict\n",
        "        predict_steps = None\n",
        "        if FLAGS.use_tpu:\n",
        "            predict_steps = int(len(predict_examples) / FLAGS.eval_batch_size)\n",
        "        predicted_result = estimator.evaluate(input_fn=predict_input_fn, steps=predict_steps)\n",
        "        output_eval_file = os.path.join(FLAGS.output_dir, \"predicted_results.txt\")\n",
        "        with codecs.open(output_eval_file, \"w\", encoding='utf-8') as writer:\n",
        "            tf.logging.info(\"***** Predict results *****\")\n",
        "            for key in sorted(predicted_result.keys()):\n",
        "                tf.logging.info(\"  %s = %s\", key, str(predicted_result[key]))\n",
        "                writer.write(\"%s = %s\\n\" % (key, str(predicted_result[key])))\n",
        "\n",
        "        result = estimator.predict(input_fn=predict_input_fn)\n",
        "        output_predict_file = os.path.join(FLAGS.output_dir, \"pred.txt\")\n",
        "\n",
        "        print('*' * 20)\n",
        "        print('type of result:%s, type of predict_examples:%s' % (type(result), type(predict_examples)))\n",
        "        print('*' * 20)\n",
        "        with codecs.open(os.path.join(FLAGS.output_dir,'label2id.pkl'), 'rb') as rf:\n",
        "            label2id = pickle.load(rf)\n",
        "            id2label = {value: key for key, value in label2id.items()}\n",
        "        with codecs.open(output_predict_file, 'w', encoding='utf-8') as writer:\n",
        "            for predict_example, prediction in zip(predict_examples, result):\n",
        "                tokens = predict_example.tokens\n",
        "                poss   = predict_example.poss\n",
        "                chunks = predict_example.chunks\n",
        "                labels = predict_example.labels\n",
        "                tokenized_tokens = predict_example.tokenized_tokens\n",
        "                tokenized_poss   = predict_example.tokenized_poss\n",
        "                tokenized_chunks = predict_example.tokenized_chunks\n",
        "                tokenized_labels = predict_example.tokenized_labels\n",
        "                text = predict_example.text\n",
        "                length = len(tokenized_tokens)\n",
        "                seq = 0\n",
        "                for token, pos, label, p_id in zip(tokenized_tokens, tokenized_poss, tokenized_labels, prediction[1:length+1]):\n",
        "                    p_label = 'O'\n",
        "                    if p_id != 0: p_label = id2label[p_id]\n",
        "                    if p_label == 'X': p_label = 'O'\n",
        "                    if label == 'X': continue\n",
        "                    org_token = tokens[seq]\n",
        "                    org_pos   = poss[seq]\n",
        "                    org_chunk = chunks[seq]\n",
        "                    org_label = labels[seq]\n",
        "                    output_line = ' '.join([org_token, org_pos, org_chunk, org_label, p_label])\n",
        "                    writer.write(output_line + '\\n')\n",
        "                    seq += 1\n",
        "                writer.write('\\n')\n",
        "\n",
        "\n",
        "def data_load():\n",
        "    processer = NerProcessor()\n",
        "    processer.get_labels()\n",
        "    processer.get_train_examples(FLAGS.data_dir)\n",
        "    print()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wsCTt3R5EiaA",
        "colab_type": "text"
      },
      "source": [
        "# run"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wswJn_eZEkbI",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "f30a20ae-4f3f-4728-abbd-630410fc97a3"
      },
      "source": [
        "if __name__ == \"__main__\":\n",
        "    print(tf.__version__)\n",
        "#     flags.DEFINE_bool(\"only_check_args\", False, \"Check.\")\n",
        "    flags.mark_flag_as_required(\"data_dir\")\n",
        "    flags.mark_flag_as_required(\"task_name\")\n",
        "    flags.mark_flag_as_required(\"vocab_file\")\n",
        "    flags.mark_flag_as_required(\"bert_config_file\")\n",
        "    flags.mark_flag_as_required(\"output_dir\")\n",
        "    tf.app.run()\n"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/absl/flags/_validators.py:359: UserWarning: Flag --data_dir has a non-None default value; therefore, mark_flag_as_required will pass even if flag is not specified in the command line!\n",
            "  'command line!' % flag_name)\n",
            "/usr/local/lib/python3.6/dist-packages/absl/flags/_validators.py:359: UserWarning: Flag --task_name has a non-None default value; therefore, mark_flag_as_required will pass even if flag is not specified in the command line!\n",
            "  'command line!' % flag_name)\n",
            "/usr/local/lib/python3.6/dist-packages/absl/flags/_validators.py:359: UserWarning: Flag --vocab_file has a non-None default value; therefore, mark_flag_as_required will pass even if flag is not specified in the command line!\n",
            "  'command line!' % flag_name)\n",
            "/usr/local/lib/python3.6/dist-packages/absl/flags/_validators.py:359: UserWarning: Flag --bert_config_file has a non-None default value; therefore, mark_flag_as_required will pass even if flag is not specified in the command line!\n",
            "  'command line!' % flag_name)\n",
            "/usr/local/lib/python3.6/dist-packages/absl/flags/_validators.py:359: UserWarning: Flag --output_dir has a non-None default value; therefore, mark_flag_as_required will pass even if flag is not specified in the command line!\n",
            "  'command line!' % flag_name)\n",
            "W0711 01:02:29.532893 140338823006080 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/bert/modeling.py:93: The name tf.gfile.GFile is deprecated. Please use tf.io.gfile.GFile instead.\n",
            "\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "1.14.0\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "W0711 01:02:30.876105 140338823006080 estimator.py:1984] Estimator's model_fn (<function model_fn_builder.<locals>.model_fn at 0x7fa2b5a11510>) includes params argument, but params are not passed to Estimator.\n",
            "I0711 01:02:30.879241 140338823006080 estimator.py:209] Using config: {'_model_dir': '/content/drive/My Drive/ner_checkpoints', '_tf_random_seed': None, '_save_summary_steps': 100, '_save_checkpoints_steps': 1000, '_save_checkpoints_secs': None, '_session_config': allow_soft_placement: true\n",
            "graph_options {\n",
            "  rewrite_options {\n",
            "    meta_optimizer_iterations: ONE\n",
            "  }\n",
            "}\n",
            ", '_keep_checkpoint_max': 20, '_keep_checkpoint_every_n_hours': 10000, '_log_step_count_steps': None, '_train_distribute': None, '_device_fn': None, '_protocol': None, '_eval_distribute': None, '_experimental_distribute': None, '_experimental_max_worker_delay_secs': None, '_service': None, '_cluster_spec': <tensorflow.python.training.server_lib.ClusterSpec object at 0x7fa2b2225eb8>, '_task_type': 'worker', '_task_id': 0, '_global_id_in_cluster': 0, '_master': '', '_evaluation_master': '', '_is_chief': True, '_num_ps_replicas': 0, '_num_worker_replicas': 1, '_tpu_config': TPUConfig(iterations_per_loop=1000, num_shards=8, num_cores_per_replica=None, per_host_input_for_training=3, tpu_job_name=None, initial_infeed_sleep_secs=None, input_partition_dims=None, eval_training_input_configuration=2), '_cluster': None}\n",
            "I0711 01:02:30.882924 140338823006080 tpu_context.py:209] _TPUContext: eval_on_tpu True\n",
            "W0711 01:02:30.885784 140338823006080 tpu_context.py:211] eval_on_tpu ignored because use_tpu is False.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "================== use WarmStartSettings from /content/drive/My Drive/ner_checkpoints\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "I0711 01:02:32.368182 140338823006080 <ipython-input-9-2f5db7366c31>:113] Writing example 0 of 14041\n",
            "I0711 01:02:32.370201 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 01:02:32.373961 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: train-0\n",
            "I0711 01:02:32.378167 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: EU rejects German call to boycott British la ##mb .\n",
            "I0711 01:02:32.385126 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 7270 22961 1528 1840 1106 21423 1418 2495 12913 119 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.387957 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.391048 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.396400 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 4 1 8 1 1 1 8 1 10 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.399037 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 01:02:32.400964 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: train-1\n",
            "I0711 01:02:32.402411 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: Peter Blackburn\n",
            "I0711 01:02:32.404586 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 1943 14428 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.406690 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.408315 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.410420 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 2 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.412744 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 01:02:32.414309 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: train-2\n",
            "I0711 01:02:32.416383 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: BR ##US ##SE ##LS 1996 - 08 - 22\n",
            "I0711 01:02:32.418234 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 26660 13329 12649 15928 1820 118 4775 118 1659 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.420351 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.422211 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.424346 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 6 10 10 10 1 10 10 10 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.429114 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 01:02:32.431130 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: train-3\n",
            "I0711 01:02:32.433119 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: The European Commission said on Thursday it disagreed with German advice to consumers to s ##hun British la ##mb until scientists determine whether mad cow disease can be transmitted to sheep .\n",
            "I0711 01:02:32.435297 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 1109 1735 2827 1163 1113 9170 1122 19786 1114 1528 5566 1106 11060 1106 188 17315 1418 2495 12913 1235 6479 4959 2480 6340 13991 3653 1169 1129 12086 1106 8892 119 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.437254 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.439949 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.442058 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 1 4 5 1 1 1 1 1 1 8 1 1 1 1 1 10 8 1 10 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.446993 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 01:02:32.448974 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: train-4\n",
            "I0711 01:02:32.451266 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: Germany ' s representative to the European Union ' s veterinary committee Werner Z ##wing ##mann said on Wednesday consumers should buy sheep ##me ##at from countries other than Britain until the scientific advice was clearer .\n",
            "I0711 01:02:32.453669 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 1860 112 188 4702 1106 1103 1735 1913 112 188 27431 3914 14651 163 7635 4119 1163 1113 9031 11060 1431 4417 8892 3263 2980 1121 2182 1168 1190 2855 1235 1103 3812 5566 1108 27830 119 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.455741 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.457848 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:32.459923 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 6 1 10 1 1 1 4 5 1 10 1 1 2 3 10 10 1 1 1 1 1 1 1 10 10 1 1 1 1 6 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:35.965183 140338823006080 <ipython-input-9-2f5db7366c31>:113] Writing example 5000 of 14041\n",
            "I0711 01:02:39.513208 140338823006080 <ipython-input-9-2f5db7366c31>:113] Writing example 10000 of 14041\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "================== train file: /content/drive/My Drive/ner_checkpoints/train.tf_record\n",
            "***** Running training *****\n",
            "  Num examples = 14041\n",
            "  Batch size = 16\n",
            "  Num steps = 2632\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "I0711 01:02:43.567397 140338823006080 <ipython-input-9-2f5db7366c31>:113] Writing example 0 of 3250\n",
            "I0711 01:02:43.569386 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 01:02:43.570629 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: dev-0\n",
            "I0711 01:02:43.576112 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: CR ##IC ##KE ##T - L ##EI ##CE ##ST ##ER ##S ##H ##IR ##E T ##A ##KE O ##VE ##R AT TO ##P A ##FT ##ER IN ##NI ##NG ##S VI ##CT ##OR ##Y .\n",
            "I0711 01:02:43.582231 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 15531 9741 22441 1942 118 149 27514 10954 9272 9637 1708 3048 18172 2036 157 1592 22441 152 17145 2069 13020 16972 2101 138 26321 9637 15969 27451 11780 1708 7118 16647 9565 3663 119 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.583781 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.586161 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.588284 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 1 10 10 10 1 4 10 10 10 10 10 10 10 10 1 10 10 1 10 10 1 1 10 1 10 10 1 10 10 10 1 10 10 10 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.591918 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 01:02:43.596188 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: dev-1\n",
            "I0711 01:02:43.599102 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: L ##ON ##D ##ON 1996 - 08 - 30\n",
            "I0711 01:02:43.601702 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 149 11414 2137 11414 1820 118 4775 118 1476 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.603861 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.606876 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.609172 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 6 10 10 10 1 10 10 10 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.614656 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 01:02:43.617642 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: dev-2\n",
            "I0711 01:02:43.620146 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: West Indian all - round ##er Phil Simmons took four for 38 on Friday as Leicestershire beat Somerset by an innings and 39 runs in two days to take over at the head of the county championship .\n",
            "I0711 01:02:43.623553 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 1537 1890 1155 118 1668 1200 5676 14068 1261 1300 1111 3383 1113 5286 1112 21854 3222 8860 1118 1126 6687 1105 3614 2326 1107 1160 1552 1106 1321 1166 1120 1103 1246 1104 1103 2514 2899 119 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.626881 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.632313 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.641937 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 8 9 1 10 10 10 2 3 1 1 1 1 1 1 1 4 1 4 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.648535 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 01:02:43.650118 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: dev-3\n",
            "I0711 01:02:43.652065 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: Their stay on top , though , may be short - lived as title rivals Essex , Derbyshire and Surrey all closed in on victory while Kent made up for lost time in their rain - affected match against Nottinghamshire .\n",
            "I0711 01:02:43.655128 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 2397 2215 1113 1499 117 1463 117 1336 1129 1603 118 2077 1112 1641 9521 8493 117 15964 1105 9757 1155 1804 1107 1113 2681 1229 5327 1189 1146 1111 1575 1159 1107 1147 4458 118 4634 1801 1222 21942 119 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.658428 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.661504 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.663854 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 1 1 1 1 1 1 1 1 1 1 10 10 1 1 1 4 1 4 1 4 1 1 1 1 1 1 4 1 1 1 1 1 1 1 1 10 10 1 1 4 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.670339 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 01:02:43.673891 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: dev-4\n",
            "I0711 01:02:43.676357 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: After bowling Somerset out for 83 on the opening morning at Grace Road , Leicestershire extended their first innings by 94 runs before being bowled out for 29 ##6 with England disc ##ard Andy C ##ad ##dick taking three for 83 .\n",
            "I0711 01:02:43.679291 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 1258 11518 8860 1149 1111 6032 1113 1103 2280 2106 1120 4378 1914 117 21854 2925 1147 1148 6687 1118 5706 2326 1196 1217 21663 1149 1111 1853 1545 1114 1652 6187 2881 4827 140 3556 25699 1781 1210 1111 6032 119 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.682294 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.685115 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:43.687239 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 1 1 4 1 1 1 1 1 1 1 1 6 7 1 4 1 1 1 1 1 1 1 1 1 1 1 1 1 10 1 6 1 10 2 3 10 10 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 01:02:46.136028 140338823006080 estimator_training.py:186] Not using Distribute Coordinator.\n",
            "I0711 01:02:46.138202 140338823006080 training.py:612] Running training and evaluation locally (non-distributed).\n",
            "I0711 01:02:46.142781 140338823006080 training.py:700] Start train and evaluate loop. The evaluate will happen after every checkpoint. Checkpoint frequency is determined based on RunConfig arguments: save_checkpoints_steps 1000 or save_checkpoints_secs None.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "***** Running evaluation *****\n",
            "  Num examples = 3250\n",
            "  Batch size = 8\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "W0711 01:02:47.806533 140338823006080 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/training/training_util.py:236: Variable.initialized_value (from tensorflow.python.ops.variables) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use Variable.read_value. Variables in 2.X are initialized automatically both in eager and graph (inside tf.defun) contexts.\n",
            "W0711 01:02:47.854762 140338823006080 deprecation.py:323] From <ipython-input-9-2f5db7366c31>:145: map_and_batch (from tensorflow.contrib.data.python.ops.batching) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use `tf.data.experimental.map_and_batch(...)`.\n",
            "W0711 01:02:47.856242 140338823006080 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/contrib/data/python/ops/batching.py:273: map_and_batch (from tensorflow.python.data.experimental.ops.batching) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use `tf.data.Dataset.map(map_func, num_parallel_calls)` followed by `tf.data.Dataset.batch(batch_size, drop_remainder)`. Static tf.data optimizations will take care of using the fused implementation.\n",
            "W0711 01:02:47.870836 140338823006080 deprecation.py:323] From <ipython-input-9-2f5db7366c31>:132: 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",
            "I0711 01:02:47.897601 140338823006080 estimator.py:1145] Calling model_fn.\n",
            "I0711 01:02:47.899058 140338823006080 tpu_estimator.py:2965] Running train on CPU\n",
            "I0711 01:02:47.900731 140338823006080 <ipython-input-9-2f5db7366c31>:282] *** Features ***\n",
            "I0711 01:02:47.902611 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_ids, shape = (16, 128)\n",
            "I0711 01:02:47.905618 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_mask, shape = (16, 128)\n",
            "I0711 01:02:47.908040 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = label_ids, shape = (16, 128)\n",
            "I0711 01:02:47.911424 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = segment_ids, shape = (16, 128)\n",
            "W0711 01:02:47.916149 140338823006080 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/bert/modeling.py:171: The name tf.variable_scope is deprecated. Please use tf.compat.v1.variable_scope instead.\n",
            "\n",
            "W0711 01:02:47.922315 140338823006080 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/bert/modeling.py:409: The name tf.get_variable is deprecated. Please use tf.compat.v1.get_variable instead.\n",
            "\n",
            "W0711 01:02:47.963440 140338823006080 deprecation_wrapper.py:119] From /usr/local/lib/python3.6/dist-packages/bert/modeling.py:490: The name tf.assert_less_equal is deprecated. Please use tf.compat.v1.assert_less_equal instead.\n",
            "\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "shape of input_ids (16, 128)\n",
            "shape of label_ids (16, 128)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "W0711 01:02:48.216416 140338823006080 deprecation.py:506] From /usr/local/lib/python3.6/dist-packages/bert/modeling.py:358: 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",
            "W0711 01:02:48.240741 140338823006080 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/bert/modeling.py:671: dense (from tensorflow.python.layers.core) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use keras.layers.dense instead.\n",
            "W0711 01:02:55.045642 140338823006080 deprecation.py:323] From <ipython-input-9-2f5db7366c31>:171: dropout (from tensorflow.python.layers.core) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use keras.layers.dropout instead.\n",
            "W0711 01:02:55.127223 140338823006080 deprecation.py:506] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/init_ops.py:1251: calling VarianceScaling.__init__ (from tensorflow.python.ops.init_ops) with dtype is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Call initializer instance with the dtype argument instead of passing it to the constructor\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "seq_length 128\n",
            "lengths Tensor(\"Sum:0\", shape=(16,), dtype=int32)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "W0711 01:02:56.285016 140338823006080 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/contrib/crf/python/ops/crf.py:213: dynamic_rnn (from tensorflow.python.ops.rnn) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use `keras.layers.RNN(cell)`, which is equivalent to this API\n",
            "W0711 01:02:56.366368 140338823006080 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:2403: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "####################\n",
            "shape of output_layer: (16, 128, 1024)\n",
            "embedding_size:1024\n",
            "seq_length:128\n",
            "shape of logit (16, 128, 11)\n",
            "shape of loss ()\n",
            "shape of per_example_loss (16,)\n",
            "num labels:11\n",
            "####################\n",
            "shape of pred_ids (16, 128)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "I0711 01:02:58.565713 140338823006080 <ipython-input-9-2f5db7366c31>:317] **** Trainable Variables ****\n",
            "I0711 01:02:58.567032 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/embeddings/word_embeddings:0, shape = (28996, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.568282 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/embeddings/token_type_embeddings:0, shape = (2, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.572048 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/embeddings/position_embeddings:0, shape = (512, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.573949 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/embeddings/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.575775 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/embeddings/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.577906 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.580357 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.582501 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.584836 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.587020 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.589499 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.591627 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.594052 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.596296 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.598650 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.600593 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.602996 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.605192 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.607631 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.611156 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.612709 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_0/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.614797 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.616441 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.618121 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.619825 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.621554 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.623549 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.625201 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.627291 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.628886 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.631006 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.633033 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.634772 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.636230 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.638003 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.639772 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.641446 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_1/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.643165 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.644862 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.646350 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.648188 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.649948 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.651690 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.653507 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.655211 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.656929 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.658555 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.660325 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.661761 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.671666 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.673974 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.675534 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.677015 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_2/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.678756 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.680603 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.682087 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.691787 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.695372 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.697661 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.699572 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.701132 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.702963 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.704458 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.706580 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.708524 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.710386 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.712300 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.714245 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.716349 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_3/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.718796 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.720863 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.722427 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.724302 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.726159 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.727967 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.729731 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.731627 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.733414 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.735236 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.737256 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.739131 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.740880 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.742439 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.744242 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.745970 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_4/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.747754 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.749603 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.751352 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.753128 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.754610 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.756534 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.758898 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.760937 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.765577 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.768547 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.770349 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.772663 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.774744 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.776730 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.778574 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.780659 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_5/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.781923 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.784683 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.787128 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.789276 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.790722 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.792820 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.794367 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.795897 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.798460 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.801039 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.803322 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.804516 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.807129 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.809690 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.811429 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.813138 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_6/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.814621 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.817217 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.818666 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.820260 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.822533 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.824508 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.826229 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.827891 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.829849 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.831377 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.833461 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.835052 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.836673 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.838769 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.840590 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.842363 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_7/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.844045 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.846109 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.847600 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.849376 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.851136 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.852922 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.854762 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.856657 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.858492 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.860126 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.861571 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.863873 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.865595 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.867170 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.869015 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.870891 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_8/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.872476 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.874102 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.875395 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.877824 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.880454 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.881758 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.886580 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.888498 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.889992 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.892911 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.896367 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.897903 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.899687 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.901110 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.903017 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.905148 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_9/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.906654 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.908389 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.909879 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.912340 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.914047 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.915754 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.917990 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.919939 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.921949 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.923803 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.925535 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.927391 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.929221 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.930882 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.932745 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.934480 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_10/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.936292 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.938145 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.939646 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.941511 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.943314 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.945110 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.947051 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.948835 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.950607 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.952363 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.954176 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.955909 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.959548 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.961705 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.967218 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.968891 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_11/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.970324 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.971810 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.975119 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.976919 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.978685 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.980151 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.981804 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.984266 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.986730 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.987959 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.990666 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.993087 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.995112 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.996894 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.998719 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:58.999921 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_12/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.003037 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.005427 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.007285 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.010120 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.011986 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.013877 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.015653 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.018680 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.019971 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.022357 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.023534 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.026541 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.028102 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.029844 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.031682 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.033579 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_13/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.035346 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.037252 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.038941 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.040656 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.042415 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.044233 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.046026 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.047634 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.049560 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.051435 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.053301 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.055559 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.057298 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.059104 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.060825 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.062315 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_14/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.063976 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.065665 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.067530 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.069317 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.070998 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.073379 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.075412 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.076767 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.079246 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.081502 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.083541 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.084913 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.086819 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.088903 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.090638 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.091768 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_15/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.094305 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.095884 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.097650 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.099867 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.101654 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.104007 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.105629 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.107413 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.109025 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.111679 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.113168 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.115326 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.116545 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.119406 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.121711 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.123945 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_16/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.126913 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.128605 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.129957 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.132334 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.134639 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.136727 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.138529 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.140254 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.141850 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.143491 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.145164 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.146549 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.148220 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.150704 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.152430 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.153940 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_17/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.155694 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.157370 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.159665 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.161330 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.163663 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.165281 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.166986 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.168515 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.171313 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.172606 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.174261 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.176049 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.178111 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.179806 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.181559 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.183025 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_18/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.184742 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.186166 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.187842 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.190349 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.192657 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.194686 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.196106 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.198515 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.200801 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.202581 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.204834 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.206490 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.207909 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.209699 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.212389 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.213831 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_19/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.215847 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.217910 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.218998 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.221660 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.224912 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.226957 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.228832 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.230638 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.232140 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.233603 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.235310 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.237107 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.238770 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.242038 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.244446 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.246391 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_20/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.248611 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.250546 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.252174 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.254014 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.256936 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.260180 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.266216 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.267853 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.269944 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.271516 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.272872 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.274849 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.276572 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.278460 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.280832 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.284209 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_21/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.286679 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.288402 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.290146 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.295768 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.298245 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.299938 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.302546 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.304284 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.306237 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.308258 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.310113 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.311463 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.312669 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.315415 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.317066 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.319649 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_22/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.322005 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/attention/self/query/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.323799 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/attention/self/query/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.327129 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/attention/self/key/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.330054 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/attention/self/key/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.333044 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/attention/self/value/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.335156 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/attention/self/value/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.336712 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/attention/output/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.338614 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/attention/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.340298 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/attention/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.342002 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/attention/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.343899 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/intermediate/dense/kernel:0, shape = (1024, 4096), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.345701 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/intermediate/dense/bias:0, shape = (4096,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.347704 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/output/dense/kernel:0, shape = (4096, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.349954 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/output/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.352212 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/output/LayerNorm/beta:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.353590 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/encoder/layer_23/output/LayerNorm/gamma:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.356368 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/pooler/dense/kernel:0, shape = (1024, 1024), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.358032 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bert/pooler/dense/bias:0, shape = (1024,), *INIT_FROM_CKPT*\n",
            "I0711 01:02:59.359386 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bi-lstm-fused-0/lstm_fused_cell/kernel:0, shape = (1152, 512)\n",
            "I0711 01:02:59.361324 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bi-lstm-fused-0/lstm_fused_cell/bias:0, shape = (512,)\n",
            "I0711 01:02:59.363539 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bi-lstm-fused-0/lstm_fused_cell_1/kernel:0, shape = (1152, 512)\n",
            "I0711 01:02:59.365838 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bi-lstm-fused-0/lstm_fused_cell_1/bias:0, shape = (512,)\n",
            "I0711 01:02:59.367653 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bi-lstm-fused-1/lstm_fused_cell/kernel:0, shape = (384, 512)\n",
            "I0711 01:02:59.369837 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bi-lstm-fused-1/lstm_fused_cell/bias:0, shape = (512,)\n",
            "I0711 01:02:59.371598 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bi-lstm-fused-1/lstm_fused_cell_1/kernel:0, shape = (384, 512)\n",
            "I0711 01:02:59.373699 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = bi-lstm-fused-1/lstm_fused_cell_1/bias:0, shape = (512,)\n",
            "I0711 01:02:59.375715 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = project/W:0, shape = (256, 11)\n",
            "I0711 01:02:59.377566 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = project/b:0, shape = (11,)\n",
            "I0711 01:02:59.379922 140338823006080 <ipython-input-9-2f5db7366c31>:323]   name = transitions:0, shape = (11, 11)\n",
            "I0711 01:03:16.568085 140338823006080 estimator.py:1147] Done calling model_fn.\n",
            "I0711 01:03:16.569667 140338823006080 estimator.py:1364] Warm-starting with WarmStartSettings: WarmStartSettings(ckpt_to_initialize_from='/content/drive/My Drive/ner_checkpoints', vars_to_warm_start='.*', var_name_to_vocab_info={}, var_name_to_prev_var_name={})\n",
            "I0711 01:03:16.572757 140338823006080 warm_starting_util.py:419] Warm-starting from: ('/content/drive/My Drive/ner_checkpoints',)\n",
            "I0711 01:03:16.575149 140338823006080 warm_starting_util.py:333] Warm-starting variables only in TRAINABLE_VARIABLES.\n",
            "I0711 01:03:18.241858 140338823006080 warm_starting_util.py:481] Warm-started 402 variables.\n",
            "I0711 01:03:18.243777 140338823006080 basic_session_run_hooks.py:541] Create CheckpointSaverHook.\n",
            "I0711 01:03:25.587305 140338823006080 monitored_session.py:240] Graph was finalized.\n",
            "W0711 01:03:26.818453 140338823006080 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/training/saver.py:1276: checkpoint_exists (from tensorflow.python.training.checkpoint_management) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use standard file APIs to check for files with this prefix.\n",
            "I0711 01:03:26.826434 140338823006080 saver.py:1280] Restoring parameters from /content/drive/My Drive/ner_checkpoints/model.ckpt-0\n",
            "W0711 01:04:06.838666 140338823006080 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/training/saver.py:1066: get_checkpoint_mtimes (from tensorflow.python.training.checkpoint_management) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use standard file utilities to get mtimes.\n",
            "I0711 01:04:08.312042 140338823006080 session_manager.py:500] Running local_init_op.\n",
            "I0711 01:04:08.642805 140338823006080 session_manager.py:502] Done running local_init_op.\n",
            "I0711 01:04:23.576631 140338823006080 basic_session_run_hooks.py:606] Saving checkpoints for 0 into /content/drive/My Drive/ner_checkpoints/model.ckpt.\n",
            "I0711 01:05:17.589041 140338823006080 basic_session_run_hooks.py:262] batch_loss = 73.551605\n",
            "I0711 01:05:43.293795 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.0389302\n",
            "I0711 01:05:43.297101 140338823006080 tpu_estimator.py:2160] examples/sec: 0.622883\n",
            "I0711 01:05:46.290543 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333695\n",
            "I0711 01:05:46.291943 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33911\n",
            "I0711 01:05:49.299835 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332299\n",
            "I0711 01:05:49.301284 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31678\n",
            "I0711 01:05:52.310923 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332106\n",
            "I0711 01:05:52.312567 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31369\n",
            "I0711 01:05:55.368504 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.327061\n",
            "I0711 01:05:55.373026 140338823006080 tpu_estimator.py:2160] examples/sec: 5.23298\n",
            "I0711 01:05:58.390802 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330871\n",
            "I0711 01:05:58.392198 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29393\n",
            "I0711 01:06:01.402566 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332032\n",
            "I0711 01:06:01.404767 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31252\n",
            "I0711 01:06:04.420441 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331388\n",
            "I0711 01:06:04.422232 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30221\n",
            "I0711 01:06:07.426148 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332674\n",
            "I0711 01:06:07.428534 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32278\n",
            "I0711 01:06:10.404848 140338823006080 basic_session_run_hooks.py:260] batch_loss = 24.58969 (52.816 sec)\n",
            "I0711 01:06:10.408343 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335322\n",
            "I0711 01:06:10.410062 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36515\n",
            "I0711 01:06:13.411357 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333004\n",
            "I0711 01:06:13.413166 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32806\n",
            "I0711 01:06:16.415053 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332912\n",
            "I0711 01:06:16.416536 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32659\n",
            "I0711 01:06:19.451493 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329358\n",
            "I0711 01:06:19.455613 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26973\n",
            "I0711 01:06:22.465319 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331788\n",
            "I0711 01:06:22.471238 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30861\n",
            "I0711 01:06:25.479352 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331775\n",
            "I0711 01:06:25.480882 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30839\n",
            "I0711 01:06:28.468584 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33454\n",
            "I0711 01:06:28.470001 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35264\n",
            "I0711 01:06:31.488195 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331173\n",
            "I0711 01:06:31.490380 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29876\n",
            "I0711 01:06:34.495912 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332472\n",
            "I0711 01:06:34.498990 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31955\n",
            "I0711 01:06:37.463297 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337\n",
            "I0711 01:06:37.464783 140338823006080 tpu_estimator.py:2160] examples/sec: 5.392\n",
            "I0711 01:06:40.454754 140338823006080 basic_session_run_hooks.py:260] batch_loss = 32.14753 (30.050 sec)\n",
            "I0711 01:06:40.457188 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334008\n",
            "I0711 01:06:40.459903 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34414\n",
            "I0711 01:06:43.385981 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341431\n",
            "I0711 01:06:43.387400 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4629\n",
            "I0711 01:06:46.364741 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335716\n",
            "I0711 01:06:46.366394 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37146\n",
            "I0711 01:06:49.333742 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336827\n",
            "I0711 01:06:49.335848 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38923\n",
            "I0711 01:06:52.318330 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335041\n",
            "I0711 01:06:52.319676 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36065\n",
            "I0711 01:06:55.287652 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336773\n",
            "I0711 01:06:55.292534 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38836\n",
            "I0711 01:06:58.263549 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336039\n",
            "I0711 01:06:58.265022 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37663\n",
            "I0711 01:07:01.207158 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339718\n",
            "I0711 01:07:01.209122 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43549\n",
            "I0711 01:07:04.181125 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336252\n",
            "I0711 01:07:04.182875 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38003\n",
            "I0711 01:07:07.137880 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338208\n",
            "I0711 01:07:07.139609 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41132\n",
            "I0711 01:07:10.114321 140338823006080 basic_session_run_hooks.py:260] batch_loss = 22.068901 (29.660 sec)\n",
            "I0711 01:07:10.120280 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335302\n",
            "I0711 01:07:10.123525 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36483\n",
            "I0711 01:07:13.059798 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340197\n",
            "I0711 01:07:13.061782 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44315\n",
            "I0711 01:07:15.967051 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343962\n",
            "I0711 01:07:15.968700 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50339\n",
            "I0711 01:07:18.914157 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339314\n",
            "I0711 01:07:18.915924 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42903\n",
            "I0711 01:07:21.872924 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337979\n",
            "I0711 01:07:21.874741 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40767\n",
            "I0711 01:07:24.832342 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33791\n",
            "I0711 01:07:24.833835 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40655\n",
            "I0711 01:07:27.817471 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334996\n",
            "I0711 01:07:27.824908 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35994\n",
            "I0711 01:07:30.784616 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337014\n",
            "I0711 01:07:30.786272 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39223\n",
            "I0711 01:07:33.717321 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341009\n",
            "I0711 01:07:33.719869 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45614\n",
            "I0711 01:07:36.652678 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340653\n",
            "I0711 01:07:36.654992 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45045\n",
            "I0711 01:07:39.585765 140338823006080 basic_session_run_hooks.py:260] batch_loss = 25.329533 (29.471 sec)\n",
            "I0711 01:07:39.592940 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340098\n",
            "I0711 01:07:39.594481 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44157\n",
            "I0711 01:07:42.546113 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338628\n",
            "I0711 01:07:42.547713 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41805\n",
            "I0711 01:07:45.484477 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340316\n",
            "I0711 01:07:45.486895 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44506\n",
            "I0711 01:07:48.374043 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.346078\n",
            "I0711 01:07:48.375506 140338823006080 tpu_estimator.py:2160] examples/sec: 5.53724\n",
            "I0711 01:07:51.297661 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342037\n",
            "I0711 01:07:51.299562 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4726\n",
            "I0711 01:07:54.205880 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343854\n",
            "I0711 01:07:54.207434 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50166\n",
            "I0711 01:07:57.139812 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340839\n",
            "I0711 01:07:57.141144 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45343\n",
            "I0711 01:08:00.069733 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341311\n",
            "I0711 01:08:00.071590 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46098\n",
            "I0711 01:08:03.021099 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338826\n",
            "I0711 01:08:03.022727 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42122\n",
            "I0711 01:08:05.990825 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336732\n",
            "I0711 01:08:05.992344 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38771\n",
            "I0711 01:08:08.913110 140338823006080 basic_session_run_hooks.py:260] batch_loss = 15.260422 (29.327 sec)\n",
            "I0711 01:08:08.919370 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341466\n",
            "I0711 01:08:08.921114 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46346\n",
            "I0711 01:08:11.859735 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3401\n",
            "I0711 01:08:11.861265 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44161\n",
            "I0711 01:08:14.773580 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34318\n",
            "I0711 01:08:14.775274 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49087\n",
            "I0711 01:08:17.745454 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336488\n",
            "I0711 01:08:17.747241 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38381\n",
            "I0711 01:08:20.686012 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340072\n",
            "I0711 01:08:20.687866 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44116\n",
            "I0711 01:08:23.628524 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339844\n",
            "I0711 01:08:23.630291 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43751\n",
            "I0711 01:08:26.506846 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.347425\n",
            "I0711 01:08:26.508476 140338823006080 tpu_estimator.py:2160] examples/sec: 5.5588\n",
            "I0711 01:08:29.415363 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343818\n",
            "I0711 01:08:29.416971 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50108\n",
            "I0711 01:08:32.398822 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335179\n",
            "I0711 01:08:32.400477 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36286\n",
            "I0711 01:08:35.392097 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334084\n",
            "I0711 01:08:35.393597 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34535\n",
            "I0711 01:08:38.427060 140338823006080 basic_session_run_hooks.py:260] batch_loss = 20.341976 (29.514 sec)\n",
            "I0711 01:08:38.434298 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328711\n",
            "I0711 01:08:38.436569 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25937\n",
            "I0711 01:08:41.369763 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34067\n",
            "I0711 01:08:41.371591 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45073\n",
            "I0711 01:08:44.329048 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337906\n",
            "I0711 01:08:44.330752 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4065\n",
            "I0711 01:08:47.318810 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334476\n",
            "I0711 01:08:47.320689 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35162\n",
            "I0711 01:08:50.290300 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336542\n",
            "I0711 01:08:50.291949 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38467\n",
            "I0711 01:08:53.267326 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3359\n",
            "I0711 01:08:53.269130 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3744\n",
            "I0711 01:08:56.281618 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331755\n",
            "I0711 01:08:56.282934 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30808\n",
            "I0711 01:08:59.202139 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342403\n",
            "I0711 01:08:59.203624 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47845\n",
            "I0711 01:09:02.207186 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332772\n",
            "I0711 01:09:02.208739 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32435\n",
            "I0711 01:09:05.207798 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333266\n",
            "I0711 01:09:05.209353 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33226\n",
            "I0711 01:09:08.188277 140338823006080 basic_session_run_hooks.py:260] batch_loss = 17.922901 (29.761 sec)\n",
            "I0711 01:09:08.193153 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334979\n",
            "I0711 01:09:08.196677 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35967\n",
            "I0711 01:09:11.185954 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334121\n",
            "I0711 01:09:11.187549 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34594\n",
            "I0711 01:09:14.177468 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334285\n",
            "I0711 01:09:14.178936 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34856\n",
            "I0711 01:09:17.187328 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332239\n",
            "I0711 01:09:17.188900 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31583\n",
            "I0711 01:09:20.179006 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334261\n",
            "I0711 01:09:20.180628 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34818\n",
            "I0711 01:09:23.175031 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33377\n",
            "I0711 01:09:23.176889 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34032\n",
            "I0711 01:09:26.145596 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336643\n",
            "I0711 01:09:26.147518 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38628\n",
            "I0711 01:09:29.098975 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338594\n",
            "I0711 01:09:29.100342 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41751\n",
            "I0711 01:09:32.074846 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336032\n",
            "I0711 01:09:32.076355 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37651\n",
            "I0711 01:09:35.055985 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335447\n",
            "I0711 01:09:35.057333 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36715\n",
            "I0711 01:09:38.022937 140338823006080 basic_session_run_hooks.py:260] batch_loss = 10.998643 (29.835 sec)\n",
            "I0711 01:09:38.027755 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336521\n",
            "I0711 01:09:38.030155 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38433\n",
            "I0711 01:09:40.994396 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337059\n",
            "I0711 01:09:40.996123 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39294\n",
            "I0711 01:09:43.992903 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3335\n",
            "I0711 01:09:43.994383 140338823006080 tpu_estimator.py:2160] examples/sec: 5.336\n",
            "I0711 01:09:46.956658 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337414\n",
            "I0711 01:09:46.959032 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39863\n",
            "I0711 01:09:49.936317 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335604\n",
            "I0711 01:09:49.937847 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36967\n",
            "I0711 01:09:52.923969 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334708\n",
            "I0711 01:09:52.925588 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35533\n",
            "I0711 01:09:55.900465 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335967\n",
            "I0711 01:09:55.902188 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37547\n",
            "I0711 01:09:58.951851 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.327769\n",
            "I0711 01:09:58.953424 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2443\n",
            "I0711 01:10:01.915206 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337407\n",
            "I0711 01:10:01.916932 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39852\n",
            "I0711 01:10:04.868780 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338568\n",
            "I0711 01:10:04.870227 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41709\n",
            "I0711 01:10:07.843253 140338823006080 basic_session_run_hooks.py:260] batch_loss = 13.5411415 (29.820 sec)\n",
            "I0711 01:10:07.850308 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335395\n",
            "I0711 01:10:07.852693 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36633\n",
            "I0711 01:10:10.854722 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332846\n",
            "I0711 01:10:10.856503 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32554\n",
            "I0711 01:10:13.816045 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337688\n",
            "I0711 01:10:13.817817 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40301\n",
            "I0711 01:10:16.794924 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335702\n",
            "I0711 01:10:16.796406 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37123\n",
            "I0711 01:10:19.766526 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336522\n",
            "I0711 01:10:19.768468 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38435\n",
            "I0711 01:10:22.723693 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338151\n",
            "I0711 01:10:22.728396 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41042\n",
            "I0711 01:10:25.708675 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335012\n",
            "I0711 01:10:25.710393 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36019\n",
            "I0711 01:10:28.707956 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333413\n",
            "I0711 01:10:28.709388 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33461\n",
            "I0711 01:10:31.739700 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329843\n",
            "I0711 01:10:31.741462 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27749\n",
            "I0711 01:10:34.725805 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334885\n",
            "I0711 01:10:34.727594 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35816\n",
            "I0711 01:10:37.719548 140338823006080 basic_session_run_hooks.py:260] batch_loss = 10.841002 (29.876 sec)\n",
            "I0711 01:10:37.722440 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33371\n",
            "I0711 01:10:37.729759 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33937\n",
            "I0711 01:10:40.754523 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329804\n",
            "I0711 01:10:40.756306 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27687\n",
            "I0711 01:10:43.730361 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33604\n",
            "I0711 01:10:43.732278 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37664\n",
            "I0711 01:10:46.700538 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336682\n",
            "I0711 01:10:46.701994 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38692\n",
            "I0711 01:10:49.651254 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338903\n",
            "I0711 01:10:49.653004 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42246\n",
            "I0711 01:10:52.636810 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334942\n",
            "I0711 01:10:52.638375 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35907\n",
            "I0711 01:10:55.569041 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341036\n",
            "I0711 01:10:55.570846 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45657\n",
            "I0711 01:10:58.523730 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338446\n",
            "I0711 01:10:58.525262 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41513\n",
            "I0711 01:11:01.517142 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33408\n",
            "I0711 01:11:01.518900 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34527\n",
            "I0711 01:11:04.498947 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335358\n",
            "I0711 01:11:04.500706 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36572\n",
            "I0711 01:11:07.479096 140338823006080 basic_session_run_hooks.py:260] batch_loss = 12.393509 (29.760 sec)\n",
            "I0711 01:11:07.482010 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335225\n",
            "I0711 01:11:07.487472 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36359\n",
            "I0711 01:11:10.453645 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336512\n",
            "I0711 01:11:10.455463 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38419\n",
            "I0711 01:11:13.417471 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337405\n",
            "I0711 01:11:13.419171 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39847\n",
            "I0711 01:11:16.354315 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340515\n",
            "I0711 01:11:16.356357 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44824\n",
            "I0711 01:11:19.320373 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337136\n",
            "I0711 01:11:19.322133 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39418\n",
            "I0711 01:11:22.284837 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337326\n",
            "I0711 01:11:22.286706 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39722\n",
            "I0711 01:11:25.278217 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334071\n",
            "I0711 01:11:25.279968 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34513\n",
            "I0711 01:11:28.256752 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33574\n",
            "I0711 01:11:28.258546 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37184\n",
            "I0711 01:11:31.199925 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339765\n",
            "I0711 01:11:31.201505 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43624\n",
            "I0711 01:11:34.169613 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336734\n",
            "I0711 01:11:34.171252 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38774\n",
            "I0711 01:11:37.183226 140338823006080 basic_session_run_hooks.py:260] batch_loss = 12.298817 (29.704 sec)\n",
            "I0711 01:11:37.188827 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331211\n",
            "I0711 01:11:37.192003 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29938\n",
            "I0711 01:11:40.125969 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340472\n",
            "I0711 01:11:40.127686 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44755\n",
            "I0711 01:11:43.064752 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340275\n",
            "I0711 01:11:43.066486 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4444\n",
            "I0711 01:11:46.092905 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330241\n",
            "I0711 01:11:46.094794 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28385\n",
            "I0711 01:11:49.039888 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339321\n",
            "I0711 01:11:49.042464 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42914\n",
            "I0711 01:11:52.001231 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337689\n",
            "I0711 01:11:52.003025 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40303\n",
            "I0711 01:11:54.932565 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34114\n",
            "I0711 01:11:54.934469 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45824\n",
            "I0711 01:11:57.890094 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338119\n",
            "I0711 01:11:57.891752 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40991\n",
            "I0711 01:12:00.821816 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341111\n",
            "I0711 01:12:00.823723 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45778\n",
            "I0711 01:12:03.809111 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334739\n",
            "I0711 01:12:03.810722 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35583\n",
            "I0711 01:12:06.782510 140338823006080 basic_session_run_hooks.py:260] batch_loss = 9.218151 (29.599 sec)\n",
            "I0711 01:12:06.785111 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336019\n",
            "I0711 01:12:06.789228 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3763\n",
            "I0711 01:12:09.823794 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329089\n",
            "I0711 01:12:09.825193 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26543\n",
            "I0711 01:12:12.847106 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330765\n",
            "I0711 01:12:12.848999 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29224\n",
            "I0711 01:12:15.839469 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334192\n",
            "I0711 01:12:15.841331 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34707\n",
            "I0711 01:12:18.825497 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334899\n",
            "I0711 01:12:18.827212 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35838\n",
            "I0711 01:12:21.768269 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339803\n",
            "I0711 01:12:21.769638 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43685\n",
            "I0711 01:12:24.757414 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33454\n",
            "I0711 01:12:24.759146 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35263\n",
            "I0711 01:12:27.718583 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337709\n",
            "I0711 01:12:27.720164 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40335\n",
            "I0711 01:12:30.713636 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333885\n",
            "I0711 01:12:30.715041 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34216\n",
            "I0711 01:12:33.689064 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336081\n",
            "I0711 01:12:33.690693 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3773\n",
            "I0711 01:12:36.637569 140338823006080 basic_session_run_hooks.py:260] batch_loss = 7.6226683 (29.855 sec)\n",
            "I0711 01:12:36.644481 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338362\n",
            "I0711 01:12:36.646441 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41379\n",
            "I0711 01:12:39.559339 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343071\n",
            "I0711 01:12:39.560668 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48913\n",
            "I0711 01:12:42.517923 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337999\n",
            "I0711 01:12:42.519442 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40799\n",
            "I0711 01:12:45.453301 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340674\n",
            "I0711 01:12:45.454722 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45079\n",
            "I0711 01:12:48.370547 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342787\n",
            "I0711 01:12:48.372029 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48459\n",
            "I0711 01:12:51.337063 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337093\n",
            "I0711 01:12:51.338696 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39349\n",
            "I0711 01:12:54.320543 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335185\n",
            "I0711 01:12:54.322535 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36297\n",
            "I0711 01:12:57.278434 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338076\n",
            "I0711 01:12:57.279927 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40922\n",
            "I0711 01:13:00.181345 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344488\n",
            "I0711 01:13:00.182772 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51181\n",
            "I0711 01:13:03.076475 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345404\n",
            "I0711 01:13:03.078090 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52646\n",
            "I0711 01:13:06.027582 140338823006080 basic_session_run_hooks.py:260] batch_loss = 6.6757984 (29.390 sec)\n",
            "I0711 01:13:06.030849 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338507\n",
            "I0711 01:13:06.032691 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41612\n",
            "I0711 01:13:09.002424 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336496\n",
            "I0711 01:13:09.004014 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38393\n",
            "I0711 01:13:11.936213 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340861\n",
            "I0711 01:13:11.937585 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45378\n",
            "I0711 01:13:14.856820 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342387\n",
            "I0711 01:13:14.858445 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47819\n",
            "I0711 01:13:17.837774 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335466\n",
            "I0711 01:13:17.839745 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36745\n",
            "I0711 01:13:20.847029 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332303\n",
            "I0711 01:13:20.848463 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31686\n",
            "I0711 01:13:23.811285 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337355\n",
            "I0711 01:13:23.812785 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39768\n",
            "I0711 01:13:26.759324 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339207\n",
            "I0711 01:13:26.760631 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42731\n",
            "I0711 01:13:29.687306 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341538\n",
            "I0711 01:13:29.689009 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4646\n",
            "I0711 01:13:32.636441 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339082\n",
            "I0711 01:13:32.637753 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42532\n",
            "I0711 01:13:35.633321 140338823006080 basic_session_run_hooks.py:260] batch_loss = 7.1598487 (29.606 sec)\n",
            "I0711 01:13:35.635966 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33338\n",
            "I0711 01:13:35.637634 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33408\n",
            "I0711 01:13:38.599525 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337432\n",
            "I0711 01:13:38.600977 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3989\n",
            "I0711 01:13:41.576791 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335882\n",
            "I0711 01:13:41.578429 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37411\n",
            "I0711 01:13:44.537175 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337794\n",
            "I0711 01:13:44.538876 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4047\n",
            "I0711 01:13:47.489863 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338673\n",
            "I0711 01:13:47.491682 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41877\n",
            "I0711 01:13:50.491609 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333147\n",
            "I0711 01:13:50.494073 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33035\n",
            "I0711 01:13:53.432562 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340019\n",
            "I0711 01:13:53.434206 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44031\n",
            "I0711 01:13:56.408993 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335974\n",
            "I0711 01:13:56.410682 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37558\n",
            "I0711 01:13:59.391477 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335296\n",
            "I0711 01:13:59.396767 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36474\n",
            "I0711 01:14:02.357644 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337135\n",
            "I0711 01:14:02.359527 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39415\n",
            "I0711 01:14:05.340858 140338823006080 basic_session_run_hooks.py:260] batch_loss = 8.630934 (29.708 sec)\n",
            "I0711 01:14:05.345908 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334647\n",
            "I0711 01:14:05.348522 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35435\n",
            "I0711 01:14:08.311723 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337166\n",
            "I0711 01:14:08.313312 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39465\n",
            "I0711 01:14:11.262980 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338838\n",
            "I0711 01:14:11.264551 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42141\n",
            "I0711 01:14:14.197679 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340754\n",
            "I0711 01:14:14.199499 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45207\n",
            "I0711 01:14:17.180477 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33525\n",
            "I0711 01:14:17.181838 140338823006080 tpu_estimator.py:2160] examples/sec: 5.364\n",
            "I0711 01:14:20.187719 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332538\n",
            "I0711 01:14:20.189183 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32062\n",
            "I0711 01:14:23.202329 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331716\n",
            "I0711 01:14:23.203828 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30746\n",
            "I0711 01:14:26.180178 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335813\n",
            "I0711 01:14:26.181761 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37301\n",
            "I0711 01:14:29.206688 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330412\n",
            "I0711 01:14:29.208181 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28659\n",
            "I0711 01:14:32.174398 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33696\n",
            "I0711 01:14:32.175991 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39135\n",
            "I0711 01:14:35.177849 140338823006080 basic_session_run_hooks.py:260] batch_loss = 7.430523 (29.837 sec)\n",
            "I0711 01:14:35.182746 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332404\n",
            "I0711 01:14:35.186173 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31846\n",
            "I0711 01:14:38.178595 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333797\n",
            "I0711 01:14:38.180009 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34075\n",
            "I0711 01:14:41.129273 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338916\n",
            "I0711 01:14:41.130955 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42265\n",
            "I0711 01:14:44.141934 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331923\n",
            "I0711 01:14:44.143231 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31077\n",
            "I0711 01:14:47.131145 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33455\n",
            "I0711 01:14:47.133085 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3528\n",
            "I0711 01:14:50.117180 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334881\n",
            "I0711 01:14:50.118602 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35809\n",
            "I0711 01:14:53.110940 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334028\n",
            "I0711 01:14:53.112638 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34445\n",
            "I0711 01:14:56.132145 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330995\n",
            "I0711 01:14:56.133837 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29593\n",
            "I0711 01:14:59.105924 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336267\n",
            "I0711 01:14:59.107275 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38028\n",
            "I0711 01:15:02.079979 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33624\n",
            "I0711 01:15:02.081389 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37984\n",
            "I0711 01:15:05.082728 140338823006080 basic_session_run_hooks.py:260] batch_loss = 9.829555 (29.905 sec)\n",
            "I0711 01:15:05.085102 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33277\n",
            "I0711 01:15:05.086695 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32432\n",
            "I0711 01:15:08.062964 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335809\n",
            "I0711 01:15:08.064417 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37294\n",
            "I0711 01:15:11.037530 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336184\n",
            "I0711 01:15:11.039024 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37894\n",
            "I0711 01:15:14.008678 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336575\n",
            "I0711 01:15:14.010001 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38521\n",
            "I0711 01:15:16.958184 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339036\n",
            "I0711 01:15:16.959937 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42457\n",
            "I0711 01:15:19.967715 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332276\n",
            "I0711 01:15:19.969412 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31642\n",
            "I0711 01:15:22.947722 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335568\n",
            "I0711 01:15:22.949224 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36909\n",
            "I0711 01:15:25.942161 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333954\n",
            "I0711 01:15:25.943762 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34327\n",
            "I0711 01:15:28.925616 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335182\n",
            "I0711 01:15:28.927136 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36292\n",
            "I0711 01:15:31.909741 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335105\n",
            "I0711 01:15:31.911444 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36167\n",
            "I0711 01:15:34.882456 140338823006080 basic_session_run_hooks.py:260] batch_loss = 5.6790314 (29.800 sec)\n",
            "I0711 01:15:34.888390 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335718\n",
            "I0711 01:15:34.890969 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37149\n",
            "I0711 01:15:37.872926 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335071\n",
            "I0711 01:15:37.874615 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36113\n",
            "I0711 01:15:40.841192 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336895\n",
            "I0711 01:15:40.842655 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39032\n",
            "I0711 01:15:43.804528 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337457\n",
            "I0711 01:15:43.806223 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39932\n",
            "I0711 01:15:46.769227 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337301\n",
            "I0711 01:15:46.771031 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39682\n",
            "I0711 01:15:49.733649 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337333\n",
            "I0711 01:15:49.740646 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39733\n",
            "I0711 01:15:52.720829 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334764\n",
            "I0711 01:15:52.723256 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35623\n",
            "I0711 01:15:55.699624 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335708\n",
            "I0711 01:15:55.701469 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37133\n",
            "I0711 01:15:58.664618 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337266\n",
            "I0711 01:15:58.666248 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39625\n",
            "I0711 01:16:01.638980 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336216\n",
            "I0711 01:16:01.640809 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37946\n",
            "I0711 01:16:04.619122 140338823006080 basic_session_run_hooks.py:260] batch_loss = 7.5189395 (29.737 sec)\n",
            "I0711 01:16:04.624980 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334925\n",
            "I0711 01:16:04.627910 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35879\n",
            "I0711 01:16:07.567139 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339855\n",
            "I0711 01:16:07.568755 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43768\n",
            "I0711 01:16:10.540411 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336321\n",
            "I0711 01:16:10.541992 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38114\n",
            "I0711 01:16:13.512409 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336478\n",
            "I0711 01:16:13.514130 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38365\n",
            "I0711 01:16:16.435526 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342102\n",
            "I0711 01:16:16.436939 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47363\n",
            "I0711 01:16:19.408964 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336314\n",
            "I0711 01:16:19.410749 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38102\n",
            "I0711 01:16:22.376706 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336949\n",
            "I0711 01:16:22.378114 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39118\n",
            "I0711 01:16:25.310593 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340852\n",
            "I0711 01:16:25.312125 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45363\n",
            "I0711 01:16:28.232181 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342284\n",
            "I0711 01:16:28.233899 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47655\n",
            "I0711 01:16:31.147623 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342992\n",
            "I0711 01:16:31.149167 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48787\n",
            "I0711 01:16:34.122749 140338823006080 basic_session_run_hooks.py:260] batch_loss = 5.766531 (29.504 sec)\n",
            "I0711 01:16:34.128708 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335446\n",
            "I0711 01:16:34.131727 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36714\n",
            "I0711 01:16:37.067415 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340289\n",
            "I0711 01:16:37.068945 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44462\n",
            "I0711 01:16:39.989695 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342195\n",
            "I0711 01:16:39.991299 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47513\n",
            "I0711 01:16:42.892697 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344479\n",
            "I0711 01:16:42.894954 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51167\n",
            "I0711 01:16:45.818510 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341778\n",
            "I0711 01:16:45.819966 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46845\n",
            "I0711 01:16:48.756314 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340392\n",
            "I0711 01:16:48.757750 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44627\n",
            "I0711 01:16:51.746518 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334424\n",
            "I0711 01:16:51.748150 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35079\n",
            "I0711 01:16:54.664485 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342704\n",
            "I0711 01:16:54.666277 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48327\n",
            "I0711 01:16:57.670342 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332684\n",
            "I0711 01:16:57.671685 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32295\n",
            "I0711 01:17:00.660282 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334462\n",
            "I0711 01:17:00.661951 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35138\n",
            "I0711 01:17:03.597528 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.9463105 (29.475 sec)\n",
            "I0711 01:17:03.603902 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339717\n",
            "I0711 01:17:03.607148 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43548\n",
            "I0711 01:17:06.569185 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337231\n",
            "I0711 01:17:06.570773 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3957\n",
            "I0711 01:17:09.478989 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343662\n",
            "I0711 01:17:09.480420 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49859\n",
            "I0711 01:17:12.372797 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345564\n",
            "I0711 01:17:12.374302 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52902\n",
            "I0711 01:17:15.275913 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344461\n",
            "I0711 01:17:15.277354 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51137\n",
            "I0711 01:17:18.174415 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345004\n",
            "I0711 01:17:18.176110 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52007\n",
            "I0711 01:17:21.069649 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3454\n",
            "I0711 01:17:21.071483 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52639\n",
            "I0711 01:17:23.971825 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344565\n",
            "I0711 01:17:23.973504 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51304\n",
            "I0711 01:17:26.960944 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334551\n",
            "I0711 01:17:26.962482 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35282\n",
            "I0711 01:17:29.881633 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3424\n",
            "I0711 01:17:29.883234 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4784\n",
            "I0711 01:17:32.824790 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.773777 (29.227 sec)\n",
            "I0711 01:17:32.832344 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338923\n",
            "I0711 01:17:32.834384 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42277\n",
            "I0711 01:17:35.727586 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345351\n",
            "I0711 01:17:35.728977 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52561\n",
            "I0711 01:17:38.647586 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342468\n",
            "I0711 01:17:38.648855 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4795\n",
            "I0711 01:17:41.552772 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344208\n",
            "I0711 01:17:41.554317 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50734\n",
            "I0711 01:17:44.483637 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341198\n",
            "I0711 01:17:44.485184 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45917\n",
            "I0711 01:17:47.422909 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340223\n",
            "I0711 01:17:47.424390 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44357\n",
            "I0711 01:17:50.359628 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340515\n",
            "I0711 01:17:50.361353 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44824\n",
            "I0711 01:17:53.255161 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345383\n",
            "I0711 01:17:53.257607 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52613\n",
            "I0711 01:17:56.171558 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342864\n",
            "I0711 01:17:56.173306 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48583\n",
            "I0711 01:17:59.048611 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.347586\n",
            "I0711 01:17:59.053781 140338823006080 tpu_estimator.py:2160] examples/sec: 5.56137\n",
            "I0711 01:18:01.966470 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.6174412 (29.142 sec)\n",
            "I0711 01:18:01.972762 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34198\n",
            "I0711 01:18:01.975348 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47168\n",
            "I0711 01:18:04.911035 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340328\n",
            "I0711 01:18:04.912845 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44524\n",
            "I0711 01:18:07.807563 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345242\n",
            "I0711 01:18:07.809029 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52388\n",
            "I0711 01:18:10.771843 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337348\n",
            "I0711 01:18:10.773720 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39756\n",
            "I0711 01:18:13.737376 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33721\n",
            "I0711 01:18:13.738914 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39535\n",
            "I0711 01:18:16.662545 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341863\n",
            "I0711 01:18:16.664309 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4698\n",
            "I0711 01:18:19.633102 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336636\n",
            "I0711 01:18:19.634570 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38618\n",
            "I0711 01:18:22.608867 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336047\n",
            "I0711 01:18:22.610233 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37676\n",
            "I0711 01:18:25.558624 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339014\n",
            "I0711 01:18:25.560118 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42422\n",
            "I0711 01:18:28.522650 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337382\n",
            "I0711 01:18:28.524488 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39811\n",
            "I0711 01:18:31.474012 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.3014708 (29.508 sec)\n",
            "I0711 01:18:31.479349 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338229\n",
            "I0711 01:18:31.482934 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41166\n",
            "I0711 01:18:34.442046 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337509\n",
            "I0711 01:18:34.443642 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40015\n",
            "I0711 01:18:37.430567 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334613\n",
            "I0711 01:18:37.432211 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35381\n",
            "I0711 01:18:40.406343 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336048\n",
            "I0711 01:18:40.407990 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37677\n",
            "I0711 01:18:43.390614 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335094\n",
            "I0711 01:18:43.392272 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3615\n",
            "I0711 01:18:46.324373 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340865\n",
            "I0711 01:18:46.326469 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45384\n",
            "I0711 01:18:49.302237 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335806\n",
            "I0711 01:18:49.303749 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3729\n",
            "I0711 01:18:52.291338 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33455\n",
            "I0711 01:18:52.293100 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3528\n",
            "I0711 01:18:55.323546 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329789\n",
            "I0711 01:18:55.325238 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27663\n",
            "I0711 01:18:58.295212 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336522\n",
            "I0711 01:18:58.297007 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38435\n",
            "I0711 01:19:01.292044 140338823006080 basic_session_run_hooks.py:260] batch_loss = 5.6486135 (29.818 sec)\n",
            "I0711 01:19:01.298422 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332964\n",
            "I0711 01:19:01.300402 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32743\n",
            "I0711 01:19:04.290292 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334245\n",
            "I0711 01:19:04.292126 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34792\n",
            "I0711 01:19:07.259586 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336776\n",
            "I0711 01:19:07.261394 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38842\n",
            "I0711 01:19:10.288291 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330181\n",
            "I0711 01:19:10.289756 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2829\n",
            "I0711 01:19:13.283724 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333838\n",
            "I0711 01:19:13.285262 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34141\n",
            "I0711 01:19:16.266247 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335287\n",
            "I0711 01:19:16.268047 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36458\n",
            "I0711 01:19:19.247933 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335376\n",
            "I0711 01:19:19.249464 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36602\n",
            "I0711 01:19:22.192010 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339667\n",
            "I0711 01:19:22.193870 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43468\n",
            "I0711 01:19:25.158601 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337086\n",
            "I0711 01:19:25.160097 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39337\n",
            "I0711 01:19:28.137721 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335668\n",
            "I0711 01:19:28.139419 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37068\n",
            "I0711 01:19:31.103864 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.8526485 (29.812 sec)\n",
            "I0711 01:19:31.109798 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336535\n",
            "I0711 01:19:31.112508 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38456\n",
            "I0711 01:19:34.054324 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339551\n",
            "I0711 01:19:34.056366 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43281\n",
            "I0711 01:19:36.990192 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340611\n",
            "I0711 01:19:36.991611 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44977\n",
            "I0711 01:19:39.946397 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33828\n",
            "I0711 01:19:39.948098 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41249\n",
            "I0711 01:19:42.954634 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332407\n",
            "I0711 01:19:42.956353 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31852\n",
            "I0711 01:19:45.918953 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337349\n",
            "I0711 01:19:45.920584 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39758\n",
            "I0711 01:19:48.834899 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342938\n",
            "I0711 01:19:48.836354 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48701\n",
            "I0711 01:19:51.769558 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340756\n",
            "I0711 01:19:51.771126 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4521\n",
            "I0711 01:19:54.733247 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33742\n",
            "I0711 01:19:54.735520 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39873\n",
            "I0711 01:19:57.638596 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344191\n",
            "I0711 01:19:57.640297 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50706\n",
            "I0711 01:20:00.580433 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.5065174 (29.477 sec)\n",
            "I0711 01:20:00.583413 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339578\n",
            "I0711 01:20:00.586796 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43325\n",
            "I0711 01:20:03.503142 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342506\n",
            "I0711 01:20:03.504463 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4801\n",
            "I0711 01:20:06.437888 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340736\n",
            "I0711 01:20:06.439263 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45178\n",
            "I0711 01:20:09.314003 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.347696\n",
            "I0711 01:20:09.319255 140338823006080 tpu_estimator.py:2160] examples/sec: 5.56313\n",
            "I0711 01:20:12.269312 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338377\n",
            "I0711 01:20:12.270880 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41403\n",
            "I0711 01:20:15.172493 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344448\n",
            "I0711 01:20:15.173967 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51116\n",
            "I0711 01:20:18.074959 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344529\n",
            "I0711 01:20:18.076360 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51246\n",
            "I0711 01:20:21.006242 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341154\n",
            "I0711 01:20:21.007942 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45846\n",
            "I0711 01:20:23.898204 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345788\n",
            "I0711 01:20:23.899622 140338823006080 tpu_estimator.py:2160] examples/sec: 5.53261\n",
            "I0711 01:20:26.862524 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337339\n",
            "I0711 01:20:26.863995 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39742\n",
            "I0711 01:20:29.803318 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.1528037 (29.223 sec)\n",
            "I0711 01:20:29.808212 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33948\n",
            "I0711 01:20:29.810559 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43168\n",
            "I0711 01:20:32.763975 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338337\n",
            "I0711 01:20:32.765711 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41338\n",
            "I0711 01:20:35.683711 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342482\n",
            "I0711 01:20:35.685405 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47971\n",
            "I0711 01:20:38.623121 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340221\n",
            "I0711 01:20:38.624603 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44354\n",
            "I0711 01:20:41.556985 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340835\n",
            "I0711 01:20:41.558597 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45336\n",
            "I0711 01:20:44.505121 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339198\n",
            "I0711 01:20:44.506518 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42718\n",
            "I0711 01:20:47.445449 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340094\n",
            "I0711 01:20:47.447260 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4415\n",
            "I0711 01:20:50.366389 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342356\n",
            "I0711 01:20:50.367714 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4777\n",
            "I0711 01:20:53.327319 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337736\n",
            "I0711 01:20:53.328759 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40377\n",
            "I0711 01:20:56.247875 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342405\n",
            "I0711 01:20:56.249512 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47847\n",
            "I0711 01:20:59.209499 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.8558141 (29.406 sec)\n",
            "I0711 01:20:59.212187 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337343\n",
            "I0711 01:20:59.214182 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39749\n",
            "I0711 01:21:02.193009 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335478\n",
            "I0711 01:21:02.194482 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36765\n",
            "I0711 01:21:05.170134 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335898\n",
            "I0711 01:21:05.171665 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37436\n",
            "I0711 01:21:08.133399 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337454\n",
            "I0711 01:21:08.135002 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39926\n",
            "I0711 01:21:11.111251 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335819\n",
            "I0711 01:21:11.112824 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37311\n",
            "I0711 01:21:14.017260 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344115\n",
            "I0711 01:21:14.018668 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50584\n",
            "I0711 01:21:16.915571 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345022\n",
            "I0711 01:21:16.917113 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52035\n",
            "I0711 01:21:19.857880 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339873\n",
            "I0711 01:21:19.859431 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43796\n",
            "I0711 01:21:22.804877 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339331\n",
            "I0711 01:21:22.806795 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4293\n",
            "I0711 01:21:25.751121 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339416\n",
            "I0711 01:21:25.752511 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43065\n",
            "I0711 01:21:28.684725 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.3507762 (29.475 sec)\n",
            "I0711 01:21:28.690283 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34023\n",
            "I0711 01:21:28.693677 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44367\n",
            "I0711 01:21:31.601271 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343528\n",
            "I0711 01:21:31.602895 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49645\n",
            "I0711 01:21:34.518007 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342846\n",
            "I0711 01:21:34.519594 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48554\n",
            "I0711 01:21:37.397577 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.347273\n",
            "I0711 01:21:37.398923 140338823006080 tpu_estimator.py:2160] examples/sec: 5.55636\n",
            "I0711 01:21:40.303873 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344087\n",
            "I0711 01:21:40.305289 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50539\n",
            "I0711 01:21:43.253170 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339059\n",
            "I0711 01:21:43.254893 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42495\n",
            "I0711 01:21:46.192985 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340155\n",
            "I0711 01:21:46.197854 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44249\n",
            "I0711 01:21:49.122315 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341394\n",
            "I0711 01:21:49.124160 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46231\n",
            "I0711 01:21:52.032399 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343613\n",
            "I0711 01:21:52.033726 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49781\n",
            "I0711 01:21:54.964122 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3411\n",
            "I0711 01:21:54.966200 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45761\n",
            "I0711 01:21:57.888143 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.2146096 (29.203 sec)\n",
            "I0711 01:21:57.891886 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341569\n",
            "I0711 01:21:57.899239 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46511\n",
            "I0711 01:22:00.805835 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343162\n",
            "I0711 01:22:00.807201 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49059\n",
            "I0711 01:22:03.779572 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336278\n",
            "I0711 01:22:03.781120 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38045\n",
            "I0711 01:22:06.780899 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333185\n",
            "I0711 01:22:06.782409 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33096\n",
            "I0711 01:22:09.764547 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335163\n",
            "I0711 01:22:09.766272 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36261\n",
            "I0711 01:22:12.750323 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334931\n",
            "I0711 01:22:12.751936 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35889\n",
            "I0711 01:22:15.695639 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339509\n",
            "I0711 01:22:15.697063 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43215\n",
            "I0711 01:22:18.699897 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332867\n",
            "I0711 01:22:18.701590 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32587\n",
            "I0711 01:22:21.687277 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33474\n",
            "I0711 01:22:21.688757 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35585\n",
            "I0711 01:22:24.680845 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334049\n",
            "I0711 01:22:24.682384 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34478\n",
            "I0711 01:22:27.622688 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.480401 (29.735 sec)\n",
            "I0711 01:22:27.625508 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339594\n",
            "I0711 01:22:27.627181 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4335\n",
            "I0711 01:22:30.592181 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33708\n",
            "I0711 01:22:30.593778 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39328\n",
            "I0711 01:22:33.507167 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343058\n",
            "I0711 01:22:33.508762 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48893\n",
            "I0711 01:22:36.436374 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341386\n",
            "I0711 01:22:36.438038 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46218\n",
            "I0711 01:22:39.366608 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341269\n",
            "I0711 01:22:39.368530 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4603\n",
            "I0711 01:22:42.339323 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336397\n",
            "I0711 01:22:42.340803 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38235\n",
            "I0711 01:22:45.291505 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338727\n",
            "I0711 01:22:45.293247 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41964\n",
            "I0711 01:22:48.245576 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338514\n",
            "I0711 01:22:48.246991 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41623\n",
            "I0711 01:22:51.183859 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340337\n",
            "I0711 01:22:51.185493 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44539\n",
            "I0711 01:22:54.088776 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344247\n",
            "I0711 01:22:54.090406 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50796\n",
            "I0711 01:22:57.038608 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.8443382 (29.416 sec)\n",
            "I0711 01:22:57.040979 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338732\n",
            "I0711 01:22:57.043996 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41971\n",
            "I0711 01:22:59.999885 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337958\n",
            "I0711 01:23:00.001255 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40733\n",
            "I0711 01:23:02.986340 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334844\n",
            "I0711 01:23:02.987964 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3575\n",
            "I0711 01:23:05.911997 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341809\n",
            "I0711 01:23:05.913450 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46895\n",
            "I0711 01:23:08.849614 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34041\n",
            "I0711 01:23:08.851263 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44656\n",
            "I0711 01:23:11.856652 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332552\n",
            "I0711 01:23:11.858198 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32084\n",
            "I0711 01:23:14.795900 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34022\n",
            "I0711 01:23:14.797723 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44352\n",
            "I0711 01:23:17.719980 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341995\n",
            "I0711 01:23:17.721306 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47191\n",
            "I0711 01:23:20.661384 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339968\n",
            "I0711 01:23:20.662827 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43949\n",
            "I0711 01:23:23.615780 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338478\n",
            "I0711 01:23:23.617280 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41566\n",
            "I0711 01:23:26.572548 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.4832706 (29.534 sec)\n",
            "I0711 01:23:26.579565 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337427\n",
            "I0711 01:23:26.583708 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39884\n",
            "I0711 01:23:29.490271 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34354\n",
            "I0711 01:23:29.491865 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49665\n",
            "I0711 01:23:32.468656 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335749\n",
            "I0711 01:23:32.470302 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37199\n",
            "I0711 01:23:35.439538 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336601\n",
            "I0711 01:23:35.440934 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38562\n",
            "I0711 01:23:38.405360 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337176\n",
            "I0711 01:23:38.406724 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39481\n",
            "I0711 01:23:41.374501 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336796\n",
            "I0711 01:23:41.376065 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38874\n",
            "I0711 01:23:44.363058 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33462\n",
            "I0711 01:23:44.364448 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35392\n",
            "I0711 01:23:47.333169 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33668\n",
            "I0711 01:23:47.334595 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38688\n",
            "I0711 01:23:50.341717 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332383\n",
            "I0711 01:23:50.343381 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31813\n",
            "I0711 01:23:53.308782 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337035\n",
            "I0711 01:23:53.310111 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39257\n",
            "I0711 01:23:56.291807 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.4849925 (29.719 sec)\n",
            "I0711 01:23:56.296913 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334661\n",
            "I0711 01:23:56.300094 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35458\n",
            "I0711 01:23:59.322790 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330481\n",
            "I0711 01:23:59.324275 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28769\n",
            "I0711 01:24:02.331506 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332364\n",
            "I0711 01:24:02.332867 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31782\n",
            "I0711 01:24:05.339870 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332408\n",
            "I0711 01:24:05.341569 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31852\n",
            "I0711 01:24:08.354872 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33168\n",
            "I0711 01:24:08.356516 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30688\n",
            "I0711 01:24:11.323747 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336826\n",
            "I0711 01:24:11.325168 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38922\n",
            "I0711 01:24:14.317375 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334047\n",
            "I0711 01:24:14.318775 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34475\n",
            "I0711 01:24:17.296135 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335705\n",
            "I0711 01:24:17.297785 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37128\n",
            "I0711 01:24:20.262226 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337144\n",
            "I0711 01:24:20.263619 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3943\n",
            "I0711 01:24:23.302645 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328899\n",
            "I0711 01:24:23.304252 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26238\n",
            "I0711 01:24:26.310850 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.224804 (30.019 sec)\n",
            "I0711 01:24:26.316404 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331811\n",
            "I0711 01:24:26.319405 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30898\n",
            "I0711 01:24:29.297767 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33543\n",
            "I0711 01:24:29.299725 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36688\n",
            "I0711 01:24:32.285868 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334647\n",
            "I0711 01:24:32.287221 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35434\n",
            "I0711 01:24:35.298304 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331962\n",
            "I0711 01:24:35.299715 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31139\n",
            "I0711 01:24:38.285537 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334759\n",
            "I0711 01:24:38.287200 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35614\n",
            "I0711 01:24:41.277746 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334199\n",
            "I0711 01:24:41.279412 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34718\n",
            "I0711 01:24:44.217624 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340156\n",
            "I0711 01:24:44.219045 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44249\n",
            "I0711 01:24:47.201825 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335104\n",
            "I0711 01:24:47.203565 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36166\n",
            "I0711 01:24:50.185893 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335109\n",
            "I0711 01:24:50.187779 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36174\n",
            "I0711 01:24:53.151309 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337218\n",
            "I0711 01:24:53.152961 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39549\n",
            "I0711 01:24:56.132232 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.3345013 (29.821 sec)\n",
            "I0711 01:24:56.139187 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334698\n",
            "I0711 01:24:56.141352 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35516\n",
            "I0711 01:24:59.138357 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33342\n",
            "I0711 01:24:59.140049 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33473\n",
            "I0711 01:25:02.112785 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336185\n",
            "I0711 01:25:02.114449 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37895\n",
            "I0711 01:25:05.058582 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339469\n",
            "I0711 01:25:05.060636 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43151\n",
            "I0711 01:25:08.035541 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33592\n",
            "I0711 01:25:08.036783 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37473\n",
            "I0711 01:25:10.986903 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338823\n",
            "I0711 01:25:10.988550 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42117\n",
            "I0711 01:25:13.951626 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337294\n",
            "I0711 01:25:13.953149 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3967\n",
            "I0711 01:25:16.898717 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339326\n",
            "I0711 01:25:16.900088 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42921\n",
            "I0711 01:25:19.793882 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345402\n",
            "I0711 01:25:19.795369 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52643\n",
            "I0711 01:25:22.734383 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340074\n",
            "I0711 01:25:22.735927 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44118\n",
            "I0711 01:25:25.674812 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.3530695 (29.543 sec)\n",
            "I0711 01:25:25.681597 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339322\n",
            "I0711 01:25:25.685339 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42915\n",
            "I0711 01:25:28.640838 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337904\n",
            "I0711 01:25:28.642665 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40647\n",
            "I0711 01:25:31.617728 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335925\n",
            "I0711 01:25:31.619407 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3748\n",
            "I0711 01:25:34.539162 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3423\n",
            "I0711 01:25:34.540487 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47679\n",
            "I0711 01:25:37.469582 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341245\n",
            "I0711 01:25:37.471112 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45993\n",
            "I0711 01:25:40.440247 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336627\n",
            "I0711 01:25:40.441826 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38603\n",
            "I0711 01:25:43.388408 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339195\n",
            "I0711 01:25:43.390013 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42712\n",
            "I0711 01:25:46.320181 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341092\n",
            "I0711 01:25:46.321982 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45747\n",
            "I0711 01:25:49.311673 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334282\n",
            "I0711 01:25:49.313984 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34851\n",
            "I0711 01:25:52.244729 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340951\n",
            "I0711 01:25:52.246460 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45521\n",
            "I0711 01:25:55.182195 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.7684621 (29.507 sec)\n",
            "I0711 01:25:55.187203 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339835\n",
            "I0711 01:25:55.189125 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43736\n",
            "I0711 01:25:58.160501 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336327\n",
            "I0711 01:25:58.165918 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38124\n",
            "I0711 01:26:01.176412 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331576\n",
            "I0711 01:26:01.181333 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30521\n",
            "I0711 01:26:04.141279 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337289\n",
            "I0711 01:26:04.142935 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39662\n",
            "I0711 01:26:07.071900 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341215\n",
            "I0711 01:26:07.073333 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45945\n",
            "I0711 01:26:10.046782 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336153\n",
            "I0711 01:26:10.048281 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37845\n",
            "I0711 01:26:12.959627 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343304\n",
            "I0711 01:26:12.961151 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49286\n",
            "I0711 01:26:15.878399 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342616\n",
            "I0711 01:26:15.879777 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48186\n",
            "I0711 01:26:18.847064 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336847\n",
            "I0711 01:26:18.848458 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38956\n",
            "I0711 01:26:21.838222 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334323\n",
            "I0711 01:26:21.839753 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34918\n",
            "I0711 01:26:24.829240 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.82575 (29.647 sec)\n",
            "I0711 01:26:24.839061 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333234\n",
            "I0711 01:26:24.841130 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33175\n",
            "I0711 01:26:27.773138 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340827\n",
            "I0711 01:26:27.774601 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45323\n",
            "I0711 01:26:30.712156 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340246\n",
            "I0711 01:26:30.713658 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44394\n",
            "I0711 01:26:33.649594 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340436\n",
            "I0711 01:26:33.651223 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44698\n",
            "I0711 01:26:36.581752 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341039\n",
            "I0711 01:26:36.583138 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45662\n",
            "I0711 01:26:39.512130 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341256\n",
            "I0711 01:26:39.513873 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4601\n",
            "I0711 01:26:42.460106 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339214\n",
            "I0711 01:26:42.461675 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42743\n",
            "I0711 01:26:45.377751 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342744\n",
            "I0711 01:26:45.379127 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48391\n",
            "I0711 01:26:48.366281 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334616\n",
            "I0711 01:26:48.367691 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35386\n",
            "I0711 01:26:51.324601 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338025\n",
            "I0711 01:26:51.326134 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4084\n",
            "I0711 01:26:54.312026 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.3089337 (29.483 sec)\n",
            "I0711 01:26:54.317854 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334085\n",
            "I0711 01:26:54.319988 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34535\n",
            "I0711 01:26:57.329479 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332052\n",
            "I0711 01:26:57.330986 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31283\n",
            "I0711 01:27:00.326219 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333697\n",
            "I0711 01:27:00.327869 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33916\n",
            "I0711 01:27:03.301988 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336042\n",
            "I0711 01:27:03.303522 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37667\n",
            "I0711 01:27:06.248278 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339411\n",
            "I0711 01:27:06.249620 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43058\n",
            "I0711 01:27:09.201945 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338568\n",
            "I0711 01:27:09.203737 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4171\n",
            "I0711 01:27:12.178428 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335966\n",
            "I0711 01:27:12.180201 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37545\n",
            "I0711 01:27:15.181195 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333026\n",
            "I0711 01:27:15.182736 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32842\n",
            "I0711 01:27:18.138504 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338138\n",
            "I0711 01:27:18.139847 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41021\n",
            "I0711 01:27:21.137966 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33339\n",
            "I0711 01:27:21.139577 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33424\n",
            "I0711 01:27:24.076451 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.4317794 (29.764 sec)\n",
            "I0711 01:27:24.079118 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340006\n",
            "I0711 01:27:24.080744 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4401\n",
            "I0711 01:27:27.081005 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333121\n",
            "I0711 01:27:27.082412 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32994\n",
            "I0711 01:27:30.111139 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330022\n",
            "I0711 01:27:30.112581 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28035\n",
            "I0711 01:27:33.038785 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341572\n",
            "I0711 01:27:33.040666 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46514\n",
            "I0711 01:27:36.019375 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335505\n",
            "I0711 01:27:36.020645 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36808\n",
            "I0711 01:27:38.970733 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338826\n",
            "I0711 01:27:38.972034 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42121\n",
            "I0711 01:27:41.896926 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341744\n",
            "I0711 01:27:41.898737 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4679\n",
            "I0711 01:27:44.845888 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339106\n",
            "I0711 01:27:44.848175 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42569\n",
            "I0711 01:27:47.859256 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331854\n",
            "I0711 01:27:47.863794 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30967\n",
            "I0711 01:27:50.872334 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33188\n",
            "I0711 01:27:50.873697 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31008\n",
            "I0711 01:27:53.868652 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.3649664 (29.792 sec)\n",
            "I0711 01:27:53.875288 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333017\n",
            "I0711 01:27:53.877155 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32827\n",
            "I0711 01:27:56.860959 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33492\n",
            "I0711 01:27:56.862545 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35872\n",
            "I0711 01:27:59.848156 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334766\n",
            "I0711 01:27:59.850552 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35626\n",
            "I0711 01:28:02.823333 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336116\n",
            "I0711 01:28:02.824952 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37786\n",
            "I0711 01:28:05.832404 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332323\n",
            "I0711 01:28:05.833844 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31717\n",
            "I0711 01:28:08.833475 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333216\n",
            "I0711 01:28:08.835183 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33145\n",
            "I0711 01:28:11.835887 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333071\n",
            "I0711 01:28:11.837655 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32914\n",
            "I0711 01:28:14.844029 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332424\n",
            "I0711 01:28:14.845448 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31879\n",
            "I0711 01:28:17.835616 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334269\n",
            "I0711 01:28:17.837231 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3483\n",
            "I0711 01:28:20.823694 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334665\n",
            "I0711 01:28:20.825212 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35463\n",
            "I0711 01:28:23.787639 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.6638088 (29.919 sec)\n",
            "I0711 01:28:23.795198 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336534\n",
            "I0711 01:28:23.798234 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38454\n",
            "I0711 01:28:26.785126 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334461\n",
            "I0711 01:28:26.786722 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35137\n",
            "I0711 01:28:29.764494 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335642\n",
            "I0711 01:28:29.766414 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37027\n",
            "I0711 01:28:32.721853 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338131\n",
            "I0711 01:28:32.723411 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4101\n",
            "I0711 01:28:35.706548 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335047\n",
            "I0711 01:28:35.708100 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36075\n",
            "I0711 01:28:38.692756 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334867\n",
            "I0711 01:28:38.694284 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35787\n",
            "I0711 01:28:41.685063 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334192\n",
            "I0711 01:28:41.686533 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34708\n",
            "I0711 01:28:44.649365 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33735\n",
            "I0711 01:28:44.650925 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3976\n",
            "I0711 01:28:47.670854 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330962\n",
            "I0711 01:28:47.672474 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2954\n",
            "I0711 01:28:50.647634 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335928\n",
            "I0711 01:28:50.649203 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37485\n",
            "I0711 01:28:53.632596 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.4944735 (29.845 sec)\n",
            "I0711 01:28:53.639858 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334205\n",
            "I0711 01:28:53.641638 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34727\n",
            "I0711 01:28:56.592795 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338644\n",
            "I0711 01:28:56.594145 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4183\n",
            "I0711 01:28:59.524598 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341092\n",
            "I0711 01:28:59.526124 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45747\n",
            "I0711 01:29:02.493606 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336818\n",
            "I0711 01:29:02.495483 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38909\n",
            "I0711 01:29:05.468016 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336195\n",
            "I0711 01:29:05.469989 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37912\n",
            "I0711 01:29:08.507031 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329049\n",
            "I0711 01:29:08.508754 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26478\n",
            "I0711 01:29:11.492398 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334968\n",
            "I0711 01:29:11.493898 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35949\n",
            "I0711 01:29:14.467460 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336129\n",
            "I0711 01:29:14.469122 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37806\n",
            "I0711 01:29:17.453998 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334844\n",
            "I0711 01:29:17.455621 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3575\n",
            "I0711 01:29:20.444299 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33441\n",
            "I0711 01:29:20.445897 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35056\n",
            "I0711 01:29:23.407958 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.223644 (29.775 sec)\n",
            "I0711 01:29:23.410655 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337112\n",
            "I0711 01:29:23.413326 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39379\n",
            "I0711 01:29:26.378891 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336901\n",
            "I0711 01:29:26.380668 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39042\n",
            "I0711 01:29:29.355247 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335989\n",
            "I0711 01:29:29.357085 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37582\n",
            "I0711 01:29:32.290451 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340681\n",
            "I0711 01:29:32.292356 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45089\n",
            "I0711 01:29:35.258501 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336928\n",
            "I0711 01:29:35.259914 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39085\n",
            "I0711 01:29:38.204464 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339442\n",
            "I0711 01:29:38.206109 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43107\n",
            "I0711 01:29:41.192700 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334653\n",
            "I0711 01:29:41.194900 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35445\n",
            "I0711 01:29:44.216777 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33067\n",
            "I0711 01:29:44.218182 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29072\n",
            "I0711 01:29:47.216023 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333419\n",
            "I0711 01:29:47.217373 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33471\n",
            "I0711 01:29:50.153509 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340435\n",
            "I0711 01:29:50.155303 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44696\n",
            "I0711 01:29:53.113729 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.879789 (29.706 sec)\n",
            "I0711 01:29:53.116373 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337504\n",
            "I0711 01:29:53.117841 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40006\n",
            "I0711 01:29:56.112269 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333791\n",
            "I0711 01:29:56.113716 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34066\n",
            "I0711 01:29:59.118941 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332591\n",
            "I0711 01:29:59.120298 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32146\n",
            "I0711 01:30:02.099376 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335522\n",
            "I0711 01:30:02.101927 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36835\n",
            "I0711 01:30:05.049798 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338933\n",
            "I0711 01:30:05.051362 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42293\n",
            "I0711 01:30:08.057786 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332454\n",
            "I0711 01:30:08.059468 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31927\n",
            "I0711 01:30:11.066350 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332381\n",
            "I0711 01:30:11.067966 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3181\n",
            "I0711 01:30:14.045238 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335704\n",
            "I0711 01:30:14.046960 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37126\n",
            "I0711 01:30:17.019097 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336254\n",
            "I0711 01:30:17.020568 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38006\n",
            "I0711 01:30:20.038899 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331149\n",
            "I0711 01:30:20.042057 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29839\n",
            "I0711 01:30:23.016980 140338823006080 basic_session_run_hooks.py:260] batch_loss = 7.72258 (29.903 sec)\n",
            "I0711 01:30:23.021331 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335288\n",
            "I0711 01:30:23.024219 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3646\n",
            "I0711 01:30:26.005035 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335163\n",
            "I0711 01:30:26.006667 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36261\n",
            "I0711 01:30:28.997865 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334128\n",
            "I0711 01:30:28.999482 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34604\n",
            "I0711 01:30:31.944061 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339421\n",
            "I0711 01:30:31.945530 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43073\n",
            "I0711 01:30:34.927499 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335193\n",
            "I0711 01:30:34.929308 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36309\n",
            "I0711 01:30:37.902760 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336096\n",
            "I0711 01:30:37.904272 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37753\n",
            "I0711 01:30:40.877519 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336162\n",
            "I0711 01:30:40.878772 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37859\n",
            "I0711 01:30:43.853474 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336027\n",
            "I0711 01:30:43.855145 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37643\n",
            "I0711 01:30:46.818135 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337313\n",
            "I0711 01:30:46.819880 140338823006080 tpu_estimator.py:2160] examples/sec: 5.397\n",
            "I0711 01:30:49.746853 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341439\n",
            "I0711 01:30:49.748418 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46302\n",
            "I0711 01:30:52.678286 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.3687277 (29.661 sec)\n",
            "I0711 01:30:52.680655 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340855\n",
            "I0711 01:30:52.682356 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45368\n",
            "I0711 01:30:55.626256 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3395\n",
            "I0711 01:30:55.627764 140338823006080 tpu_estimator.py:2160] examples/sec: 5.432\n",
            "I0711 01:30:58.564436 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340347\n",
            "I0711 01:30:58.565862 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44555\n",
            "I0711 01:31:01.508553 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339651\n",
            "I0711 01:31:01.510178 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43442\n",
            "I0711 01:31:04.398608 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34602\n",
            "I0711 01:31:04.400257 140338823006080 tpu_estimator.py:2160] examples/sec: 5.53632\n",
            "I0711 01:31:07.346155 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339264\n",
            "I0711 01:31:07.347618 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42822\n",
            "I0711 01:31:10.289614 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339733\n",
            "I0711 01:31:10.290968 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43573\n",
            "I0711 01:31:13.249355 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337868\n",
            "I0711 01:31:13.250716 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40589\n",
            "I0711 01:31:16.178108 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341448\n",
            "I0711 01:31:16.179714 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46316\n",
            "I0711 01:31:19.122710 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3396\n",
            "I0711 01:31:19.124010 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4336\n",
            "I0711 01:31:22.084927 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.339849 (29.407 sec)\n",
            "I0711 01:31:22.088647 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337165\n",
            "I0711 01:31:22.090502 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39464\n",
            "I0711 01:31:25.096105 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332503\n",
            "I0711 01:31:25.097690 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32005\n",
            "I0711 01:31:28.087968 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334241\n",
            "I0711 01:31:28.089643 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34786\n",
            "I0711 01:31:31.030408 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339851\n",
            "I0711 01:31:31.031801 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43762\n",
            "I0711 01:31:33.998758 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336887\n",
            "I0711 01:31:34.000422 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39019\n",
            "I0711 01:31:36.939738 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34002\n",
            "I0711 01:31:36.941122 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44032\n",
            "I0711 01:31:39.876119 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340561\n",
            "I0711 01:31:39.877604 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44898\n",
            "I0711 01:31:42.800929 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3419\n",
            "I0711 01:31:42.802772 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4704\n",
            "I0711 01:31:45.734531 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340893\n",
            "I0711 01:31:45.736266 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45429\n",
            "I0711 01:31:48.698844 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337336\n",
            "I0711 01:31:48.700623 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39737\n",
            "I0711 01:31:51.660099 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.0989728 (29.575 sec)\n",
            "I0711 01:31:51.664613 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337176\n",
            "I0711 01:31:51.667979 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39482\n",
            "I0711 01:31:54.627352 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337526\n",
            "I0711 01:31:54.628692 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40041\n",
            "I0711 01:31:57.608372 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335472\n",
            "I0711 01:31:57.613430 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36755\n",
            "I0711 01:32:00.600985 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33414\n",
            "I0711 01:32:00.602750 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34624\n",
            "I0711 01:32:03.562086 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337714\n",
            "I0711 01:32:03.564300 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40342\n",
            "I0711 01:32:06.548504 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334855\n",
            "I0711 01:32:06.550132 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35769\n",
            "I0711 01:32:09.527897 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335628\n",
            "I0711 01:32:09.529257 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37005\n",
            "I0711 01:32:12.527979 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333337\n",
            "I0711 01:32:12.529689 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33339\n",
            "I0711 01:32:15.507811 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335579\n",
            "I0711 01:32:15.509366 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36926\n",
            "I0711 01:32:18.467010 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33793\n",
            "I0711 01:32:18.468430 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40688\n",
            "I0711 01:32:21.424979 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.4043505 (29.765 sec)\n",
            "I0711 01:32:21.431709 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337428\n",
            "I0711 01:32:21.433837 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39884\n",
            "I0711 01:32:24.406541 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336037\n",
            "I0711 01:32:24.408423 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37659\n",
            "I0711 01:32:27.376862 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33666\n",
            "I0711 01:32:27.378271 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38655\n",
            "I0711 01:32:30.321265 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339632\n",
            "I0711 01:32:30.322590 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43412\n",
            "I0711 01:32:33.341513 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331089\n",
            "I0711 01:32:33.342864 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29742\n",
            "I0711 01:32:36.328756 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334761\n",
            "I0711 01:32:36.330257 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35618\n",
            "I0711 01:32:39.281142 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338705\n",
            "I0711 01:32:39.282610 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41929\n",
            "I0711 01:32:42.277149 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333791\n",
            "I0711 01:32:42.278904 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34066\n",
            "I0711 01:32:45.242022 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337268\n",
            "I0711 01:32:45.243687 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39629\n",
            "I0711 01:32:48.187129 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339554\n",
            "I0711 01:32:48.188951 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43286\n",
            "I0711 01:32:51.172439 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.8712714 (29.747 sec)\n",
            "I0711 01:32:51.174940 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334685\n",
            "I0711 01:32:51.179725 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35496\n",
            "I0711 01:32:54.115356 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340091\n",
            "I0711 01:32:54.117159 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44146\n",
            "I0711 01:32:57.113990 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333487\n",
            "I0711 01:32:57.115638 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33579\n",
            "I0711 01:33:00.096121 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33534\n",
            "I0711 01:33:00.097708 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36544\n",
            "I0711 01:33:03.105373 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332299\n",
            "I0711 01:33:03.106946 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31678\n",
            "I0711 01:33:06.075341 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336701\n",
            "I0711 01:33:06.076669 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38722\n",
            "I0711 01:33:09.101688 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330431\n",
            "I0711 01:33:09.103334 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2869\n",
            "I0711 01:33:12.103788 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3331\n",
            "I0711 01:33:12.105254 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3296\n",
            "I0711 01:33:15.142584 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329079\n",
            "I0711 01:33:15.144191 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26526\n",
            "I0711 01:33:18.098551 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338304\n",
            "I0711 01:33:18.100039 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41286\n",
            "I0711 01:33:21.076184 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.8585818 (29.904 sec)\n",
            "I0711 01:33:21.085188 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334879\n",
            "I0711 01:33:21.089282 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35807\n",
            "I0711 01:33:24.027510 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339805\n",
            "I0711 01:33:24.028866 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43689\n",
            "I0711 01:33:26.985800 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338035\n",
            "I0711 01:33:26.987236 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40856\n",
            "I0711 01:33:30.007879 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330897\n",
            "I0711 01:33:30.009164 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29436\n",
            "I0711 01:33:33.045387 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.32922\n",
            "I0711 01:33:33.047209 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26752\n",
            "I0711 01:33:36.002817 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338128\n",
            "I0711 01:33:36.004388 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41006\n",
            "I0711 01:33:38.968183 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33723\n",
            "I0711 01:33:38.969558 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39568\n",
            "I0711 01:33:41.949146 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335458\n",
            "I0711 01:33:41.950603 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36733\n",
            "I0711 01:33:44.885210 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340598\n",
            "I0711 01:33:44.886765 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44957\n",
            "I0711 01:33:47.879231 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334001\n",
            "I0711 01:33:47.880727 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34401\n",
            "I0711 01:33:50.812688 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.7776024 (29.737 sec)\n",
            "I0711 01:33:50.818619 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340205\n",
            "I0711 01:33:50.821927 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44327\n",
            "I0711 01:33:53.787682 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3368\n",
            "I0711 01:33:53.789146 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3888\n",
            "I0711 01:33:56.763362 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33606\n",
            "I0711 01:33:56.764832 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37696\n",
            "I0711 01:33:59.739535 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336005\n",
            "I0711 01:33:59.741122 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37607\n",
            "I0711 01:34:02.677979 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34031\n",
            "I0711 01:34:02.679585 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44497\n",
            "I0711 01:34:05.662577 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335054\n",
            "I0711 01:34:05.664862 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36086\n",
            "I0711 01:34:08.641967 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335643\n",
            "I0711 01:34:08.643632 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37028\n",
            "I0711 01:34:11.607167 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33725\n",
            "I0711 01:34:11.608619 140338823006080 tpu_estimator.py:2160] examples/sec: 5.396\n",
            "I0711 01:34:14.593049 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334899\n",
            "I0711 01:34:14.594732 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35839\n",
            "I0711 01:34:17.553910 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337742\n",
            "I0711 01:34:17.557499 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40387\n",
            "I0711 01:34:20.506346 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.474779 (29.694 sec)\n",
            "I0711 01:34:20.512567 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337996\n",
            "I0711 01:34:20.514107 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40794\n",
            "I0711 01:34:23.447556 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340713\n",
            "I0711 01:34:23.449058 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4514\n",
            "I0711 01:34:26.486272 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329091\n",
            "I0711 01:34:26.487969 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26546\n",
            "I0711 01:34:29.441240 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338409\n",
            "I0711 01:34:29.442500 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41455\n",
            "I0711 01:34:32.420382 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335678\n",
            "I0711 01:34:32.422157 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37085\n",
            "I0711 01:34:35.406972 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334821\n",
            "I0711 01:34:35.408689 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35714\n",
            "I0711 01:34:38.390483 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33517\n",
            "I0711 01:34:38.391759 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36271\n",
            "I0711 01:34:41.421654 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329908\n",
            "I0711 01:34:41.423022 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27853\n",
            "I0711 01:34:44.382009 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337793\n",
            "I0711 01:34:44.383626 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40469\n",
            "I0711 01:34:47.341359 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337927\n",
            "I0711 01:34:47.342936 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40683\n",
            "I0711 01:34:50.314459 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.1620953 (29.808 sec)\n",
            "I0711 01:34:50.321905 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335495\n",
            "I0711 01:34:50.323327 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36792\n",
            "I0711 01:34:53.285856 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33739\n",
            "I0711 01:34:53.287213 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39824\n",
            "I0711 01:34:56.244037 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338043\n",
            "I0711 01:34:56.245711 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40869\n",
            "I0711 01:34:59.215798 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336504\n",
            "I0711 01:34:59.217371 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38406\n",
            "I0711 01:35:02.219546 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332921\n",
            "I0711 01:35:02.221093 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32673\n",
            "I0711 01:35:05.207127 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334717\n",
            "I0711 01:35:05.208680 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35547\n",
            "I0711 01:35:08.167107 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337845\n",
            "I0711 01:35:08.168922 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40552\n",
            "I0711 01:35:11.153324 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334875\n",
            "I0711 01:35:11.155348 140338823006080 tpu_estimator.py:2160] examples/sec: 5.358\n",
            "I0711 01:35:14.140178 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334798\n",
            "I0711 01:35:14.141581 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35677\n",
            "I0711 01:35:17.116892 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335935\n",
            "I0711 01:35:17.118370 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37495\n",
            "I0711 01:35:20.074443 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.6258981 (29.760 sec)\n",
            "I0711 01:35:20.082063 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337248\n",
            "I0711 01:35:20.085244 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39596\n",
            "I0711 01:35:23.068826 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334827\n",
            "I0711 01:35:23.070423 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35724\n",
            "I0711 01:35:26.013672 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339559\n",
            "I0711 01:35:26.015201 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43295\n",
            "I0711 01:35:28.993829 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335551\n",
            "I0711 01:35:28.995252 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36882\n",
            "I0711 01:35:31.960114 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337133\n",
            "I0711 01:35:31.965098 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39413\n",
            "I0711 01:35:34.955415 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333853\n",
            "I0711 01:35:34.957086 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34164\n",
            "I0711 01:35:37.946492 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334321\n",
            "I0711 01:35:37.948222 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34913\n",
            "I0711 01:35:40.960713 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331764\n",
            "I0711 01:35:40.962118 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30822\n",
            "I0711 01:35:43.969721 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332332\n",
            "I0711 01:35:43.971141 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31731\n",
            "I0711 01:35:46.978338 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332385\n",
            "I0711 01:35:46.979690 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31816\n",
            "I0711 01:35:49.951011 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.4546359 (29.877 sec)\n",
            "I0711 01:35:49.959621 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335424\n",
            "I0711 01:35:49.961107 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36678\n",
            "I0711 01:35:52.940301 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3355\n",
            "I0711 01:35:52.941671 140338823006080 tpu_estimator.py:2160] examples/sec: 5.368\n",
            "I0711 01:35:55.910659 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336652\n",
            "I0711 01:35:55.912278 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38644\n",
            "I0711 01:35:58.921618 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332117\n",
            "I0711 01:35:58.923172 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31387\n",
            "I0711 01:36:01.903187 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335407\n",
            "I0711 01:36:01.904783 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36652\n",
            "I0711 01:36:04.901816 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333473\n",
            "I0711 01:36:04.903464 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33556\n",
            "I0711 01:36:07.881894 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335573\n",
            "I0711 01:36:07.884179 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36917\n",
            "I0711 01:36:10.841751 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337851\n",
            "I0711 01:36:10.843482 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40562\n",
            "I0711 01:36:13.808103 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337111\n",
            "I0711 01:36:13.809736 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39377\n",
            "I0711 01:36:16.753268 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339546\n",
            "I0711 01:36:16.754634 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43273\n",
            "I0711 01:36:19.718151 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2341156 (29.767 sec)\n",
            "I0711 01:36:19.726713 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336314\n",
            "I0711 01:36:19.728423 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38102\n",
            "I0711 01:36:22.723329 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333701\n",
            "I0711 01:36:22.724905 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33922\n",
            "I0711 01:36:25.674411 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338859\n",
            "I0711 01:36:25.676122 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42175\n",
            "I0711 01:36:28.628576 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338498\n",
            "I0711 01:36:28.630106 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41597\n",
            "I0711 01:36:31.565484 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340498\n",
            "I0711 01:36:31.566784 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44796\n",
            "I0711 01:36:34.509275 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339704\n",
            "I0711 01:36:34.510616 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43526\n",
            "I0711 01:36:37.425736 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342873\n",
            "I0711 01:36:37.427455 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48597\n",
            "I0711 01:36:40.366584 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340039\n",
            "I0711 01:36:40.368111 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44063\n",
            "I0711 01:36:43.354181 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334722\n",
            "I0711 01:36:43.355897 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35555\n",
            "I0711 01:36:46.286100 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34107\n",
            "I0711 01:36:46.287570 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45711\n",
            "I0711 01:36:49.221161 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0192666 (29.503 sec)\n",
            "I0711 01:36:49.227121 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340027\n",
            "I0711 01:36:49.229020 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44044\n",
            "I0711 01:36:52.193164 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337144\n",
            "I0711 01:36:52.195083 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39431\n",
            "I0711 01:36:55.158132 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337268\n",
            "I0711 01:36:55.159773 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39629\n",
            "I0711 01:36:58.111798 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338559\n",
            "I0711 01:36:58.113146 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41694\n",
            "I0711 01:37:01.041586 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341322\n",
            "I0711 01:37:01.043184 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46115\n",
            "I0711 01:37:03.954886 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343255\n",
            "I0711 01:37:03.961419 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49208\n",
            "I0711 01:37:06.951713 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333688\n",
            "I0711 01:37:06.953622 140338823006080 tpu_estimator.py:2160] examples/sec: 5.339\n",
            "I0711 01:37:09.989399 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329196\n",
            "I0711 01:37:09.991011 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26714\n",
            "I0711 01:37:12.918516 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341406\n",
            "I0711 01:37:12.920347 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4625\n",
            "I0711 01:37:15.869589 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338855\n",
            "I0711 01:37:15.871195 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42169\n",
            "I0711 01:37:18.821669 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.00215 (29.601 sec)\n",
            "I0711 01:37:18.828836 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337934\n",
            "I0711 01:37:18.830733 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40695\n",
            "I0711 01:37:21.891932 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.326472\n",
            "I0711 01:37:21.893794 140338823006080 tpu_estimator.py:2160] examples/sec: 5.22355\n",
            "I0711 01:37:24.869422 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335839\n",
            "I0711 01:37:24.871137 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37342\n",
            "I0711 01:37:27.831579 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337589\n",
            "I0711 01:37:27.833035 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40142\n",
            "I0711 01:37:30.828878 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333635\n",
            "I0711 01:37:30.830328 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33816\n",
            "I0711 01:37:33.766222 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340446\n",
            "I0711 01:37:33.767570 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44714\n",
            "I0711 01:37:36.709896 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339709\n",
            "I0711 01:37:36.711258 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43534\n",
            "I0711 01:37:39.669990 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337826\n",
            "I0711 01:37:39.671591 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40522\n",
            "I0711 01:37:42.618319 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339181\n",
            "I0711 01:37:42.620114 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4269\n",
            "I0711 01:37:45.531568 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343252\n",
            "I0711 01:37:45.533143 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49203\n",
            "I0711 01:37:48.525575 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.8784065 (29.704 sec)\n",
            "I0711 01:37:48.533487 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333126\n",
            "I0711 01:37:48.535107 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33002\n",
            "I0711 01:37:51.513182 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335603\n",
            "I0711 01:37:51.514834 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36965\n",
            "I0711 01:37:54.501455 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334646\n",
            "I0711 01:37:54.502973 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35434\n",
            "I0711 01:37:57.492307 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334354\n",
            "I0711 01:37:57.494216 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34966\n",
            "I0711 01:38:00.447891 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338336\n",
            "I0711 01:38:00.449301 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41338\n",
            "I0711 01:38:03.395960 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339215\n",
            "I0711 01:38:03.400130 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42744\n",
            "I0711 01:38:06.369321 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336327\n",
            "I0711 01:38:06.371151 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38122\n",
            "I0711 01:38:09.320896 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338788\n",
            "I0711 01:38:09.323454 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42062\n",
            "I0711 01:38:12.261502 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340063\n",
            "I0711 01:38:12.263059 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44101\n",
            "I0711 01:38:15.271261 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33226\n",
            "I0711 01:38:15.273021 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31616\n",
            "I0711 01:38:18.224985 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.471735 (29.699 sec)\n",
            "I0711 01:38:18.227895 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338216\n",
            "I0711 01:38:18.232411 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41145\n",
            "I0711 01:38:21.176007 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3392\n",
            "I0711 01:38:21.177691 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4272\n",
            "I0711 01:38:24.193949 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331353\n",
            "I0711 01:38:24.195469 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30166\n",
            "I0711 01:38:27.154944 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337725\n",
            "I0711 01:38:27.156812 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40359\n",
            "I0711 01:38:30.140393 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334958\n",
            "I0711 01:38:30.142023 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35933\n",
            "I0711 01:38:33.071979 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341108\n",
            "I0711 01:38:33.073559 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45773\n",
            "I0711 01:38:35.999736 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341558\n",
            "I0711 01:38:36.001328 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46492\n",
            "I0711 01:38:38.913866 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343163\n",
            "I0711 01:38:38.915529 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4906\n",
            "I0711 01:38:41.813147 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344911\n",
            "I0711 01:38:41.814529 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51857\n",
            "I0711 01:38:44.743226 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341294\n",
            "I0711 01:38:44.744913 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4607\n",
            "I0711 01:38:47.703282 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2462244 (29.478 sec)\n",
            "I0711 01:38:47.708783 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337225\n",
            "I0711 01:38:47.710791 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39559\n",
            "I0711 01:38:50.680048 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33653\n",
            "I0711 01:38:50.681852 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38448\n",
            "I0711 01:38:53.631466 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338817\n",
            "I0711 01:38:53.633166 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42107\n",
            "I0711 01:38:56.581157 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339025\n",
            "I0711 01:38:56.582851 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42441\n",
            "I0711 01:38:59.574831 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334031\n",
            "I0711 01:38:59.576394 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3445\n",
            "I0711 01:39:02.548251 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336318\n",
            "I0711 01:39:02.549865 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38109\n",
            "I0711 01:39:05.615579 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.326014\n",
            "I0711 01:39:05.617146 140338823006080 tpu_estimator.py:2160] examples/sec: 5.21623\n",
            "I0711 01:39:08.632056 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331512\n",
            "I0711 01:39:08.633747 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30419\n",
            "I0711 01:39:11.638122 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332665\n",
            "I0711 01:39:11.639842 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32264\n",
            "I0711 01:39:14.676894 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329077\n",
            "I0711 01:39:14.678213 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26523\n",
            "I0711 01:39:17.628412 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.6407909 (29.925 sec)\n",
            "I0711 01:39:17.630806 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338535\n",
            "I0711 01:39:17.634537 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41656\n",
            "I0711 01:39:20.651577 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331047\n",
            "I0711 01:39:20.653032 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29675\n",
            "I0711 01:39:23.604797 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338606\n",
            "I0711 01:39:23.606129 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4177\n",
            "I0711 01:39:26.567279 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337556\n",
            "I0711 01:39:26.569004 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4009\n",
            "I0711 01:39:29.569600 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333078\n",
            "I0711 01:39:29.571194 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32925\n",
            "I0711 01:39:32.540589 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336584\n",
            "I0711 01:39:32.542474 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38535\n",
            "I0711 01:39:35.519193 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33573\n",
            "I0711 01:39:35.520894 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37168\n",
            "I0711 01:39:38.494794 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336073\n",
            "I0711 01:39:38.496479 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37716\n",
            "I0711 01:39:41.434471 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340164\n",
            "I0711 01:39:41.436209 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44263\n",
            "I0711 01:39:44.376454 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339903\n",
            "I0711 01:39:44.377787 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43845\n",
            "I0711 01:39:47.386225 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.057551 (29.758 sec)\n",
            "I0711 01:39:47.391816 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331637\n",
            "I0711 01:39:47.395263 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30619\n",
            "I0711 01:39:50.356131 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337351\n",
            "I0711 01:39:50.357614 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39761\n",
            "I0711 01:39:53.376710 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331076\n",
            "I0711 01:39:53.378157 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29721\n",
            "I0711 01:39:56.387715 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332101\n",
            "I0711 01:39:56.389144 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31361\n",
            "I0711 01:39:59.415750 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330247\n",
            "I0711 01:39:59.417383 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28395\n",
            "I0711 01:40:02.416569 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333238\n",
            "I0711 01:40:02.418239 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33181\n",
            "I0711 01:40:05.435462 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331247\n",
            "I0711 01:40:05.436997 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29995\n",
            "I0711 01:40:08.408822 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336322\n",
            "I0711 01:40:08.410193 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38114\n",
            "I0711 01:40:11.444649 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329402\n",
            "I0711 01:40:11.447037 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27044\n",
            "I0711 01:40:14.455419 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332142\n",
            "I0711 01:40:14.456928 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31427\n",
            "I0711 01:40:17.455043 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.891535 (30.069 sec)\n",
            "I0711 01:40:17.457529 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333089\n",
            "I0711 01:40:17.462827 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32943\n",
            "I0711 01:40:20.451217 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334057\n",
            "I0711 01:40:20.453178 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34492\n",
            "I0711 01:40:23.468903 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331367\n",
            "I0711 01:40:23.470414 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30187\n",
            "I0711 01:40:26.439030 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336685\n",
            "I0711 01:40:26.440357 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38695\n",
            "I0711 01:40:29.410521 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336528\n",
            "I0711 01:40:29.411820 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38445\n",
            "I0711 01:40:32.388156 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335843\n",
            "I0711 01:40:32.389490 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37349\n",
            "I0711 01:40:35.350638 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337552\n",
            "I0711 01:40:35.351989 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40083\n",
            "I0711 01:40:38.348858 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333528\n",
            "I0711 01:40:38.350648 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33645\n",
            "I0711 01:40:41.318140 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336791\n",
            "I0711 01:40:41.319440 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38865\n",
            "I0711 01:40:44.328242 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332212\n",
            "I0711 01:40:44.329995 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31539\n",
            "I0711 01:40:47.312558 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.0513704 (29.858 sec)\n",
            "I0711 01:40:47.318482 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334416\n",
            "I0711 01:40:47.320473 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35066\n",
            "I0711 01:40:50.356402 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329176\n",
            "I0711 01:40:50.357820 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26682\n",
            "I0711 01:40:53.318779 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337563\n",
            "I0711 01:40:53.320013 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40101\n",
            "I0711 01:40:56.253715 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340732\n",
            "I0711 01:40:56.255479 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45172\n",
            "I0711 01:40:59.219252 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337201\n",
            "I0711 01:40:59.220733 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39522\n",
            "I0711 01:41:02.189891 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336626\n",
            "I0711 01:41:02.191320 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38601\n",
            "I0711 01:41:05.179762 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334462\n",
            "I0711 01:41:05.181085 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35139\n",
            "I0711 01:41:08.182573 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333023\n",
            "I0711 01:41:08.183965 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32836\n",
            "I0711 01:41:11.150362 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336951\n",
            "I0711 01:41:11.151664 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39121\n",
            "I0711 01:41:14.147740 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33363\n",
            "I0711 01:41:14.149177 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33809\n",
            "I0711 01:41:17.175343 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.0464501 (29.863 sec)\n",
            "I0711 01:41:17.182347 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329573\n",
            "I0711 01:41:17.183754 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27316\n",
            "I0711 01:41:20.157438 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336074\n",
            "I0711 01:41:20.158975 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37719\n",
            "I0711 01:41:23.156768 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333413\n",
            "I0711 01:41:23.158061 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33461\n",
            "I0711 01:41:26.136209 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33563\n",
            "I0711 01:41:26.137892 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37008\n",
            "I0711 01:41:29.080029 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339697\n",
            "I0711 01:41:29.081557 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43515\n",
            "I0711 01:41:32.051663 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336512\n",
            "I0711 01:41:32.053170 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3842\n",
            "I0711 01:41:35.020024 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336884\n",
            "I0711 01:41:35.021679 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39015\n",
            "I0711 01:41:37.994032 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33626\n",
            "I0711 01:41:37.995576 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38016\n",
            "I0711 01:41:40.959697 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337189\n",
            "I0711 01:41:40.961135 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39503\n",
            "I0711 01:41:43.936474 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335925\n",
            "I0711 01:41:43.938017 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3748\n",
            "I0711 01:41:46.893658 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.84066916 (29.718 sec)\n",
            "I0711 01:41:46.896361 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33785\n",
            "I0711 01:41:46.900713 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4056\n",
            "I0711 01:41:49.871488 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336121\n",
            "I0711 01:41:49.872927 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37794\n",
            "I0711 01:41:52.849222 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335829\n",
            "I0711 01:41:52.850820 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37326\n",
            "I0711 01:41:55.804029 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338433\n",
            "I0711 01:41:55.805310 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41493\n",
            "I0711 01:41:58.773420 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336769\n",
            "I0711 01:41:58.774733 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3883\n",
            "I0711 01:42:01.764586 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334313\n",
            "I0711 01:42:01.765862 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34901\n",
            "I0711 01:42:04.734966 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336661\n",
            "I0711 01:42:04.736675 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38657\n",
            "I0711 01:42:07.721109 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334882\n",
            "I0711 01:42:07.722577 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3581\n",
            "I0711 01:42:10.667980 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339346\n",
            "I0711 01:42:10.669934 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42954\n",
            "I0711 01:42:13.651144 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335212\n",
            "I0711 01:42:13.653608 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3634\n",
            "I0711 01:42:16.617467 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2219353 (29.724 sec)\n",
            "I0711 01:42:16.621793 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336632\n",
            "I0711 01:42:16.625170 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38611\n",
            "I0711 01:42:19.607161 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334967\n",
            "I0711 01:42:19.608718 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35946\n",
            "I0711 01:42:22.567836 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337757\n",
            "I0711 01:42:22.569219 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40412\n",
            "I0711 01:42:25.559858 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334216\n",
            "I0711 01:42:25.561401 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34745\n",
            "I0711 01:42:28.572909 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331895\n",
            "I0711 01:42:28.574359 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31032\n",
            "I0711 01:42:31.528253 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338368\n",
            "I0711 01:42:31.529567 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41388\n",
            "I0711 01:42:34.500931 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336393\n",
            "I0711 01:42:34.502298 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38229\n",
            "I0711 01:42:37.456414 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338355\n",
            "I0711 01:42:37.457830 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41368\n",
            "I0711 01:42:40.426288 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336724\n",
            "I0711 01:42:40.427899 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38758\n",
            "I0711 01:42:43.430274 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332889\n",
            "I0711 01:42:43.431629 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32622\n",
            "I0711 01:42:46.422646 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.0997863 (29.805 sec)\n",
            "I0711 01:42:46.429894 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333373\n",
            "I0711 01:42:46.431413 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33396\n",
            "I0711 01:42:49.431827 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333125\n",
            "I0711 01:42:49.433248 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33\n",
            "I0711 01:42:52.446386 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331717\n",
            "I0711 01:42:52.447948 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30746\n",
            "I0711 01:42:55.439528 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334095\n",
            "I0711 01:42:55.440913 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34553\n",
            "I0711 01:42:58.416311 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335945\n",
            "I0711 01:42:58.417809 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37511\n",
            "I0711 01:43:01.375113 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337967\n",
            "I0711 01:43:01.376539 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40747\n",
            "I0711 01:43:04.334091 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337956\n",
            "I0711 01:43:04.335632 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4073\n",
            "I0711 01:43:07.331346 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333641\n",
            "I0711 01:43:07.332950 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33826\n",
            "I0711 01:43:10.308635 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335878\n",
            "I0711 01:43:10.310094 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37405\n",
            "I0711 01:43:13.290265 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33538\n",
            "I0711 01:43:13.292046 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36607\n",
            "I0711 01:43:16.279468 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.885793 (29.857 sec)\n",
            "I0711 01:43:16.281823 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334268\n",
            "I0711 01:43:16.283744 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34829\n",
            "I0711 01:43:19.238257 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33826\n",
            "I0711 01:43:19.240128 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41216\n",
            "I0711 01:43:22.199848 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337647\n",
            "I0711 01:43:22.201705 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40235\n",
            "I0711 01:43:25.195687 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333796\n",
            "I0711 01:43:25.197113 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34074\n",
            "I0711 01:43:28.161023 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337242\n",
            "I0711 01:43:28.162545 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39587\n",
            "I0711 01:43:31.127985 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337037\n",
            "I0711 01:43:31.129670 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39259\n",
            "I0711 01:43:34.094494 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337089\n",
            "I0711 01:43:34.096084 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39343\n",
            "I0711 01:43:37.055846 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337685\n",
            "I0711 01:43:37.057318 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40297\n",
            "I0711 01:43:40.008750 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338649\n",
            "I0711 01:43:40.010480 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41839\n",
            "I0711 01:43:43.023469 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331707\n",
            "I0711 01:43:43.024922 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30731\n",
            "I0711 01:43:45.957196 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.7883735 (29.678 sec)\n",
            "I0711 01:43:45.965221 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339963\n",
            "I0711 01:43:45.966782 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43941\n",
            "I0711 01:43:48.915130 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33897\n",
            "I0711 01:43:48.916639 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42352\n",
            "I0711 01:43:51.879820 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337301\n",
            "I0711 01:43:51.881235 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39682\n",
            "I0711 01:43:54.863115 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335209\n",
            "I0711 01:43:54.864589 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36335\n",
            "I0711 01:43:57.807443 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339621\n",
            "I0711 01:43:57.808812 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43394\n",
            "I0711 01:44:00.728460 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342353\n",
            "I0711 01:44:00.729783 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47765\n",
            "I0711 01:44:03.750632 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330881\n",
            "I0711 01:44:03.755805 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2941\n",
            "I0711 01:44:06.738206 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334727\n",
            "I0711 01:44:06.739742 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35563\n",
            "I0711 01:44:09.742220 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332886\n",
            "I0711 01:44:09.743609 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32618\n",
            "I0711 01:44:12.709016 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337061\n",
            "I0711 01:44:12.710489 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39297\n",
            "I0711 01:44:15.649024 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0693243 (29.692 sec)\n",
            "I0711 01:44:15.652872 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339689\n",
            "I0711 01:44:15.655352 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43502\n",
            "I0711 01:44:18.653386 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33328\n",
            "I0711 01:44:18.655333 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33248\n",
            "I0711 01:44:21.596325 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339793\n",
            "I0711 01:44:21.597687 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43669\n",
            "I0711 01:44:24.512807 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342877\n",
            "I0711 01:44:24.514560 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48604\n",
            "I0711 01:44:27.509172 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33374\n",
            "I0711 01:44:27.510588 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33984\n",
            "I0711 01:44:30.489400 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335547\n",
            "I0711 01:44:30.491015 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36875\n",
            "I0711 01:44:33.470912 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335398\n",
            "I0711 01:44:33.472524 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36637\n",
            "I0711 01:44:36.407295 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340559\n",
            "I0711 01:44:36.408904 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44894\n",
            "I0711 01:44:39.352909 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339481\n",
            "I0711 01:44:39.354592 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4317\n",
            "I0711 01:44:42.374826 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330922\n",
            "I0711 01:44:42.376468 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29475\n",
            "I0711 01:44:45.340992 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.230225 (29.692 sec)\n",
            "I0711 01:44:45.343993 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336791\n",
            "I0711 01:44:45.348670 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38865\n",
            "I0711 01:44:48.318010 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336246\n",
            "I0711 01:44:48.319598 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37994\n",
            "I0711 01:44:51.321550 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332937\n",
            "I0711 01:44:51.323121 140338823006080 tpu_estimator.py:2160] examples/sec: 5.327\n",
            "I0711 01:44:54.318708 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333652\n",
            "I0711 01:44:54.320303 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33843\n",
            "I0711 01:44:57.293981 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336102\n",
            "I0711 01:44:57.295680 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37764\n",
            "I0711 01:45:00.271543 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335844\n",
            "I0711 01:45:00.272857 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37351\n",
            "I0711 01:45:03.206366 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340738\n",
            "I0711 01:45:03.208094 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45181\n",
            "I0711 01:45:06.187616 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335439\n",
            "I0711 01:45:06.189220 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36702\n",
            "I0711 01:45:09.157689 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336684\n",
            "I0711 01:45:09.159616 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38695\n",
            "I0711 01:45:12.115277 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338113\n",
            "I0711 01:45:12.116602 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40981\n",
            "I0711 01:45:15.075728 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.060453 (29.735 sec)\n",
            "I0711 01:45:15.082122 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337057\n",
            "I0711 01:45:15.083769 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39291\n",
            "I0711 01:45:18.075687 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334054\n",
            "I0711 01:45:18.077411 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34487\n",
            "I0711 01:45:21.054341 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335722\n",
            "I0711 01:45:21.056327 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37155\n",
            "I0711 01:45:24.056803 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333061\n",
            "I0711 01:45:24.058372 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32897\n",
            "I0711 01:45:27.034596 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335812\n",
            "I0711 01:45:27.035956 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37299\n",
            "I0711 01:45:30.018349 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335163\n",
            "I0711 01:45:30.019933 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36261\n",
            "I0711 01:45:33.048287 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330031\n",
            "I0711 01:45:33.049733 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2805\n",
            "I0711 01:45:36.031572 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335196\n",
            "I0711 01:45:36.033201 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36314\n",
            "I0711 01:45:39.028920 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333635\n",
            "I0711 01:45:39.030356 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33816\n",
            "I0711 01:45:42.046972 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331334\n",
            "I0711 01:45:42.048424 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30135\n",
            "I0711 01:45:45.043728 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.5563898 (29.968 sec)\n",
            "I0711 01:45:45.050584 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332937\n",
            "I0711 01:45:45.052486 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32699\n",
            "I0711 01:45:48.071751 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330994\n",
            "I0711 01:45:48.073163 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29591\n",
            "I0711 01:45:51.074681 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333007\n",
            "I0711 01:45:51.076483 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32811\n",
            "I0711 01:45:54.088357 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331821\n",
            "I0711 01:45:54.089771 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30914\n",
            "I0711 01:45:57.044850 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338242\n",
            "I0711 01:45:57.046607 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41187\n",
            "I0711 01:46:00.208326 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.316112\n",
            "I0711 01:46:00.209811 140338823006080 tpu_estimator.py:2160] examples/sec: 5.05779\n",
            "I0711 01:46:03.211278 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332999\n",
            "I0711 01:46:03.212855 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32798\n",
            "I0711 01:46:06.181031 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336733\n",
            "I0711 01:46:06.182674 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38773\n",
            "I0711 01:46:09.165960 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335014\n",
            "I0711 01:46:09.167938 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36023\n",
            "I0711 01:46:12.185959 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331123\n",
            "I0711 01:46:12.187443 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29796\n",
            "I0711 01:46:15.134248 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.2701516 (30.091 sec)\n",
            "I0711 01:46:15.144053 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338053\n",
            "I0711 01:46:15.145507 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40886\n",
            "I0711 01:46:18.149394 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332743\n",
            "I0711 01:46:18.151621 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32388\n",
            "I0711 01:46:21.186267 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.32929\n",
            "I0711 01:46:21.187928 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26863\n",
            "I0711 01:46:24.137543 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338834\n",
            "I0711 01:46:24.139015 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42134\n",
            "I0711 01:46:27.118287 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335486\n",
            "I0711 01:46:27.119539 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36777\n",
            "I0711 01:46:30.126901 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332382\n",
            "I0711 01:46:30.128389 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31811\n",
            "I0711 01:46:33.125177 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333523\n",
            "I0711 01:46:33.126452 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33637\n",
            "I0711 01:46:36.097265 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336464\n",
            "I0711 01:46:36.098566 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38343\n",
            "I0711 01:46:39.064697 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336994\n",
            "I0711 01:46:39.066272 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3919\n",
            "I0711 01:46:42.047809 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335214\n",
            "I0711 01:46:42.049157 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36343\n",
            "I0711 01:46:45.045415 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.5977225 (29.911 sec)\n",
            "I0711 01:46:45.051734 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332908\n",
            "I0711 01:46:45.053259 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32653\n",
            "I0711 01:46:48.044365 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334158\n",
            "I0711 01:46:48.046238 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34653\n",
            "I0711 01:46:51.010369 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33714\n",
            "I0711 01:46:51.011951 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39424\n",
            "I0711 01:46:53.969458 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337944\n",
            "I0711 01:46:53.970875 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4071\n",
            "I0711 01:46:56.917132 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339251\n",
            "I0711 01:46:56.918423 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42801\n",
            "I0711 01:46:59.891138 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336251\n",
            "I0711 01:46:59.892458 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38002\n",
            "I0711 01:47:02.953174 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.326592\n",
            "I0711 01:47:02.955139 140338823006080 tpu_estimator.py:2160] examples/sec: 5.22548\n",
            "I0711 01:47:05.944131 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334325\n",
            "I0711 01:47:05.945549 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34919\n",
            "I0711 01:47:08.922740 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335734\n",
            "I0711 01:47:08.924150 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37175\n",
            "I0711 01:47:11.925516 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33302\n",
            "I0711 01:47:11.927224 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32832\n",
            "I0711 01:47:14.940738 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.6630929 (29.895 sec)\n",
            "I0711 01:47:14.947887 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330897\n",
            "I0711 01:47:14.949446 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29435\n",
            "I0711 01:47:17.918030 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336655\n",
            "I0711 01:47:17.920213 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38648\n",
            "I0711 01:47:20.922133 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33288\n",
            "I0711 01:47:20.923859 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32608\n",
            "I0711 01:47:23.910141 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334671\n",
            "I0711 01:47:23.911545 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35473\n",
            "I0711 01:47:26.911649 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333161\n",
            "I0711 01:47:26.913174 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33058\n",
            "I0711 01:47:29.837665 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34176\n",
            "I0711 01:47:29.839267 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46816\n",
            "I0711 01:47:32.838546 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333244\n",
            "I0711 01:47:32.840031 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33191\n",
            "I0711 01:47:35.884815 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.32826\n",
            "I0711 01:47:35.886195 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25216\n",
            "I0711 01:47:38.889920 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332773\n",
            "I0711 01:47:38.891505 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32436\n",
            "I0711 01:47:41.875932 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33489\n",
            "I0711 01:47:41.877690 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35824\n",
            "I0711 01:47:44.880966 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.7872014 (29.940 sec)\n",
            "I0711 01:47:44.883421 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332504\n",
            "I0711 01:47:44.885066 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32007\n",
            "I0711 01:47:47.881368 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333559\n",
            "I0711 01:47:47.883129 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33695\n",
            "I0711 01:47:50.934697 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.327516\n",
            "I0711 01:47:50.936143 140338823006080 tpu_estimator.py:2160] examples/sec: 5.24026\n",
            "I0711 01:47:53.942569 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332456\n",
            "I0711 01:47:53.944137 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3193\n",
            "I0711 01:47:56.911447 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336828\n",
            "I0711 01:47:56.913251 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38924\n",
            "I0711 01:47:59.929694 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331324\n",
            "I0711 01:47:59.931354 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30119\n",
            "I0711 01:48:02.914967 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334977\n",
            "I0711 01:48:02.916558 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35963\n",
            "I0711 01:48:05.899831 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335024\n",
            "I0711 01:48:05.901157 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36038\n",
            "I0711 01:48:08.842334 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339847\n",
            "I0711 01:48:08.843724 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43755\n",
            "I0711 01:48:11.840629 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333523\n",
            "I0711 01:48:11.841964 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33637\n",
            "I0711 01:48:14.826866 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.1143312 (29.946 sec)\n",
            "I0711 01:48:14.831507 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334384\n",
            "I0711 01:48:14.835246 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35015\n",
            "I0711 01:48:17.798693 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336985\n",
            "I0711 01:48:17.800159 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39176\n",
            "I0711 01:48:20.769493 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336609\n",
            "I0711 01:48:20.772104 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38575\n",
            "I0711 01:48:23.739733 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336669\n",
            "I0711 01:48:23.741162 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3867\n",
            "I0711 01:48:26.743423 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332922\n",
            "I0711 01:48:26.745114 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32675\n",
            "I0711 01:48:29.740417 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33367\n",
            "I0711 01:48:29.742382 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33872\n",
            "I0711 01:48:32.739959 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333381\n",
            "I0711 01:48:32.741882 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3341\n",
            "I0711 01:48:35.714806 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336153\n",
            "I0711 01:48:35.716564 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37845\n",
            "I0711 01:48:38.680266 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337223\n",
            "I0711 01:48:38.681797 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39557\n",
            "I0711 01:48:41.637545 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338145\n",
            "I0711 01:48:41.638951 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41033\n",
            "I0711 01:48:44.637200 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.1592255 (29.810 sec)\n",
            "I0711 01:48:44.642382 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332791\n",
            "I0711 01:48:44.645763 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32466\n",
            "I0711 01:48:47.604731 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337571\n",
            "I0711 01:48:47.606110 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40113\n",
            "I0711 01:48:50.612606 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332463\n",
            "I0711 01:48:50.614044 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31941\n",
            "I0711 01:48:53.607922 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333852\n",
            "I0711 01:48:53.609381 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34163\n",
            "I0711 01:48:56.588189 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335548\n",
            "I0711 01:48:56.589952 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36877\n",
            "I0711 01:48:59.563578 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336082\n",
            "I0711 01:48:59.565191 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37732\n",
            "I0711 01:49:02.560588 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333671\n",
            "I0711 01:49:02.562059 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33873\n",
            "I0711 01:49:05.521614 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33772\n",
            "I0711 01:49:05.522956 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40352\n",
            "I0711 01:49:08.450739 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341404\n",
            "I0711 01:49:08.452132 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46246\n",
            "I0711 01:49:11.431911 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335431\n",
            "I0711 01:49:11.433259 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36689\n",
            "I0711 01:49:14.432530 140338823006080 basic_session_run_hooks.py:260] batch_loss = 10.355345 (29.795 sec)\n",
            "I0711 01:49:14.439048 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332571\n",
            "I0711 01:49:14.441300 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32113\n",
            "I0711 01:49:17.438511 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333373\n",
            "I0711 01:49:17.441009 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33397\n",
            "I0711 01:49:20.417520 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335675\n",
            "I0711 01:49:20.418857 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3708\n",
            "I0711 01:49:23.412607 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333877\n",
            "I0711 01:49:23.414146 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34203\n",
            "I0711 01:49:26.382338 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336739\n",
            "I0711 01:49:26.383764 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38782\n",
            "I0711 01:49:29.362296 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335572\n",
            "I0711 01:49:29.363594 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36915\n",
            "I0711 01:49:32.311576 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339066\n",
            "I0711 01:49:32.313187 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42506\n",
            "I0711 01:49:35.272689 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33771\n",
            "I0711 01:49:35.274305 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40335\n",
            "I0711 01:49:38.214285 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339953\n",
            "I0711 01:49:38.215872 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43924\n",
            "I0711 01:49:41.195901 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335387\n",
            "I0711 01:49:41.197410 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3662\n",
            "I0711 01:49:44.161375 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0636771 (29.729 sec)\n",
            "I0711 01:49:44.164616 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336848\n",
            "I0711 01:49:44.167618 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38956\n",
            "I0711 01:49:47.142160 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335848\n",
            "I0711 01:49:47.143563 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37356\n",
            "I0711 01:49:50.097405 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338377\n",
            "I0711 01:49:50.098825 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41403\n",
            "I0711 01:49:53.102324 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332801\n",
            "I0711 01:49:53.104452 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32482\n",
            "I0711 01:49:56.078225 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336023\n",
            "I0711 01:49:56.079774 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37637\n",
            "I0711 01:49:59.019099 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340037\n",
            "I0711 01:49:59.020778 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44059\n",
            "I0711 01:50:01.996510 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335858\n",
            "I0711 01:50:01.998116 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37372\n",
            "I0711 01:50:04.944600 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339213\n",
            "I0711 01:50:04.946348 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42741\n",
            "I0711 01:50:07.898234 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33856\n",
            "I0711 01:50:07.899989 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41696\n",
            "I0711 01:50:10.859762 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337664\n",
            "I0711 01:50:10.861276 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40263\n",
            "I0711 01:50:13.834064 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2491674 (29.673 sec)\n",
            "I0711 01:50:13.843268 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335265\n",
            "I0711 01:50:13.844682 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36424\n",
            "I0711 01:50:16.791334 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339109\n",
            "I0711 01:50:16.792573 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42574\n",
            "I0711 01:50:19.761542 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336678\n",
            "I0711 01:50:19.763180 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38685\n",
            "I0711 01:50:22.715910 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338484\n",
            "I0711 01:50:22.718328 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41574\n",
            "I0711 01:50:25.646133 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341272\n",
            "I0711 01:50:25.647628 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46036\n",
            "I0711 01:50:28.600723 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338454\n",
            "I0711 01:50:28.602477 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41527\n",
            "I0711 01:50:31.566845 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337142\n",
            "I0711 01:50:31.568297 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39427\n",
            "I0711 01:50:34.513346 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339389\n",
            "I0711 01:50:34.514951 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43022\n",
            "I0711 01:50:37.460322 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339331\n",
            "I0711 01:50:37.462043 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4293\n",
            "I0711 01:50:40.465444 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332757\n",
            "I0711 01:50:40.466894 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32411\n",
            "I0711 01:50:43.434595 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.79689217 (29.601 sec)\n",
            "I0711 01:50:43.437097 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336516\n",
            "I0711 01:50:43.438585 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38425\n",
            "I0711 01:50:46.393415 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338266\n",
            "I0711 01:50:46.395241 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41225\n",
            "I0711 01:50:49.333536 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340117\n",
            "I0711 01:50:49.335404 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44187\n",
            "I0711 01:50:52.275691 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33989\n",
            "I0711 01:50:52.277133 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43824\n",
            "I0711 01:50:55.191534 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342958\n",
            "I0711 01:50:55.192942 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48732\n",
            "I0711 01:50:58.105813 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34313\n",
            "I0711 01:50:58.107091 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49009\n",
            "I0711 01:51:01.056929 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338866\n",
            "I0711 01:51:01.058478 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42185\n",
            "I0711 01:51:03.993353 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340544\n",
            "I0711 01:51:03.995098 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44871\n",
            "I0711 01:51:06.937810 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339618\n",
            "I0711 01:51:06.939427 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43389\n",
            "I0711 01:51:09.871358 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34088\n",
            "I0711 01:51:09.872695 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45408\n",
            "I0711 01:51:12.813969 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.6647692 (29.379 sec)\n",
            "I0711 01:51:12.817543 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339423\n",
            "I0711 01:51:12.821572 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43078\n",
            "I0711 01:51:15.790443 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336372\n",
            "I0711 01:51:15.792041 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38196\n",
            "I0711 01:51:18.751333 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33774\n",
            "I0711 01:51:18.753023 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40384\n",
            "I0711 01:51:21.696598 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339521\n",
            "I0711 01:51:21.698150 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43234\n",
            "I0711 01:51:24.652468 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33831\n",
            "I0711 01:51:24.653896 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41296\n",
            "I0711 01:51:27.597318 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339576\n",
            "I0711 01:51:27.598842 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43321\n",
            "I0711 01:51:30.580770 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335182\n",
            "I0711 01:51:30.582191 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36291\n",
            "I0711 01:51:33.509997 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341398\n",
            "I0711 01:51:33.511511 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46237\n",
            "I0711 01:51:36.571551 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.326627\n",
            "I0711 01:51:36.576658 140338823006080 tpu_estimator.py:2160] examples/sec: 5.22604\n",
            "I0711 01:51:39.549993 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335751\n",
            "I0711 01:51:39.551557 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37202\n",
            "I0711 01:51:42.506908 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0878384 (29.693 sec)\n",
            "I0711 01:51:42.513769 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337411\n",
            "I0711 01:51:42.516912 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39857\n",
            "I0711 01:51:45.467426 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338553\n",
            "I0711 01:51:45.469138 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41684\n",
            "I0711 01:51:48.441062 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336288\n",
            "I0711 01:51:48.442754 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3806\n",
            "I0711 01:51:51.417321 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335994\n",
            "I0711 01:51:51.418730 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37591\n",
            "I0711 01:51:54.362114 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339582\n",
            "I0711 01:51:54.363766 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43332\n",
            "I0711 01:51:57.342792 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335497\n",
            "I0711 01:51:57.344286 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36795\n",
            "I0711 01:52:00.317019 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336216\n",
            "I0711 01:52:00.318354 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37945\n",
            "I0711 01:52:03.267921 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338888\n",
            "I0711 01:52:03.269391 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42221\n",
            "I0711 01:52:06.215779 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339229\n",
            "I0711 01:52:06.217337 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42766\n",
            "I0711 01:52:09.211130 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333847\n",
            "I0711 01:52:09.212548 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34155\n",
            "I0711 01:52:12.199057 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.6612536 (29.692 sec)\n",
            "I0711 01:52:12.205538 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33401\n",
            "I0711 01:52:12.207547 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34416\n",
            "I0711 01:52:15.243157 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329152\n",
            "I0711 01:52:15.244661 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26643\n",
            "I0711 01:52:18.254818 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332048\n",
            "I0711 01:52:18.256287 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31277\n",
            "I0711 01:52:21.271050 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331535\n",
            "I0711 01:52:21.272944 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30455\n",
            "I0711 01:52:24.282916 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332021\n",
            "I0711 01:52:24.285470 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31234\n",
            "I0711 01:52:27.298663 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331593\n",
            "I0711 01:52:27.300064 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30549\n",
            "I0711 01:52:30.299878 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333197\n",
            "I0711 01:52:30.301265 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33115\n",
            "I0711 01:52:33.282013 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33533\n",
            "I0711 01:52:33.283645 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36529\n",
            "I0711 01:52:36.305023 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330797\n",
            "I0711 01:52:36.306364 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29275\n",
            "I0711 01:52:39.306279 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333189\n",
            "I0711 01:52:39.308046 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33102\n",
            "I0711 01:52:42.308748 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.735527 (30.110 sec)\n",
            "I0711 01:52:42.312698 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332631\n",
            "I0711 01:52:42.317328 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32209\n",
            "I0711 01:52:45.312291 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333375\n",
            "I0711 01:52:45.313712 140338823006080 tpu_estimator.py:2160] examples/sec: 5.334\n",
            "I0711 01:52:48.317396 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33276\n",
            "I0711 01:52:48.318959 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32416\n",
            "I0711 01:52:51.312743 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333854\n",
            "I0711 01:52:51.314197 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34166\n",
            "I0711 01:52:54.263916 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338854\n",
            "I0711 01:52:54.265346 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42166\n",
            "I0711 01:52:57.255036 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33435\n",
            "I0711 01:52:57.256611 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3496\n",
            "I0711 01:53:00.231616 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335922\n",
            "I0711 01:53:00.233030 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37475\n",
            "I0711 01:53:03.210825 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33566\n",
            "I0711 01:53:03.212169 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37056\n",
            "I0711 01:53:06.196099 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33498\n",
            "I0711 01:53:06.197784 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35967\n",
            "I0711 01:53:09.195606 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333384\n",
            "I0711 01:53:09.197308 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33414\n",
            "I0711 01:53:12.148511 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.90118265 (29.840 sec)\n",
            "I0711 01:53:12.151296 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338342\n",
            "I0711 01:53:12.152681 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41347\n",
            "I0711 01:53:15.118944 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336957\n",
            "I0711 01:53:15.120311 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39132\n",
            "I0711 01:53:18.092853 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33626\n",
            "I0711 01:53:18.094543 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38016\n",
            "I0711 01:53:21.087546 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333922\n",
            "I0711 01:53:21.088992 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34275\n",
            "I0711 01:53:24.160384 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.325437\n",
            "I0711 01:53:24.161824 140338823006080 tpu_estimator.py:2160] examples/sec: 5.207\n",
            "I0711 01:53:27.158849 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333504\n",
            "I0711 01:53:27.160428 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33607\n",
            "I0711 01:53:30.159967 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333209\n",
            "I0711 01:53:30.161770 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33135\n",
            "I0711 01:53:33.167722 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332467\n",
            "I0711 01:53:33.169057 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31946\n",
            "I0711 01:53:36.159522 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334251\n",
            "I0711 01:53:36.161234 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34801\n",
            "I0711 01:53:39.126194 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337089\n",
            "I0711 01:53:39.127785 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39342\n",
            "I0711 01:53:42.110671 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2961059 (29.962 sec)\n",
            "I0711 01:53:42.118084 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334225\n",
            "I0711 01:53:42.120226 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34759\n",
            "I0711 01:53:45.099727 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335392\n",
            "I0711 01:53:45.101369 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36627\n",
            "I0711 01:53:48.116802 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331449\n",
            "I0711 01:53:48.118110 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30318\n",
            "I0711 01:53:51.143528 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330381\n",
            "I0711 01:53:51.144896 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28609\n",
            "I0711 01:53:54.150973 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332514\n",
            "I0711 01:53:54.158463 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32022\n",
            "I0711 01:53:57.131904 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335463\n",
            "I0711 01:53:57.133402 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36741\n",
            "I0711 01:54:00.165599 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329631\n",
            "I0711 01:54:00.166945 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2741\n",
            "I0711 01:54:03.164853 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333418\n",
            "I0711 01:54:03.166167 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33469\n",
            "I0711 01:54:06.181393 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3315\n",
            "I0711 01:54:06.182656 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30401\n",
            "I0711 01:54:09.143956 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337547\n",
            "I0711 01:54:09.145776 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40076\n",
            "I0711 01:54:12.129506 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.337727 (30.019 sec)\n",
            "I0711 01:54:12.134361 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334399\n",
            "I0711 01:54:12.137898 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35039\n",
            "I0711 01:54:15.140454 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332664\n",
            "I0711 01:54:15.142312 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32262\n",
            "I0711 01:54:18.104307 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337398\n",
            "I0711 01:54:18.105754 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39837\n",
            "I0711 01:54:21.080295 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336024\n",
            "I0711 01:54:21.081768 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37638\n",
            "I0711 01:54:24.057405 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33589\n",
            "I0711 01:54:24.058858 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37424\n",
            "I0711 01:54:27.018110 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337763\n",
            "I0711 01:54:27.020638 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4042\n",
            "I0711 01:54:30.004498 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334857\n",
            "I0711 01:54:30.006283 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35771\n",
            "I0711 01:54:32.986171 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335383\n",
            "I0711 01:54:32.987726 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36613\n",
            "I0711 01:54:35.984548 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333515\n",
            "I0711 01:54:35.986203 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33623\n",
            "I0711 01:54:38.960847 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335979\n",
            "I0711 01:54:38.962418 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37567\n",
            "I0711 01:54:41.913810 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.1454513 (29.784 sec)\n",
            "I0711 01:54:41.922758 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337616\n",
            "I0711 01:54:41.924725 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40186\n",
            "I0711 01:54:44.894799 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336481\n",
            "I0711 01:54:44.896264 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38369\n",
            "I0711 01:54:47.859221 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337325\n",
            "I0711 01:54:47.861084 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39721\n",
            "I0711 01:54:50.832335 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336346\n",
            "I0711 01:54:50.833785 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38153\n",
            "I0711 01:54:53.791765 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337912\n",
            "I0711 01:54:53.793357 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40659\n",
            "I0711 01:54:56.752614 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337735\n",
            "I0711 01:54:56.754313 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40376\n",
            "I0711 01:54:59.693698 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340018\n",
            "I0711 01:54:59.695403 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44029\n",
            "I0711 01:55:02.665569 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33649\n",
            "I0711 01:55:02.667186 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38384\n",
            "I0711 01:55:05.646679 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335434\n",
            "I0711 01:55:05.648250 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36694\n",
            "I0711 01:55:08.685683 140338823006080 basic_session_run_hooks.py:606] Saving checkpoints for 1000 into /content/drive/My Drive/ner_checkpoints/model.ckpt.\n",
            "I0711 01:55:38.830018 140338823006080 estimator.py:1145] Calling model_fn.\n",
            "I0711 01:55:38.832021 140338823006080 tpu_estimator.py:2965] Running eval on CPU\n",
            "I0711 01:55:38.833962 140338823006080 <ipython-input-9-2f5db7366c31>:282] *** Features ***\n",
            "I0711 01:55:38.843168 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_ids, shape = (?, 128)\n",
            "I0711 01:55:38.846459 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_mask, shape = (?, 128)\n",
            "I0711 01:55:38.849276 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = label_ids, shape = (?, 128)\n",
            "I0711 01:55:38.854788 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = segment_ids, shape = (?, 128)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "shape of input_ids (?, 128)\n",
            "shape of label_ids (?, 128)\n",
            "seq_length 128\n",
            "lengths Tensor(\"Sum:0\", shape=(?,), dtype=int32)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "W0711 01:55:45.818780 140338823006080 deprecation_wrapper.py:119] From bert-ner/tf_metrics.py:151: The name tf.diag_part is deprecated. Please use tf.linalg.tensor_diag_part instead.\n",
            "\n",
            "W0711 01:55:45.835066 140338823006080 deprecation.py:323] From bert-ner/tf_metrics.py:139: to_float (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"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "####################\n",
            "shape of output_layer: (?, 128, 1024)\n",
            "embedding_size:1024\n",
            "seq_length:128\n",
            "shape of logit (?, 128, 11)\n",
            "shape of loss ()\n",
            "shape of per_example_loss (?,)\n",
            "num labels:11\n",
            "####################\n",
            "shape of pred_ids (?, 128)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "I0711 01:55:46.294131 140338823006080 estimator.py:1147] Done calling model_fn.\n",
            "I0711 01:55:46.324961 140338823006080 evaluation.py:255] Starting evaluation at 2019-07-11T01:55:46Z\n",
            "I0711 01:55:47.568299 140338823006080 monitored_session.py:240] Graph was finalized.\n",
            "I0711 01:55:47.582003 140338823006080 saver.py:1280] Restoring parameters from /content/drive/My Drive/ner_checkpoints/model.ckpt-1000\n",
            "I0711 01:55:55.701409 140338823006080 session_manager.py:500] Running local_init_op.\n",
            "I0711 01:55:55.803561 140338823006080 session_manager.py:502] Done running local_init_op.\n",
            "I0711 01:56:02.926589 140338823006080 evaluation.py:167] Evaluation [10/100]\n",
            "I0711 01:56:08.176839 140338823006080 evaluation.py:167] Evaluation [20/100]\n",
            "I0711 01:56:13.389809 140338823006080 evaluation.py:167] Evaluation [30/100]\n",
            "I0711 01:56:18.638427 140338823006080 evaluation.py:167] Evaluation [40/100]\n",
            "I0711 01:56:23.982215 140338823006080 evaluation.py:167] Evaluation [50/100]\n",
            "I0711 01:56:28.843925 140338823006080 evaluation.py:167] Evaluation [60/100]\n",
            "I0711 01:56:33.725836 140338823006080 evaluation.py:167] Evaluation [70/100]\n",
            "I0711 01:56:38.994539 140338823006080 evaluation.py:167] Evaluation [80/100]\n",
            "I0711 01:56:44.276308 140338823006080 evaluation.py:167] Evaluation [90/100]\n",
            "I0711 01:56:49.981227 140338823006080 evaluation.py:167] Evaluation [100/100]\n",
            "I0711 01:56:50.224136 140338823006080 evaluation.py:275] Finished evaluation at 2019-07-11-01:56:50\n",
            "I0711 01:56:50.225663 140338823006080 estimator.py:2039] Saving dict for global step 1000: eval_accuracy = 0.9853557, eval_f = 0.9264958, eval_loss = 1.358573, eval_precision = 0.93402267, eval_recall = 0.91908914, global_step = 1000, loss = 1.358573\n",
            "I0711 01:56:51.376900 140338823006080 estimator.py:2099] Saving 'checkpoint_path' summary for global step 1000: /content/drive/My Drive/ner_checkpoints/model.ckpt-1000\n",
            "I0711 01:56:51.384837 140338823006080 error_handling.py:96] evaluation_loop marked as finished\n",
            "W0711 01:56:51.392540 140338823006080 deprecation.py:323] From /usr/local/lib/python3.6/dist-packages/tensorflow/python/summary/summary_iterator.py:68: tf_record_iterator (from tensorflow.python.lib.io.tf_record) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use eager execution and: \n",
            "`tf.data.TFRecordDataset(path)`\n",
            "I0711 01:56:51.416856 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.00945446\n",
            "I0711 01:56:51.418735 140338823006080 tpu_estimator.py:2160] examples/sec: 0.151271\n",
            "I0711 01:56:54.459213 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.8239682 (132.545 sec)\n",
            "I0711 01:56:54.464311 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328146\n",
            "I0711 01:56:54.466696 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25034\n",
            "I0711 01:56:57.398672 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340795\n",
            "I0711 01:56:57.400171 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45272\n",
            "I0711 01:57:00.363066 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337341\n",
            "I0711 01:57:00.364900 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39746\n",
            "I0711 01:57:03.336394 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336315\n",
            "I0711 01:57:03.337975 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38104\n",
            "I0711 01:57:06.320196 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335147\n",
            "I0711 01:57:06.321597 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36234\n",
            "I0711 01:57:09.252947 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340971\n",
            "I0711 01:57:09.254708 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45553\n",
            "I0711 01:57:12.175099 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34222\n",
            "I0711 01:57:12.176848 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47551\n",
            "I0711 01:57:15.134461 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33791\n",
            "I0711 01:57:15.135749 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40656\n",
            "I0711 01:57:18.158869 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330643\n",
            "I0711 01:57:18.160294 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29029\n",
            "I0711 01:57:21.161276 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333069\n",
            "I0711 01:57:21.162857 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3291\n",
            "I0711 01:57:24.138998 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.6612072 (29.680 sec)\n",
            "I0711 01:57:24.144455 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335212\n",
            "I0711 01:57:24.146579 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36339\n",
            "I0711 01:57:27.071595 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341628\n",
            "I0711 01:57:27.073051 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46605\n",
            "I0711 01:57:30.059035 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334733\n",
            "I0711 01:57:30.061159 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35573\n",
            "I0711 01:57:33.040301 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335429\n",
            "I0711 01:57:33.041624 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36687\n",
            "I0711 01:57:35.977435 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340471\n",
            "I0711 01:57:35.978975 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44754\n",
            "I0711 01:57:39.003849 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330423\n",
            "I0711 01:57:39.005422 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28676\n",
            "I0711 01:57:41.988001 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335103\n",
            "I0711 01:57:41.989682 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36165\n",
            "I0711 01:57:44.991431 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332948\n",
            "I0711 01:57:44.993035 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32716\n",
            "I0711 01:57:47.917266 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341792\n",
            "I0711 01:57:47.918699 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46868\n",
            "I0711 01:57:50.869580 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338714\n",
            "I0711 01:57:50.871090 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41942\n",
            "I0711 01:57:53.851520 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.6545749 (29.712 sec)\n",
            "I0711 01:57:53.858303 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334651\n",
            "I0711 01:57:53.860680 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35441\n",
            "I0711 01:57:56.817987 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337809\n",
            "I0711 01:57:56.819483 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40495\n",
            "I0711 01:57:59.762314 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33964\n",
            "I0711 01:57:59.764164 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43424\n",
            "I0711 01:58:02.673273 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34353\n",
            "I0711 01:58:02.674676 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49649\n",
            "I0711 01:58:05.639956 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337071\n",
            "I0711 01:58:05.641632 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39313\n",
            "I0711 01:58:08.606551 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337092\n",
            "I0711 01:58:08.608165 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39347\n",
            "I0711 01:58:11.542193 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340643\n",
            "I0711 01:58:11.543745 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45029\n",
            "I0711 01:58:14.459247 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342806\n",
            "I0711 01:58:14.460905 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4849\n",
            "I0711 01:58:17.495759 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329363\n",
            "I0711 01:58:17.498591 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2698\n",
            "I0711 01:58:20.455116 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337888\n",
            "I0711 01:58:20.456670 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40621\n",
            "I0711 01:58:23.396472 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.89812565 (29.545 sec)\n",
            "I0711 01:58:23.401338 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339399\n",
            "I0711 01:58:23.403819 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43038\n",
            "I0711 01:58:26.309637 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343848\n",
            "I0711 01:58:26.311297 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50156\n",
            "I0711 01:58:29.234323 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341919\n",
            "I0711 01:58:29.243170 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47071\n",
            "I0711 01:58:32.243599 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332309\n",
            "I0711 01:58:32.245098 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31694\n",
            "I0711 01:58:35.259767 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331562\n",
            "I0711 01:58:35.261216 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30499\n",
            "I0711 01:58:38.249182 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334499\n",
            "I0711 01:58:38.250805 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35199\n",
            "I0711 01:58:41.213798 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337306\n",
            "I0711 01:58:41.215090 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3969\n",
            "I0711 01:58:44.221929 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332427\n",
            "I0711 01:58:44.223436 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31884\n",
            "I0711 01:58:47.186944 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337275\n",
            "I0711 01:58:47.188614 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39639\n",
            "I0711 01:58:50.204123 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331435\n",
            "I0711 01:58:50.205951 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30296\n",
            "I0711 01:58:53.206224 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.04044 (29.810 sec)\n",
            "I0711 01:58:53.209437 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332737\n",
            "I0711 01:58:53.215917 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3238\n",
            "I0711 01:58:56.196257 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334807\n",
            "I0711 01:58:56.197601 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35692\n",
            "I0711 01:58:59.173035 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335938\n",
            "I0711 01:58:59.174481 140338823006080 tpu_estimator.py:2160] examples/sec: 5.375\n",
            "I0711 01:59:02.148815 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336047\n",
            "I0711 01:59:02.150438 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37675\n",
            "I0711 01:59:05.131141 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335305\n",
            "I0711 01:59:05.132523 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36488\n",
            "I0711 01:59:08.124610 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334065\n",
            "I0711 01:59:08.126135 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34503\n",
            "I0711 01:59:11.080848 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338264\n",
            "I0711 01:59:11.082423 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41223\n",
            "I0711 01:59:14.046407 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3372\n",
            "I0711 01:59:14.048098 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3952\n",
            "I0711 01:59:17.018169 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336517\n",
            "I0711 01:59:17.019860 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38427\n",
            "I0711 01:59:20.007527 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334505\n",
            "I0711 01:59:20.008973 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35208\n",
            "I0711 01:59:23.011658 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.626686 (29.805 sec)\n",
            "I0711 01:59:23.018040 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332171\n",
            "I0711 01:59:23.019959 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31473\n",
            "I0711 01:59:26.033934 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331583\n",
            "I0711 01:59:26.035632 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30532\n",
            "I0711 01:59:29.085831 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.32766\n",
            "I0711 01:59:29.087380 140338823006080 tpu_estimator.py:2160] examples/sec: 5.24256\n",
            "I0711 01:59:32.090050 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332864\n",
            "I0711 01:59:32.091921 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32582\n",
            "I0711 01:59:35.068025 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335815\n",
            "I0711 01:59:35.069627 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37304\n",
            "I0711 01:59:38.045587 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33583\n",
            "I0711 01:59:38.047439 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37329\n",
            "I0711 01:59:41.047184 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333156\n",
            "I0711 01:59:41.048490 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33049\n",
            "I0711 01:59:44.020929 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336274\n",
            "I0711 01:59:44.022289 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38039\n",
            "I0711 01:59:47.017103 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333762\n",
            "I0711 01:59:47.018384 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34019\n",
            "I0711 01:59:49.990878 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336269\n",
            "I0711 01:59:49.992645 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38031\n",
            "I0711 01:59:52.994372 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.8934824 (29.983 sec)\n",
            "I0711 01:59:53.001616 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332155\n",
            "I0711 01:59:53.003407 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31448\n",
            "I0711 01:59:55.971484 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336708\n",
            "I0711 01:59:55.972762 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38733\n",
            "I0711 01:59:58.911383 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340146\n",
            "I0711 01:59:58.912988 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44234\n",
            "I0711 02:00:01.877484 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337147\n",
            "I0711 02:00:01.878862 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39435\n",
            "I0711 02:00:04.896043 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331281\n",
            "I0711 02:00:04.897752 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3005\n",
            "I0711 02:00:07.882569 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334834\n",
            "I0711 02:00:07.883986 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35735\n",
            "I0711 02:00:10.858062 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336078\n",
            "I0711 02:00:10.859593 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37724\n",
            "I0711 02:00:13.858463 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333291\n",
            "I0711 02:00:13.860125 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33266\n",
            "I0711 02:00:16.862875 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332843\n",
            "I0711 02:00:16.864363 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32549\n",
            "I0711 02:00:19.883545 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331054\n",
            "I0711 02:00:19.885241 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29686\n",
            "I0711 02:00:22.892257 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0629411 (29.898 sec)\n",
            "I0711 02:00:22.899651 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331555\n",
            "I0711 02:00:22.901962 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30489\n",
            "I0711 02:00:25.881587 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335351\n",
            "I0711 02:00:25.883049 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36562\n",
            "I0711 02:00:28.885747 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332871\n",
            "I0711 02:00:28.887319 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32594\n",
            "I0711 02:00:31.876023 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334416\n",
            "I0711 02:00:31.877390 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35065\n",
            "I0711 02:00:34.837701 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337647\n",
            "I0711 02:00:34.840228 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40236\n",
            "I0711 02:00:37.828604 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334358\n",
            "I0711 02:00:37.829977 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34973\n",
            "I0711 02:00:40.802758 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336218\n",
            "I0711 02:00:40.804109 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37949\n",
            "I0711 02:00:43.809369 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332602\n",
            "I0711 02:00:43.811023 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32164\n",
            "I0711 02:00:46.779886 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336639\n",
            "I0711 02:00:46.781399 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38622\n",
            "I0711 02:00:49.763448 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335177\n",
            "I0711 02:00:49.764878 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36283\n",
            "I0711 02:00:52.723344 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2656655 (29.831 sec)\n",
            "I0711 02:00:52.725758 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337566\n",
            "I0711 02:00:52.727252 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40106\n",
            "I0711 02:00:55.713374 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334713\n",
            "I0711 02:00:55.715310 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35541\n",
            "I0711 02:00:58.696046 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335272\n",
            "I0711 02:00:58.697667 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36435\n",
            "I0711 02:01:01.637431 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339976\n",
            "I0711 02:01:01.638939 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43961\n",
            "I0711 02:01:04.628716 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334306\n",
            "I0711 02:01:04.630345 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34889\n",
            "I0711 02:01:07.623624 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333902\n",
            "I0711 02:01:07.625000 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34243\n",
            "I0711 02:01:10.569469 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339461\n",
            "I0711 02:01:10.571489 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43137\n",
            "I0711 02:01:13.513124 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339716\n",
            "I0711 02:01:13.514734 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43545\n",
            "I0711 02:01:16.436065 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342122\n",
            "I0711 02:01:16.437784 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47396\n",
            "I0711 02:01:19.396230 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337823\n",
            "I0711 02:01:19.397932 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40516\n",
            "I0711 02:01:22.341051 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.75371265 (29.618 sec)\n",
            "I0711 02:01:22.343584 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339282\n",
            "I0711 02:01:22.345313 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42851\n",
            "I0711 02:01:25.273107 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341373\n",
            "I0711 02:01:25.274919 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46197\n",
            "I0711 02:01:28.241276 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336891\n",
            "I0711 02:01:28.242968 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39026\n",
            "I0711 02:01:31.200459 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337935\n",
            "I0711 02:01:31.201871 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40697\n",
            "I0711 02:01:34.152201 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338776\n",
            "I0711 02:01:34.154019 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42042\n",
            "I0711 02:01:37.134368 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335326\n",
            "I0711 02:01:37.136139 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36521\n",
            "I0711 02:01:40.108692 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336208\n",
            "I0711 02:01:40.109958 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37932\n",
            "I0711 02:01:43.081991 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336332\n",
            "I0711 02:01:43.083391 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38132\n",
            "I0711 02:01:45.991284 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343734\n",
            "I0711 02:01:45.992955 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49975\n",
            "I0711 02:01:48.897491 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344079\n",
            "I0711 02:01:48.899044 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50527\n",
            "I0711 02:01:51.837040 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.24831176 (29.496 sec)\n",
            "I0711 02:01:51.839428 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339909\n",
            "I0711 02:01:51.845619 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43855\n",
            "I0711 02:01:54.799440 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337841\n",
            "I0711 02:01:54.800987 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40546\n",
            "I0711 02:01:57.730329 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341198\n",
            "I0711 02:01:57.731853 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45916\n",
            "I0711 02:02:00.640475 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343625\n",
            "I0711 02:02:00.642361 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49801\n",
            "I0711 02:02:03.546155 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344148\n",
            "I0711 02:02:03.547602 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50637\n",
            "I0711 02:02:06.468324 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342225\n",
            "I0711 02:02:06.469581 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4756\n",
            "I0711 02:02:09.463318 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333883\n",
            "I0711 02:02:09.464848 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34213\n",
            "I0711 02:02:12.411834 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339151\n",
            "I0711 02:02:12.413319 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42642\n",
            "I0711 02:02:15.320443 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343809\n",
            "I0711 02:02:15.321797 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50095\n",
            "I0711 02:02:18.272027 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338796\n",
            "I0711 02:02:18.273614 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42074\n",
            "I0711 02:02:21.267535 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.77185214 (29.431 sec)\n",
            "I0711 02:02:21.273436 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333181\n",
            "I0711 02:02:21.277014 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33089\n",
            "I0711 02:02:24.192713 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342546\n",
            "I0711 02:02:24.194356 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48073\n",
            "I0711 02:02:27.165322 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33641\n",
            "I0711 02:02:27.166959 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38257\n",
            "I0711 02:02:30.145194 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335594\n",
            "I0711 02:02:30.149570 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3695\n",
            "I0711 02:02:33.078155 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340943\n",
            "I0711 02:02:33.082915 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45509\n",
            "I0711 02:02:36.010500 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341016\n",
            "I0711 02:02:36.012064 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45625\n",
            "I0711 02:02:39.028890 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33132\n",
            "I0711 02:02:39.034851 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30112\n",
            "I0711 02:02:42.022787 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333999\n",
            "I0711 02:02:42.024249 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34399\n",
            "I0711 02:02:44.965796 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339787\n",
            "I0711 02:02:44.967264 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43659\n",
            "I0711 02:02:47.935844 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336698\n",
            "I0711 02:02:47.940293 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38716\n",
            "I0711 02:02:50.852769 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.3141117 (29.585 sec)\n",
            "I0711 02:02:50.862677 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341659\n",
            "I0711 02:02:50.865480 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46654\n",
            "I0711 02:02:53.775492 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343314\n",
            "I0711 02:02:53.777039 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49302\n",
            "I0711 02:02:56.734349 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337964\n",
            "I0711 02:02:56.736186 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40742\n",
            "I0711 02:02:59.640843 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344063\n",
            "I0711 02:02:59.642260 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50501\n",
            "I0711 02:03:02.553916 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343285\n",
            "I0711 02:03:02.555981 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49255\n",
            "I0711 02:03:05.468602 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343086\n",
            "I0711 02:03:05.473011 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48938\n",
            "I0711 02:03:08.387406 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342606\n",
            "I0711 02:03:08.388799 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4817\n",
            "I0711 02:03:11.280151 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345696\n",
            "I0711 02:03:11.281748 140338823006080 tpu_estimator.py:2160] examples/sec: 5.53114\n",
            "I0711 02:03:14.171797 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345829\n",
            "I0711 02:03:14.173522 140338823006080 tpu_estimator.py:2160] examples/sec: 5.53327\n",
            "I0711 02:03:17.114996 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339756\n",
            "I0711 02:03:17.116442 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43609\n",
            "I0711 02:03:20.065013 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.72142506 (29.212 sec)\n",
            "I0711 02:03:20.068434 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33859\n",
            "I0711 02:03:20.070484 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41744\n",
            "I0711 02:03:23.002620 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340807\n",
            "I0711 02:03:23.004169 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45292\n",
            "I0711 02:03:25.967915 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337239\n",
            "I0711 02:03:25.969802 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39582\n",
            "I0711 02:03:28.913489 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339491\n",
            "I0711 02:03:28.914913 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43185\n",
            "I0711 02:03:31.863332 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339004\n",
            "I0711 02:03:31.864810 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42406\n",
            "I0711 02:03:34.808688 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339514\n",
            "I0711 02:03:34.810000 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43223\n",
            "I0711 02:03:37.770857 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337596\n",
            "I0711 02:03:37.772652 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40154\n",
            "I0711 02:03:40.745044 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336224\n",
            "I0711 02:03:40.746863 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37958\n",
            "I0711 02:03:43.794751 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.327899\n",
            "I0711 02:03:43.796234 140338823006080 tpu_estimator.py:2160] examples/sec: 5.24638\n",
            "I0711 02:03:46.751937 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338156\n",
            "I0711 02:03:46.753312 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41049\n",
            "I0711 02:03:49.707944 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.37439203 (29.643 sec)\n",
            "I0711 02:03:49.719098 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337305\n",
            "I0711 02:03:49.720630 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39688\n",
            "I0711 02:03:52.667090 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338932\n",
            "I0711 02:03:52.668694 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42292\n",
            "I0711 02:03:55.671471 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332843\n",
            "I0711 02:03:55.673022 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32548\n",
            "I0711 02:03:58.650375 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335698\n",
            "I0711 02:03:58.651744 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37117\n",
            "I0711 02:04:01.623278 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336376\n",
            "I0711 02:04:01.631144 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38202\n",
            "I0711 02:04:04.627962 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332814\n",
            "I0711 02:04:04.631176 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32502\n",
            "I0711 02:04:07.606596 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335719\n",
            "I0711 02:04:07.608382 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3715\n",
            "I0711 02:04:10.597374 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334365\n",
            "I0711 02:04:10.598773 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34983\n",
            "I0711 02:04:13.572636 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336103\n",
            "I0711 02:04:13.574267 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37764\n",
            "I0711 02:04:16.568755 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333769\n",
            "I0711 02:04:16.570263 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34031\n",
            "I0711 02:04:19.546705 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.403664 (29.839 sec)\n",
            "I0711 02:04:19.555646 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334934\n",
            "I0711 02:04:19.557331 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35894\n",
            "I0711 02:04:22.530397 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336019\n",
            "I0711 02:04:22.531965 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3763\n",
            "I0711 02:04:25.538709 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332412\n",
            "I0711 02:04:25.540270 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31858\n",
            "I0711 02:04:28.514007 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336103\n",
            "I0711 02:04:28.515805 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37764\n",
            "I0711 02:04:31.474655 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337764\n",
            "I0711 02:04:31.476009 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40422\n",
            "I0711 02:04:34.465673 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334334\n",
            "I0711 02:04:34.467323 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34935\n",
            "I0711 02:04:37.396133 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341257\n",
            "I0711 02:04:37.397748 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46012\n",
            "I0711 02:04:40.353815 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338094\n",
            "I0711 02:04:40.355498 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40951\n",
            "I0711 02:04:43.312000 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338043\n",
            "I0711 02:04:43.313720 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40869\n",
            "I0711 02:04:46.277041 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337259\n",
            "I0711 02:04:46.278415 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39614\n",
            "I0711 02:04:49.248926 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.61057997 (29.702 sec)\n",
            "I0711 02:04:49.258188 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335485\n",
            "I0711 02:04:49.259868 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36776\n",
            "I0711 02:04:52.184446 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341694\n",
            "I0711 02:04:52.185766 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46711\n",
            "I0711 02:04:55.141209 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338206\n",
            "I0711 02:04:55.143011 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4113\n",
            "I0711 02:04:58.080408 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340222\n",
            "I0711 02:04:58.082233 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44355\n",
            "I0711 02:05:01.031795 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33883\n",
            "I0711 02:05:01.033264 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42129\n",
            "I0711 02:05:03.962016 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341264\n",
            "I0711 02:05:03.963493 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46022\n",
            "I0711 02:05:06.873151 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343526\n",
            "I0711 02:05:06.874816 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49641\n",
            "I0711 02:05:09.788009 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343053\n",
            "I0711 02:05:09.789603 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48885\n",
            "I0711 02:05:12.721244 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340929\n",
            "I0711 02:05:12.722698 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45486\n",
            "I0711 02:05:15.633570 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343377\n",
            "I0711 02:05:15.634999 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49402\n",
            "I0711 02:05:18.560002 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.782533 (29.311 sec)\n",
            "I0711 02:05:18.567812 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340797\n",
            "I0711 02:05:18.570770 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45276\n",
            "I0711 02:05:21.480957 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343261\n",
            "I0711 02:05:21.482372 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49218\n",
            "I0711 02:05:24.459263 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335773\n",
            "I0711 02:05:24.460774 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37237\n",
            "I0711 02:05:27.402813 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339717\n",
            "I0711 02:05:27.404192 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43546\n",
            "I0711 02:05:30.322666 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342482\n",
            "I0711 02:05:30.324245 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47971\n",
            "I0711 02:05:33.308572 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334911\n",
            "I0711 02:05:33.309889 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35858\n",
            "I0711 02:05:36.241546 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34095\n",
            "I0711 02:05:36.243111 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45521\n",
            "I0711 02:05:39.210363 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336832\n",
            "I0711 02:05:39.211779 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38931\n",
            "I0711 02:05:42.179420 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336809\n",
            "I0711 02:05:42.181135 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38894\n",
            "I0711 02:05:45.141384 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33761\n",
            "I0711 02:05:45.143588 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40177\n",
            "I0711 02:05:48.084453 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.7173867 (29.524 sec)\n",
            "I0711 02:05:48.089751 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339172\n",
            "I0711 02:05:48.094835 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42675\n",
            "I0711 02:05:51.012341 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342171\n",
            "I0711 02:05:51.013849 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47474\n",
            "I0711 02:05:53.955337 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339805\n",
            "I0711 02:05:53.956704 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43687\n",
            "I0711 02:05:56.909617 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338468\n",
            "I0711 02:05:56.911139 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41548\n",
            "I0711 02:05:59.855625 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339443\n",
            "I0711 02:05:59.856971 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43109\n",
            "I0711 02:06:02.781661 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341758\n",
            "I0711 02:06:02.783199 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46812\n",
            "I0711 02:06:05.746968 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337239\n",
            "I0711 02:06:05.748547 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39582\n",
            "I0711 02:06:08.744880 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333565\n",
            "I0711 02:06:08.746460 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33704\n",
            "I0711 02:06:11.677385 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341006\n",
            "I0711 02:06:11.678747 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4561\n",
            "I0711 02:06:14.619483 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339888\n",
            "I0711 02:06:14.620782 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43821\n",
            "I0711 02:06:17.547364 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.29989576 (29.463 sec)\n",
            "I0711 02:06:17.555853 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340619\n",
            "I0711 02:06:17.560056 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4499\n",
            "I0711 02:06:20.497633 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339869\n",
            "I0711 02:06:20.499161 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4379\n",
            "I0711 02:06:23.442578 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339579\n",
            "I0711 02:06:23.444131 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43326\n",
            "I0711 02:06:26.384144 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339944\n",
            "I0711 02:06:26.385761 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4391\n",
            "I0711 02:06:29.307893 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342022\n",
            "I0711 02:06:29.309407 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47236\n",
            "I0711 02:06:32.265738 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338086\n",
            "I0711 02:06:32.267058 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40937\n",
            "I0711 02:06:35.192396 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341685\n",
            "I0711 02:06:35.194107 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46696\n",
            "I0711 02:06:38.124500 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341063\n",
            "I0711 02:06:38.126407 140338823006080 tpu_estimator.py:2160] examples/sec: 5.457\n",
            "I0711 02:06:41.032511 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343868\n",
            "I0711 02:06:41.033713 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50189\n",
            "I0711 02:06:43.972740 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340129\n",
            "I0711 02:06:43.974447 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44206\n",
            "I0711 02:06:46.886940 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.13204575 (29.340 sec)\n",
            "I0711 02:06:46.890546 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342695\n",
            "I0711 02:06:46.893500 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48313\n",
            "I0711 02:06:49.799738 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343745\n",
            "I0711 02:06:49.801363 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49992\n",
            "I0711 02:06:52.740787 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340016\n",
            "I0711 02:06:52.742348 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44026\n",
            "I0711 02:06:55.655026 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343147\n",
            "I0711 02:06:55.656519 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49035\n",
            "I0711 02:06:58.604706 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339012\n",
            "I0711 02:06:58.606697 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4242\n",
            "I0711 02:07:01.583341 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335733\n",
            "I0711 02:07:01.584821 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37172\n",
            "I0711 02:07:04.525398 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339892\n",
            "I0711 02:07:04.527035 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43827\n",
            "I0711 02:07:07.422826 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345136\n",
            "I0711 02:07:07.424377 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52218\n",
            "I0711 02:07:10.348823 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341783\n",
            "I0711 02:07:10.350859 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46853\n",
            "I0711 02:07:13.258377 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343676\n",
            "I0711 02:07:13.259814 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49882\n",
            "I0711 02:07:16.189422 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.24780583 (29.302 sec)\n",
            "I0711 02:07:16.196983 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340296\n",
            "I0711 02:07:16.199173 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44474\n",
            "I0711 02:07:19.179579 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335287\n",
            "I0711 02:07:19.180902 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3646\n",
            "I0711 02:07:22.161375 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33536\n",
            "I0711 02:07:22.163046 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36576\n",
            "I0711 02:07:25.157109 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333812\n",
            "I0711 02:07:25.159143 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34099\n",
            "I0711 02:07:28.123268 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337134\n",
            "I0711 02:07:28.125057 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39415\n",
            "I0711 02:07:31.097297 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336245\n",
            "I0711 02:07:31.099104 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37993\n",
            "I0711 02:07:34.078214 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335472\n",
            "I0711 02:07:34.079560 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36755\n",
            "I0711 02:07:37.053613 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336089\n",
            "I0711 02:07:37.055536 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37742\n",
            "I0711 02:07:40.010000 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338248\n",
            "I0711 02:07:40.014177 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41197\n",
            "I0711 02:07:42.975223 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337243\n",
            "I0711 02:07:42.976788 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39589\n",
            "I0711 02:07:45.889496 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.77829576 (29.700 sec)\n",
            "I0711 02:07:45.892143 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342834\n",
            "I0711 02:07:45.896441 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48534\n",
            "I0711 02:07:48.810273 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342676\n",
            "I0711 02:07:48.811923 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48282\n",
            "I0711 02:07:51.784191 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336255\n",
            "I0711 02:07:51.785510 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38008\n",
            "I0711 02:07:54.694480 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343613\n",
            "I0711 02:07:54.695975 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49782\n",
            "I0711 02:07:57.611177 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342852\n",
            "I0711 02:07:57.612519 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48563\n",
            "I0711 02:08:00.582608 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336535\n",
            "I0711 02:08:00.584052 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38456\n",
            "I0711 02:08:03.532877 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338949\n",
            "I0711 02:08:03.534287 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42318\n",
            "I0711 02:08:06.461793 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341424\n",
            "I0711 02:08:06.463218 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46279\n",
            "I0711 02:08:09.376095 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343145\n",
            "I0711 02:08:09.377495 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49032\n",
            "I0711 02:08:12.351034 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336131\n",
            "I0711 02:08:12.352596 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3781\n",
            "I0711 02:08:15.277511 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.13700533 (29.388 sec)\n",
            "I0711 02:08:15.285452 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340791\n",
            "I0711 02:08:15.287820 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45265\n",
            "I0711 02:08:18.235255 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339002\n",
            "I0711 02:08:18.236716 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42404\n",
            "I0711 02:08:21.146760 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343468\n",
            "I0711 02:08:21.148405 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49548\n",
            "I0711 02:08:24.072889 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341747\n",
            "I0711 02:08:24.074387 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46795\n",
            "I0711 02:08:27.002346 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341357\n",
            "I0711 02:08:27.003808 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46171\n",
            "I0711 02:08:29.945378 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339788\n",
            "I0711 02:08:29.946998 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43661\n",
            "I0711 02:08:32.888138 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339822\n",
            "I0711 02:08:32.889668 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43716\n",
            "I0711 02:08:35.868786 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335496\n",
            "I0711 02:08:35.870192 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36794\n",
            "I0711 02:08:38.856953 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334647\n",
            "I0711 02:08:38.858608 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35435\n",
            "I0711 02:08:41.779308 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342198\n",
            "I0711 02:08:41.780981 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47516\n",
            "I0711 02:08:44.736245 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.09891105 (29.459 sec)\n",
            "I0711 02:08:44.744673 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337224\n",
            "I0711 02:08:44.745929 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39558\n",
            "I0711 02:08:47.677395 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340975\n",
            "I0711 02:08:47.678915 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4556\n",
            "I0711 02:08:50.659261 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335369\n",
            "I0711 02:08:50.660955 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36591\n",
            "I0711 02:08:53.667866 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332377\n",
            "I0711 02:08:53.669137 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31802\n",
            "I0711 02:08:56.652678 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335024\n",
            "I0711 02:08:56.654509 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36038\n",
            "I0711 02:08:59.662549 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332237\n",
            "I0711 02:08:59.664858 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31579\n",
            "I0711 02:09:02.680140 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331402\n",
            "I0711 02:09:02.681765 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30243\n",
            "I0711 02:09:05.694212 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331777\n",
            "I0711 02:09:05.695824 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30843\n",
            "I0711 02:09:08.724240 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330028\n",
            "I0711 02:09:08.725683 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28045\n",
            "I0711 02:09:11.700353 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336006\n",
            "I0711 02:09:11.702246 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37609\n",
            "I0711 02:09:14.674568 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.3100853 (29.938 sec)\n",
            "I0711 02:09:14.685998 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334937\n",
            "I0711 02:09:14.687770 140338823006080 tpu_estimator.py:2160] examples/sec: 5.359\n",
            "I0711 02:09:17.649482 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337438\n",
            "I0711 02:09:17.651121 140338823006080 tpu_estimator.py:2160] examples/sec: 5.399\n",
            "I0711 02:09:20.675803 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330433\n",
            "I0711 02:09:20.677479 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28692\n",
            "I0711 02:09:23.652811 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335909\n",
            "I0711 02:09:23.654539 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37454\n",
            "I0711 02:09:26.664521 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332041\n",
            "I0711 02:09:26.666209 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31266\n",
            "I0711 02:09:29.664378 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333346\n",
            "I0711 02:09:29.665947 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33353\n",
            "I0711 02:09:32.671890 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332506\n",
            "I0711 02:09:32.673563 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3201\n",
            "I0711 02:09:35.681336 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33228\n",
            "I0711 02:09:35.683341 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31649\n",
            "I0711 02:09:38.670351 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334559\n",
            "I0711 02:09:38.671885 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35295\n",
            "I0711 02:09:41.658105 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334706\n",
            "I0711 02:09:41.659748 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35529\n",
            "I0711 02:09:44.653208 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.0143876 (29.979 sec)\n",
            "I0711 02:09:44.661841 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332944\n",
            "I0711 02:09:44.664416 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3271\n",
            "I0711 02:09:47.641456 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335584\n",
            "I0711 02:09:47.642750 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36934\n",
            "I0711 02:09:50.676987 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329434\n",
            "I0711 02:09:50.678435 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27094\n",
            "I0711 02:09:53.679747 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33303\n",
            "I0711 02:09:53.681571 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32848\n",
            "I0711 02:09:56.644373 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337307\n",
            "I0711 02:09:56.645784 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39691\n",
            "I0711 02:09:59.596969 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338692\n",
            "I0711 02:09:59.598785 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41908\n",
            "I0711 02:10:02.584953 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33467\n",
            "I0711 02:10:02.586515 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35471\n",
            "I0711 02:10:05.553249 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336895\n",
            "I0711 02:10:05.554656 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39032\n",
            "I0711 02:10:08.540659 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334736\n",
            "I0711 02:10:08.542199 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35577\n",
            "I0711 02:10:11.536765 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333768\n",
            "I0711 02:10:11.538275 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34029\n",
            "I0711 02:10:14.547241 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.4567218 (29.894 sec)\n",
            "I0711 02:10:14.554736 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331343\n",
            "I0711 02:10:14.557645 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30149\n",
            "I0711 02:10:17.572766 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331343\n",
            "I0711 02:10:17.574592 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30149\n",
            "I0711 02:10:20.564771 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334231\n",
            "I0711 02:10:20.566450 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34769\n",
            "I0711 02:10:23.553023 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334641\n",
            "I0711 02:10:23.554621 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35426\n",
            "I0711 02:10:26.521384 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33688\n",
            "I0711 02:10:26.523041 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39008\n",
            "I0711 02:10:29.475713 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338485\n",
            "I0711 02:10:29.476998 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41576\n",
            "I0711 02:10:32.438167 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337559\n",
            "I0711 02:10:32.439517 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40095\n",
            "I0711 02:10:35.384113 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339453\n",
            "I0711 02:10:35.385789 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43125\n",
            "I0711 02:10:38.329526 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339512\n",
            "I0711 02:10:38.330812 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43219\n",
            "I0711 02:10:41.227813 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345026\n",
            "I0711 02:10:41.229478 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52041\n",
            "I0711 02:10:44.182183 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.3781016 (29.635 sec)\n",
            "I0711 02:10:44.184750 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338187\n",
            "I0711 02:10:44.188729 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41099\n",
            "I0711 02:10:47.130519 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339476\n",
            "I0711 02:10:47.131916 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43162\n",
            "I0711 02:10:50.067813 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340449\n",
            "I0711 02:10:50.069316 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44719\n",
            "I0711 02:10:53.058970 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334315\n",
            "I0711 02:10:53.060518 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34904\n",
            "I0711 02:10:56.030780 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336496\n",
            "I0711 02:10:56.032593 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38394\n",
            "I0711 02:10:58.946379 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342981\n",
            "I0711 02:10:58.948227 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4877\n",
            "I0711 02:11:01.929328 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335234\n",
            "I0711 02:11:01.931547 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36375\n",
            "I0711 02:11:04.869980 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340071\n",
            "I0711 02:11:04.871252 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44113\n",
            "I0711 02:11:07.842404 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336422\n",
            "I0711 02:11:07.843986 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38275\n",
            "I0711 02:11:10.744510 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344578\n",
            "I0711 02:11:10.745910 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51324\n",
            "I0711 02:11:13.688958 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.5480092 (29.507 sec)\n",
            "I0711 02:11:13.697587 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338648\n",
            "I0711 02:11:13.699183 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41836\n",
            "I0711 02:11:16.622767 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341837\n",
            "I0711 02:11:16.624140 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46939\n",
            "I0711 02:11:19.573653 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338891\n",
            "I0711 02:11:19.575529 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42225\n",
            "I0711 02:11:22.514460 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340038\n",
            "I0711 02:11:22.516102 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4406\n",
            "I0711 02:11:25.456980 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339849\n",
            "I0711 02:11:25.458615 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43758\n",
            "I0711 02:11:28.386954 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341292\n",
            "I0711 02:11:28.388308 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46067\n",
            "I0711 02:11:31.328593 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339947\n",
            "I0711 02:11:31.330167 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43915\n",
            "I0711 02:11:34.300326 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336513\n",
            "I0711 02:11:34.301808 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3842\n",
            "I0711 02:11:37.270151 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336718\n",
            "I0711 02:11:37.271751 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38749\n",
            "I0711 02:11:40.209799 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34017\n",
            "I0711 02:11:40.211421 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44272\n",
            "I0711 02:11:43.145175 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.24644017 (29.456 sec)\n",
            "I0711 02:11:43.151364 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339951\n",
            "I0711 02:11:43.155273 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43921\n",
            "I0711 02:11:46.069237 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342725\n",
            "I0711 02:11:46.070565 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4836\n",
            "I0711 02:11:49.033732 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337325\n",
            "I0711 02:11:49.035030 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39721\n",
            "I0711 02:11:51.961495 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34156\n",
            "I0711 02:11:51.963045 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46496\n",
            "I0711 02:11:54.942847 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335414\n",
            "I0711 02:11:54.944179 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36662\n",
            "I0711 02:11:57.899289 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338246\n",
            "I0711 02:11:57.900696 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41194\n",
            "I0711 02:12:00.862275 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337499\n",
            "I0711 02:12:00.863649 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39998\n",
            "I0711 02:12:03.851981 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33448\n",
            "I0711 02:12:03.853518 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35167\n",
            "I0711 02:12:06.833362 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335413\n",
            "I0711 02:12:06.834811 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3666\n",
            "I0711 02:12:09.797485 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337366\n",
            "I0711 02:12:09.798801 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39786\n",
            "I0711 02:12:12.787644 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.078815 (29.642 sec)\n",
            "I0711 02:12:12.798002 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333299\n",
            "I0711 02:12:12.800331 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33278\n",
            "I0711 02:12:15.747129 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339073\n",
            "I0711 02:12:15.748411 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42518\n",
            "I0711 02:12:18.708716 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337648\n",
            "I0711 02:12:18.710679 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40237\n",
            "I0711 02:12:21.705400 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333697\n",
            "I0711 02:12:21.706674 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33916\n",
            "I0711 02:12:24.647417 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339902\n",
            "I0711 02:12:24.648808 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43843\n",
            "I0711 02:12:27.647426 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333336\n",
            "I0711 02:12:27.648893 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33338\n",
            "I0711 02:12:30.630435 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33523\n",
            "I0711 02:12:30.631886 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36368\n",
            "I0711 02:12:33.577568 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339326\n",
            "I0711 02:12:33.579019 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42921\n",
            "I0711 02:12:36.556972 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335625\n",
            "I0711 02:12:36.558391 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36999\n",
            "I0711 02:12:39.561040 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332889\n",
            "I0711 02:12:39.562379 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32623\n",
            "I0711 02:12:42.507063 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.36695528 (29.719 sec)\n",
            "I0711 02:12:42.513634 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338765\n",
            "I0711 02:12:42.516170 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42023\n",
            "I0711 02:12:45.484860 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336478\n",
            "I0711 02:12:45.489704 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38364\n",
            "I0711 02:12:48.442125 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338151\n",
            "I0711 02:12:48.445469 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41042\n",
            "I0711 02:12:51.427908 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334922\n",
            "I0711 02:12:51.429583 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35875\n",
            "I0711 02:12:54.414224 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33486\n",
            "I0711 02:12:54.415785 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35776\n",
            "I0711 02:12:57.364594 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338937\n",
            "I0711 02:12:57.369210 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42299\n",
            "I0711 02:13:00.368121 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332949\n",
            "I0711 02:13:00.369724 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32718\n",
            "I0711 02:13:03.364942 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333677\n",
            "I0711 02:13:03.367105 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33883\n",
            "I0711 02:13:06.351789 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334803\n",
            "I0711 02:13:06.353279 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35685\n",
            "I0711 02:13:09.395718 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328531\n",
            "I0711 02:13:09.397365 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25649\n",
            "I0711 02:13:12.401775 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.1448789 (29.895 sec)\n",
            "I0711 02:13:12.411460 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331606\n",
            "I0711 02:13:12.413488 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3057\n",
            "I0711 02:13:15.395364 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335122\n",
            "I0711 02:13:15.397087 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36196\n",
            "I0711 02:13:18.372204 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335923\n",
            "I0711 02:13:18.373881 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37477\n",
            "I0711 02:13:21.364916 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334138\n",
            "I0711 02:13:21.366553 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34621\n",
            "I0711 02:13:24.360057 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333874\n",
            "I0711 02:13:24.361753 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34199\n",
            "I0711 02:13:27.371794 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332038\n",
            "I0711 02:13:27.373207 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31261\n",
            "I0711 02:13:30.380853 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332327\n",
            "I0711 02:13:30.382723 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31723\n",
            "I0711 02:13:33.372187 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3343\n",
            "I0711 02:13:33.373490 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3488\n",
            "I0711 02:13:36.356376 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335103\n",
            "I0711 02:13:36.358131 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36165\n",
            "I0711 02:13:39.348533 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334208\n",
            "I0711 02:13:39.350275 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34732\n",
            "I0711 02:13:42.305447 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.94571733 (29.904 sec)\n",
            "I0711 02:13:42.308057 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337888\n",
            "I0711 02:13:42.313216 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40621\n",
            "I0711 02:13:45.324784 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331481\n",
            "I0711 02:13:45.326340 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3037\n",
            "I0711 02:13:48.279098 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338493\n",
            "I0711 02:13:48.280736 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41589\n",
            "I0711 02:13:51.283773 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33281\n",
            "I0711 02:13:51.285335 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32496\n",
            "I0711 02:13:54.289421 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332707\n",
            "I0711 02:13:54.291105 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32331\n",
            "I0711 02:13:57.295517 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332665\n",
            "I0711 02:13:57.296899 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32265\n",
            "I0711 02:14:00.271483 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336019\n",
            "I0711 02:14:00.273187 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3763\n",
            "I0711 02:14:03.280514 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332343\n",
            "I0711 02:14:03.281869 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31749\n",
            "I0711 02:14:06.244533 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337371\n",
            "I0711 02:14:06.245760 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39794\n",
            "I0711 02:14:09.181117 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340533\n",
            "I0711 02:14:09.182927 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44853\n",
            "I0711 02:14:12.126777 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.64989984 (29.821 sec)\n",
            "I0711 02:14:12.134344 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338614\n",
            "I0711 02:14:12.136201 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41782\n",
            "I0711 02:14:15.154896 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331064\n",
            "I0711 02:14:15.156426 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29703\n",
            "I0711 02:14:18.144850 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334459\n",
            "I0711 02:14:18.146517 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35134\n",
            "I0711 02:14:21.133468 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334604\n",
            "I0711 02:14:21.138505 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35367\n",
            "I0711 02:14:24.130066 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333705\n",
            "I0711 02:14:24.134869 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33929\n",
            "I0711 02:14:27.136147 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332657\n",
            "I0711 02:14:27.137744 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32252\n",
            "I0711 02:14:30.123311 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334765\n",
            "I0711 02:14:30.124799 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35624\n",
            "I0711 02:14:33.088261 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337276\n",
            "I0711 02:14:33.089966 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39641\n",
            "I0711 02:14:36.099865 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332046\n",
            "I0711 02:14:36.101343 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31273\n",
            "I0711 02:14:39.087312 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334737\n",
            "I0711 02:14:39.088744 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3558\n",
            "I0711 02:14:42.042147 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.807512 (29.915 sec)\n",
            "I0711 02:14:42.050965 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337419\n",
            "I0711 02:14:42.054385 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3987\n",
            "I0711 02:14:45.048215 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333644\n",
            "I0711 02:14:45.049873 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3383\n",
            "I0711 02:14:47.976928 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341445\n",
            "I0711 02:14:47.978527 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46312\n",
            "I0711 02:14:50.947523 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336637\n",
            "I0711 02:14:50.949121 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38619\n",
            "I0711 02:14:53.989650 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328712\n",
            "I0711 02:14:53.990959 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25939\n",
            "I0711 02:14:56.969123 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335632\n",
            "I0711 02:14:56.970720 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37011\n",
            "I0711 02:14:59.978826 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332255\n",
            "I0711 02:14:59.980509 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31608\n",
            "I0711 02:15:02.924648 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339463\n",
            "I0711 02:15:02.926120 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43141\n",
            "I0711 02:15:05.920894 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33375\n",
            "I0711 02:15:05.922871 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34\n",
            "I0711 02:15:08.906129 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334986\n",
            "I0711 02:15:08.907438 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35977\n",
            "I0711 02:15:11.912450 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.59831285 (29.870 sec)\n",
            "I0711 02:15:11.920847 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331823\n",
            "I0711 02:15:11.925058 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30917\n",
            "I0711 02:15:14.901968 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335329\n",
            "I0711 02:15:14.903378 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36526\n",
            "I0711 02:15:17.882256 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335544\n",
            "I0711 02:15:17.883905 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36871\n",
            "I0711 02:15:20.858473 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335989\n",
            "I0711 02:15:20.860027 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37583\n",
            "I0711 02:15:23.808830 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338935\n",
            "I0711 02:15:23.810247 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42296\n",
            "I0711 02:15:26.775641 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337075\n",
            "I0711 02:15:26.777326 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3932\n",
            "I0711 02:15:29.751539 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336022\n",
            "I0711 02:15:29.753137 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37636\n",
            "I0711 02:15:32.760981 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332297\n",
            "I0711 02:15:32.762377 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31674\n",
            "I0711 02:15:35.747653 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334817\n",
            "I0711 02:15:35.749263 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35708\n",
            "I0711 02:15:38.746977 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333402\n",
            "I0711 02:15:38.748404 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33444\n",
            "I0711 02:15:41.739524 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.9113767 (29.827 sec)\n",
            "I0711 02:15:41.745433 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333506\n",
            "I0711 02:15:41.749582 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33609\n",
            "I0711 02:15:44.728729 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335197\n",
            "I0711 02:15:44.730493 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36315\n",
            "I0711 02:15:47.722006 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334088\n",
            "I0711 02:15:47.723431 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3454\n",
            "I0711 02:15:50.697309 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336099\n",
            "I0711 02:15:50.698881 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37759\n",
            "I0711 02:15:53.630670 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340906\n",
            "I0711 02:15:53.632394 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45449\n",
            "I0711 02:15:56.580579 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338996\n",
            "I0711 02:15:56.582189 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42394\n",
            "I0711 02:15:59.549247 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336855\n",
            "I0711 02:15:59.550863 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38967\n",
            "I0711 02:16:02.502257 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338632\n",
            "I0711 02:16:02.503819 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41811\n",
            "I0711 02:16:05.423392 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342328\n",
            "I0711 02:16:05.424867 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47725\n",
            "I0711 02:16:08.328647 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344205\n",
            "I0711 02:16:08.330044 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50727\n",
            "I0711 02:16:11.263853 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.7414733 (29.524 sec)\n",
            "I0711 02:16:11.271854 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339781\n",
            "I0711 02:16:11.274209 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43649\n",
            "I0711 02:16:14.244844 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336346\n",
            "I0711 02:16:14.246195 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38154\n",
            "I0711 02:16:17.187627 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339813\n",
            "I0711 02:16:17.188950 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43701\n",
            "I0711 02:16:20.161372 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336281\n",
            "I0711 02:16:20.162741 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3805\n",
            "I0711 02:16:23.159439 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333547\n",
            "I0711 02:16:23.160946 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33675\n",
            "I0711 02:16:26.086637 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341632\n",
            "I0711 02:16:26.088285 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46612\n",
            "I0711 02:16:29.036499 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338994\n",
            "I0711 02:16:29.038187 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4239\n",
            "I0711 02:16:32.033619 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333652\n",
            "I0711 02:16:32.035404 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33844\n",
            "I0711 02:16:35.055531 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330919\n",
            "I0711 02:16:35.056970 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29471\n",
            "I0711 02:16:38.023239 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336959\n",
            "I0711 02:16:38.024666 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39135\n",
            "I0711 02:16:40.986488 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.7199047 (29.723 sec)\n",
            "I0711 02:16:40.993289 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336683\n",
            "I0711 02:16:40.996257 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38693\n",
            "I0711 02:16:43.936866 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339728\n",
            "I0711 02:16:43.938403 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43565\n",
            "I0711 02:16:46.870457 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340884\n",
            "I0711 02:16:46.872014 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45415\n",
            "I0711 02:16:49.814409 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339677\n",
            "I0711 02:16:49.816100 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43483\n",
            "I0711 02:16:52.793988 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335616\n",
            "I0711 02:16:52.795583 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36986\n",
            "I0711 02:16:55.754139 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337831\n",
            "I0711 02:16:55.755714 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4053\n",
            "I0711 02:16:58.696445 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339859\n",
            "I0711 02:16:58.698179 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43775\n",
            "I0711 02:17:01.679797 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335204\n",
            "I0711 02:17:01.681190 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36326\n",
            "I0711 02:17:04.669094 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334527\n",
            "I0711 02:17:04.670557 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35243\n",
            "I0711 02:17:07.677878 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332349\n",
            "I0711 02:17:07.679835 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31759\n",
            "I0711 02:17:10.682059 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0999974 (29.696 sec)\n",
            "I0711 02:17:10.686990 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332331\n",
            "I0711 02:17:10.690282 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3173\n",
            "I0711 02:17:13.687048 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333322\n",
            "I0711 02:17:13.688508 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33315\n",
            "I0711 02:17:16.682789 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333814\n",
            "I0711 02:17:16.684145 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34103\n",
            "I0711 02:17:19.628788 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339434\n",
            "I0711 02:17:19.630160 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43095\n",
            "I0711 02:17:22.651443 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330841\n",
            "I0711 02:17:22.653026 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29346\n",
            "I0711 02:17:25.589343 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340371\n",
            "I0711 02:17:25.590855 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44594\n",
            "I0711 02:17:28.547795 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338014\n",
            "I0711 02:17:28.549397 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40823\n",
            "I0711 02:17:31.518162 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33666\n",
            "I0711 02:17:31.519590 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38656\n",
            "I0711 02:17:34.510990 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334133\n",
            "I0711 02:17:34.512669 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34613\n",
            "I0711 02:17:37.458757 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339245\n",
            "I0711 02:17:37.460108 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42791\n",
            "I0711 02:17:40.450781 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.7909148 (29.769 sec)\n",
            "I0711 02:17:40.459289 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333278\n",
            "I0711 02:17:40.461551 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33244\n",
            "I0711 02:17:43.444830 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334945\n",
            "I0711 02:17:43.446396 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35911\n",
            "I0711 02:17:46.433207 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334627\n",
            "I0711 02:17:46.434937 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35403\n",
            "I0711 02:17:49.383296 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338974\n",
            "I0711 02:17:49.388683 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42359\n",
            "I0711 02:17:52.355339 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336482\n",
            "I0711 02:17:52.357284 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38371\n",
            "I0711 02:17:55.297863 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339828\n",
            "I0711 02:17:55.299635 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43725\n",
            "I0711 02:17:58.249397 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338809\n",
            "I0711 02:17:58.251047 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42095\n",
            "I0711 02:18:01.231626 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335315\n",
            "I0711 02:18:01.233094 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36503\n",
            "I0711 02:18:04.282596 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.327766\n",
            "I0711 02:18:04.284263 140338823006080 tpu_estimator.py:2160] examples/sec: 5.24425\n",
            "I0711 02:18:07.214973 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341023\n",
            "I0711 02:18:07.216583 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45636\n",
            "I0711 02:18:10.183462 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.1882038 (29.733 sec)\n",
            "I0711 02:18:10.188889 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33628\n",
            "I0711 02:18:10.192738 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38049\n",
            "I0711 02:18:13.167663 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335683\n",
            "I0711 02:18:13.169305 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37094\n",
            "I0711 02:18:16.159305 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33427\n",
            "I0711 02:18:16.160817 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34831\n",
            "I0711 02:18:19.153868 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33394\n",
            "I0711 02:18:19.155380 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34305\n",
            "I0711 02:18:22.111580 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338105\n",
            "I0711 02:18:22.113185 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40968\n",
            "I0711 02:18:25.114304 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333021\n",
            "I0711 02:18:25.115922 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32834\n",
            "I0711 02:18:28.095571 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335433\n",
            "I0711 02:18:28.097030 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36693\n",
            "I0711 02:18:31.060364 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337285\n",
            "I0711 02:18:31.061995 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39655\n",
            "I0711 02:18:34.029647 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336781\n",
            "I0711 02:18:34.031032 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3885\n",
            "I0711 02:18:37.026633 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333669\n",
            "I0711 02:18:37.028411 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3387\n",
            "I0711 02:18:40.002787 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.65572834 (29.819 sec)\n",
            "I0711 02:18:40.011338 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33504\n",
            "I0711 02:18:40.012851 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36065\n",
            "I0711 02:18:43.007121 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333804\n",
            "I0711 02:18:43.008725 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34087\n",
            "I0711 02:18:45.962808 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33833\n",
            "I0711 02:18:45.964379 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41327\n",
            "I0711 02:18:48.956271 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334064\n",
            "I0711 02:18:48.957911 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34503\n",
            "I0711 02:18:51.949846 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334046\n",
            "I0711 02:18:51.951724 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34473\n",
            "I0711 02:18:54.940276 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334407\n",
            "I0711 02:18:54.942131 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35052\n",
            "I0711 02:18:57.903551 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337461\n",
            "I0711 02:18:57.905221 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39937\n",
            "I0711 02:19:00.893208 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334498\n",
            "I0711 02:19:00.894987 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35197\n",
            "I0711 02:19:03.845324 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33874\n",
            "I0711 02:19:03.847361 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41984\n",
            "I0711 02:19:06.802579 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338137\n",
            "I0711 02:19:06.804343 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41019\n",
            "I0711 02:19:09.812776 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.67524767 (29.810 sec)\n",
            "I0711 02:19:09.819353 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331477\n",
            "I0711 02:19:09.822838 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30363\n",
            "I0711 02:19:12.795315 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336032\n",
            "I0711 02:19:12.796630 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37651\n",
            "I0711 02:19:15.823549 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330225\n",
            "I0711 02:19:15.825185 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28361\n",
            "I0711 02:19:18.797701 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336229\n",
            "I0711 02:19:18.799134 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37967\n",
            "I0711 02:19:21.779722 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335341\n",
            "I0711 02:19:21.781299 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36546\n",
            "I0711 02:19:24.727966 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339187\n",
            "I0711 02:19:24.729295 140338823006080 tpu_estimator.py:2160] examples/sec: 5.427\n",
            "I0711 02:19:27.699146 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336564\n",
            "I0711 02:19:27.700839 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38503\n",
            "I0711 02:19:30.667866 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336849\n",
            "I0711 02:19:30.669675 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38958\n",
            "I0711 02:19:33.650222 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335308\n",
            "I0711 02:19:33.651930 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36493\n",
            "I0711 02:19:36.628204 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335801\n",
            "I0711 02:19:36.629637 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37282\n",
            "I0711 02:19:39.574645 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.801574 (29.762 sec)\n",
            "I0711 02:19:39.581037 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338649\n",
            "I0711 02:19:39.584169 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41839\n",
            "I0711 02:19:42.545100 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337382\n",
            "I0711 02:19:42.546399 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39811\n",
            "I0711 02:19:45.531694 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334825\n",
            "I0711 02:19:45.533338 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3572\n",
            "I0711 02:19:48.499586 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336935\n",
            "I0711 02:19:48.501096 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39096\n",
            "I0711 02:19:51.502489 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333016\n",
            "I0711 02:19:51.503813 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32825\n",
            "I0711 02:19:54.521312 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33125\n",
            "I0711 02:19:54.523286 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29999\n",
            "I0711 02:19:57.524852 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332942\n",
            "I0711 02:19:57.526140 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32707\n",
            "I0711 02:20:00.560311 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329445\n",
            "I0711 02:20:00.561867 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27111\n",
            "I0711 02:20:03.528370 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336914\n",
            "I0711 02:20:03.529826 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39063\n",
            "I0711 02:20:06.513942 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33495\n",
            "I0711 02:20:06.515749 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35919\n",
            "I0711 02:20:09.443389 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.0906897 (29.869 sec)\n",
            "I0711 02:20:09.450172 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340634\n",
            "I0711 02:20:09.453130 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45015\n",
            "I0711 02:20:12.392809 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339769\n",
            "I0711 02:20:12.394332 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43631\n",
            "I0711 02:20:15.387212 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333957\n",
            "I0711 02:20:15.388580 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34331\n",
            "I0711 02:20:18.364152 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33593\n",
            "I0711 02:20:18.365696 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37488\n",
            "I0711 02:20:21.333436 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336769\n",
            "I0711 02:20:21.335118 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3883\n",
            "I0711 02:20:24.333289 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33335\n",
            "I0711 02:20:24.334729 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3336\n",
            "I0711 02:20:27.308371 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336117\n",
            "I0711 02:20:27.309719 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37788\n",
            "I0711 02:20:30.299181 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334365\n",
            "I0711 02:20:30.300537 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34984\n",
            "I0711 02:20:33.280289 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335447\n",
            "I0711 02:20:33.281697 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36715\n",
            "I0711 02:20:36.268971 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334592\n",
            "I0711 02:20:36.270680 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35348\n",
            "I0711 02:20:39.229462 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.6338944 (29.786 sec)\n",
            "I0711 02:20:39.231976 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337492\n",
            "I0711 02:20:39.234026 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39988\n",
            "I0711 02:20:42.209125 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335898\n",
            "I0711 02:20:42.210889 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37436\n",
            "I0711 02:20:45.164971 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338308\n",
            "I0711 02:20:45.166705 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41294\n",
            "I0711 02:20:48.121788 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338202\n",
            "I0711 02:20:48.123342 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41124\n",
            "I0711 02:20:51.090656 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336828\n",
            "I0711 02:20:51.092343 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38925\n",
            "I0711 02:20:54.085253 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333931\n",
            "I0711 02:20:54.086675 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3429\n",
            "I0711 02:20:57.082664 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333625\n",
            "I0711 02:20:57.084210 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33799\n",
            "I0711 02:21:00.058300 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336071\n",
            "I0711 02:21:00.059859 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37713\n",
            "I0711 02:21:03.061995 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332913\n",
            "I0711 02:21:03.063626 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32661\n",
            "I0711 02:21:06.071399 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332305\n",
            "I0711 02:21:06.073002 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31688\n",
            "I0711 02:21:09.036884 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.9696269 (29.807 sec)\n",
            "I0711 02:21:09.044879 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336291\n",
            "I0711 02:21:09.046882 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38065\n",
            "I0711 02:21:12.075377 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329983\n",
            "I0711 02:21:12.077597 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27973\n",
            "I0711 02:21:15.037084 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337649\n",
            "I0711 02:21:15.038584 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40238\n",
            "I0711 02:21:17.993355 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338269\n",
            "I0711 02:21:17.994946 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4123\n",
            "I0711 02:21:20.975168 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335358\n",
            "I0711 02:21:20.976453 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36573\n",
            "I0711 02:21:23.919385 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339645\n",
            "I0711 02:21:23.920958 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43432\n",
            "I0711 02:21:26.914387 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333891\n",
            "I0711 02:21:26.915896 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34226\n",
            "I0711 02:21:29.899177 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335029\n",
            "I0711 02:21:29.900766 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36046\n",
            "I0711 02:21:32.833131 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340841\n",
            "I0711 02:21:32.834945 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45346\n",
            "I0711 02:21:35.788682 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33835\n",
            "I0711 02:21:35.790202 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4136\n",
            "I0711 02:21:38.747736 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.90881157 (29.711 sec)\n",
            "I0711 02:21:38.755254 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337387\n",
            "I0711 02:21:38.757435 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3982\n",
            "I0711 02:21:41.701560 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339097\n",
            "I0711 02:21:41.703173 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42556\n",
            "I0711 02:21:44.688736 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334766\n",
            "I0711 02:21:44.690094 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35626\n",
            "I0711 02:21:47.593322 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34429\n",
            "I0711 02:21:47.597991 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50863\n",
            "I0711 02:21:50.527619 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340789\n",
            "I0711 02:21:50.529277 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45262\n",
            "I0711 02:21:53.431392 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344383\n",
            "I0711 02:21:53.433005 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51013\n",
            "I0711 02:21:56.335550 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344338\n",
            "I0711 02:21:56.337062 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50942\n",
            "I0711 02:21:59.256796 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342312\n",
            "I0711 02:21:59.258279 140338823006080 tpu_estimator.py:2160] examples/sec: 5.477\n",
            "I0711 02:22:02.181337 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341952\n",
            "I0711 02:22:02.183152 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47124\n",
            "I0711 02:22:05.152273 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33658\n",
            "I0711 02:22:05.153928 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38528\n",
            "I0711 02:22:08.087380 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2339449 (29.340 sec)\n",
            "I0711 02:22:08.089817 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340419\n",
            "I0711 02:22:08.092144 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4467\n",
            "I0711 02:22:11.031840 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3399\n",
            "I0711 02:22:11.033638 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43839\n",
            "I0711 02:22:13.979809 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339224\n",
            "I0711 02:22:13.981538 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42759\n",
            "I0711 02:22:16.964665 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335018\n",
            "I0711 02:22:16.966041 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36029\n",
            "I0711 02:22:19.961735 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333657\n",
            "I0711 02:22:19.963042 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33851\n",
            "I0711 02:22:22.934174 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336436\n",
            "I0711 02:22:22.935908 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38297\n",
            "I0711 02:22:26.019639 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.324094\n",
            "I0711 02:22:26.021023 140338823006080 tpu_estimator.py:2160] examples/sec: 5.18551\n",
            "I0711 02:22:29.011893 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3342\n",
            "I0711 02:22:29.013722 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3472\n",
            "I0711 02:22:32.019353 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332498\n",
            "I0711 02:22:32.020818 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31997\n",
            "I0711 02:22:35.066035 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328238\n",
            "I0711 02:22:35.067588 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25181\n",
            "I0711 02:22:38.063925 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.302005 (29.977 sec)\n",
            "I0711 02:22:38.070271 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332857\n",
            "I0711 02:22:38.074131 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32571\n",
            "I0711 02:22:41.033151 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337518\n",
            "I0711 02:22:41.034715 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40028\n",
            "I0711 02:22:43.982457 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339056\n",
            "I0711 02:22:43.983900 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42489\n",
            "I0711 02:22:46.915194 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340982\n",
            "I0711 02:22:46.916885 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45571\n",
            "I0711 02:22:49.875906 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337749\n",
            "I0711 02:22:49.877567 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40399\n",
            "I0711 02:22:52.833417 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338126\n",
            "I0711 02:22:52.834867 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41002\n",
            "I0711 02:22:55.743754 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343597\n",
            "I0711 02:22:55.745816 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49755\n",
            "I0711 02:22:58.674841 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341174\n",
            "I0711 02:22:58.680299 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45878\n",
            "I0711 02:23:01.629212 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338481\n",
            "I0711 02:23:01.630473 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4157\n",
            "I0711 02:23:04.620005 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334358\n",
            "I0711 02:23:04.621601 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34973\n",
            "I0711 02:23:07.586990 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2368665 (29.523 sec)\n",
            "I0711 02:23:07.596287 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33599\n",
            "I0711 02:23:07.597606 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37584\n",
            "I0711 02:23:10.579158 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335247\n",
            "I0711 02:23:10.580884 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36396\n",
            "I0711 02:23:13.589107 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332232\n",
            "I0711 02:23:13.591271 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31571\n",
            "I0711 02:23:16.572928 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335136\n",
            "I0711 02:23:16.574303 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36217\n",
            "I0711 02:23:19.545048 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336461\n",
            "I0711 02:23:19.546237 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38338\n",
            "I0711 02:23:22.515302 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336677\n",
            "I0711 02:23:22.517005 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38684\n",
            "I0711 02:23:25.528521 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331873\n",
            "I0711 02:23:25.530292 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30997\n",
            "I0711 02:23:28.470452 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339913\n",
            "I0711 02:23:28.471901 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43861\n",
            "I0711 02:23:31.455783 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334971\n",
            "I0711 02:23:31.457489 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35953\n",
            "I0711 02:23:34.432372 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335948\n",
            "I0711 02:23:34.434029 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37516\n",
            "I0711 02:23:37.411318 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.0913894 (29.824 sec)\n",
            "I0711 02:23:37.415968 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335169\n",
            "I0711 02:23:37.420493 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36271\n",
            "I0711 02:23:40.395553 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335622\n",
            "I0711 02:23:40.397395 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36995\n",
            "I0711 02:23:43.365396 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336713\n",
            "I0711 02:23:43.367058 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38741\n",
            "I0711 02:23:46.350976 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334941\n",
            "I0711 02:23:46.352566 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35905\n",
            "I0711 02:23:49.274531 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342048\n",
            "I0711 02:23:49.275940 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47277\n",
            "I0711 02:23:52.172213 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345112\n",
            "I0711 02:23:52.173600 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52179\n",
            "I0711 02:23:55.149040 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335925\n",
            "I0711 02:23:55.150488 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3748\n",
            "I0711 02:23:58.127594 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335735\n",
            "I0711 02:23:58.129112 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37175\n",
            "I0711 02:24:01.105893 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335759\n",
            "I0711 02:24:01.107686 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37214\n",
            "I0711 02:24:04.066586 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337756\n",
            "I0711 02:24:04.068319 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4041\n",
            "I0711 02:24:07.018450 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.4411349 (29.607 sec)\n",
            "I0711 02:24:07.024187 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338224\n",
            "I0711 02:24:07.027539 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41158\n",
            "I0711 02:24:09.944381 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342342\n",
            "I0711 02:24:09.946046 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47747\n",
            "I0711 02:24:12.939917 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333827\n",
            "I0711 02:24:12.941667 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34123\n",
            "I0711 02:24:15.997685 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.327029\n",
            "I0711 02:24:15.999209 140338823006080 tpu_estimator.py:2160] examples/sec: 5.23246\n",
            "I0711 02:24:19.018966 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330992\n",
            "I0711 02:24:19.020605 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29587\n",
            "I0711 02:24:22.048130 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330124\n",
            "I0711 02:24:22.049543 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28199\n",
            "I0711 02:24:25.014416 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337114\n",
            "I0711 02:24:25.016021 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39382\n",
            "I0711 02:24:27.956107 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339945\n",
            "I0711 02:24:27.957612 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43911\n",
            "I0711 02:24:30.936224 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335557\n",
            "I0711 02:24:30.937786 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36891\n",
            "I0711 02:24:33.917973 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335372\n",
            "I0711 02:24:33.919827 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36596\n",
            "I0711 02:24:36.924495 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.9347622 (29.906 sec)\n",
            "I0711 02:24:36.931823 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331805\n",
            "I0711 02:24:36.934834 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30888\n",
            "I0711 02:24:39.893018 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337694\n",
            "I0711 02:24:39.894835 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40311\n",
            "I0711 02:24:42.874101 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335451\n",
            "I0711 02:24:42.876169 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36722\n",
            "I0711 02:24:45.831034 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338187\n",
            "I0711 02:24:45.832159 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41099\n",
            "I0711 02:24:48.861001 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330043\n",
            "I0711 02:24:48.865003 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28068\n",
            "I0711 02:24:51.838165 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335886\n",
            "I0711 02:24:51.839828 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37417\n",
            "I0711 02:24:54.817040 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335696\n",
            "I0711 02:24:54.818568 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37113\n",
            "I0711 02:24:57.794683 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335835\n",
            "I0711 02:24:57.796126 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37336\n",
            "I0711 02:25:00.807503 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331913\n",
            "I0711 02:25:00.809015 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31062\n",
            "I0711 02:25:03.840860 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329677\n",
            "I0711 02:25:03.842389 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27484\n",
            "I0711 02:25:06.829109 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.77212524 (29.905 sec)\n",
            "I0711 02:25:06.835741 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333901\n",
            "I0711 02:25:06.839133 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34242\n",
            "I0711 02:25:09.834654 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333454\n",
            "I0711 02:25:09.836413 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33527\n",
            "I0711 02:25:12.829057 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333954\n",
            "I0711 02:25:12.830607 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34326\n",
            "I0711 02:25:15.850698 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330939\n",
            "I0711 02:25:15.852684 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29503\n",
            "I0711 02:25:18.837235 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334844\n",
            "I0711 02:25:18.838610 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35751\n",
            "I0711 02:25:21.826111 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334572\n",
            "I0711 02:25:21.827737 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35315\n",
            "I0711 02:25:24.850497 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330652\n",
            "I0711 02:25:24.852231 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29043\n",
            "I0711 02:25:27.836468 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334904\n",
            "I0711 02:25:27.838237 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35847\n",
            "I0711 02:25:30.846519 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332209\n",
            "I0711 02:25:30.848038 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31534\n",
            "I0711 02:25:33.810752 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337354\n",
            "I0711 02:25:33.812296 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39766\n",
            "I0711 02:25:36.840497 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.3363287 (30.011 sec)\n",
            "I0711 02:25:36.843501 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329732\n",
            "I0711 02:25:36.852104 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27572\n",
            "I0711 02:25:39.887154 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.32856\n",
            "I0711 02:25:39.888621 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25696\n",
            "I0711 02:25:42.808137 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342345\n",
            "I0711 02:25:42.809941 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47752\n",
            "I0711 02:25:45.850576 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.32868\n",
            "I0711 02:25:45.852198 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25889\n",
            "I0711 02:25:48.852822 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33309\n",
            "I0711 02:25:48.854336 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32944\n",
            "I0711 02:25:51.807923 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338397\n",
            "I0711 02:25:51.809679 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41436\n",
            "I0711 02:25:54.811876 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332892\n",
            "I0711 02:25:54.813395 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32628\n",
            "I0711 02:25:57.840136 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330227\n",
            "I0711 02:25:57.841843 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28363\n",
            "I0711 02:26:00.815134 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336139\n",
            "I0711 02:26:00.816646 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37822\n",
            "I0711 02:26:03.795968 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335477\n",
            "I0711 02:26:03.797632 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36763\n",
            "I0711 02:26:06.769176 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.1771026 (29.929 sec)\n",
            "I0711 02:26:06.774377 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335739\n",
            "I0711 02:26:06.778987 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37182\n",
            "I0711 02:26:09.767901 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334061\n",
            "I0711 02:26:09.769324 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34497\n",
            "I0711 02:26:12.801828 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329602\n",
            "I0711 02:26:12.803266 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27364\n",
            "I0711 02:26:15.746123 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339639\n",
            "I0711 02:26:15.747886 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43423\n",
            "I0711 02:26:18.712328 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337135\n",
            "I0711 02:26:18.713949 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39417\n",
            "I0711 02:26:21.740889 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330187\n",
            "I0711 02:26:21.742280 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28299\n",
            "I0711 02:26:24.706274 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337225\n",
            "I0711 02:26:24.708011 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3956\n",
            "I0711 02:26:27.693540 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334761\n",
            "I0711 02:26:27.695226 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35617\n",
            "I0711 02:26:30.690701 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333641\n",
            "I0711 02:26:30.692111 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33826\n",
            "I0711 02:26:33.691911 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333199\n",
            "I0711 02:26:33.693477 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33118\n",
            "I0711 02:26:36.641951 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.18321896 (29.873 sec)\n",
            "I0711 02:26:36.649030 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338161\n",
            "I0711 02:26:36.651975 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41057\n",
            "I0711 02:26:39.618520 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336765\n",
            "I0711 02:26:39.619973 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38824\n",
            "I0711 02:26:42.591576 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336361\n",
            "I0711 02:26:42.593021 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38177\n",
            "I0711 02:26:45.566574 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336126\n",
            "I0711 02:26:45.568007 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37801\n",
            "I0711 02:26:48.548102 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335401\n",
            "I0711 02:26:48.549508 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36642\n",
            "I0711 02:26:51.513944 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337173\n",
            "I0711 02:26:51.515413 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39476\n",
            "I0711 02:26:54.486126 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336451\n",
            "I0711 02:26:54.487660 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38321\n",
            "I0711 02:26:57.474142 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334674\n",
            "I0711 02:26:57.475610 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35478\n",
            "I0711 02:27:00.432424 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33804\n",
            "I0711 02:27:00.433995 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40864\n",
            "I0711 02:27:03.403896 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336525\n",
            "I0711 02:27:03.405437 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3844\n",
            "I0711 02:27:06.376432 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.1740975 (29.734 sec)\n",
            "I0711 02:27:06.381062 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336061\n",
            "I0711 02:27:06.382650 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37698\n",
            "I0711 02:27:09.360581 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335461\n",
            "I0711 02:27:09.361939 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36738\n",
            "I0711 02:27:12.333673 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336347\n",
            "I0711 02:27:12.335133 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38156\n",
            "I0711 02:27:15.316499 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335246\n",
            "I0711 02:27:15.318048 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36394\n",
            "I0711 02:27:18.314416 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33357\n",
            "I0711 02:27:18.317948 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33713\n",
            "I0711 02:27:21.292909 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335737\n",
            "I0711 02:27:21.294559 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3718\n",
            "I0711 02:27:24.271778 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3357\n",
            "I0711 02:27:24.273417 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37119\n",
            "I0711 02:27:27.260309 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334622\n",
            "I0711 02:27:27.262140 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35395\n",
            "I0711 02:27:30.227224 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33704\n",
            "I0711 02:27:30.228698 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39264\n",
            "I0711 02:27:33.177572 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33894\n",
            "I0711 02:27:33.179011 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42304\n",
            "I0711 02:27:36.135445 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.81614995 (29.759 sec)\n",
            "I0711 02:27:36.137813 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337811\n",
            "I0711 02:27:36.143560 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40497\n",
            "I0711 02:27:39.101553 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337412\n",
            "I0711 02:27:39.103333 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3986\n",
            "I0711 02:27:42.100322 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333474\n",
            "I0711 02:27:42.101605 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33558\n",
            "I0711 02:27:45.081638 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335422\n",
            "I0711 02:27:45.083165 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36675\n",
            "I0711 02:27:48.093020 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33207\n",
            "I0711 02:27:48.094688 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31313\n",
            "I0711 02:27:51.080977 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334682\n",
            "I0711 02:27:51.082662 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35491\n",
            "I0711 02:27:54.046514 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337204\n",
            "I0711 02:27:54.048117 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39526\n",
            "I0711 02:27:57.022000 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336081\n",
            "I0711 02:27:57.023435 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3773\n",
            "I0711 02:28:00.014618 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334169\n",
            "I0711 02:28:00.016663 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34671\n",
            "I0711 02:28:02.984901 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336657\n",
            "I0711 02:28:02.986568 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38652\n",
            "I0711 02:28:05.999159 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.7255807 (29.864 sec)\n",
            "I0711 02:28:06.006645 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33093\n",
            "I0711 02:28:06.010152 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29488\n",
            "I0711 02:28:08.954370 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339246\n",
            "I0711 02:28:08.956038 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42794\n",
            "I0711 02:28:11.948371 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333994\n",
            "I0711 02:28:11.949905 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34391\n",
            "I0711 02:28:14.935459 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334776\n",
            "I0711 02:28:14.937000 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35641\n",
            "I0711 02:28:17.931130 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333817\n",
            "I0711 02:28:17.932578 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34108\n",
            "I0711 02:28:20.902917 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336498\n",
            "I0711 02:28:20.904548 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38397\n",
            "I0711 02:28:23.887865 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335018\n",
            "I0711 02:28:23.889495 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36029\n",
            "I0711 02:28:26.899060 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332091\n",
            "I0711 02:28:26.900554 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31346\n",
            "I0711 02:28:29.878854 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335593\n",
            "I0711 02:28:29.880661 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36949\n",
            "I0711 02:28:32.874423 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333834\n",
            "I0711 02:28:32.875998 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34135\n",
            "I0711 02:28:35.830635 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.7586427 (29.831 sec)\n",
            "I0711 02:28:35.840267 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337166\n",
            "I0711 02:28:35.841990 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39465\n",
            "I0711 02:28:38.816781 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33596\n",
            "I0711 02:28:38.818323 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37536\n",
            "I0711 02:28:41.784214 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336996\n",
            "I0711 02:28:41.785825 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39193\n",
            "I0711 02:28:44.723398 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340227\n",
            "I0711 02:28:44.725007 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44364\n",
            "I0711 02:28:47.679349 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3383\n",
            "I0711 02:28:47.680934 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4128\n",
            "I0711 02:28:50.645911 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337108\n",
            "I0711 02:28:50.647599 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39372\n",
            "I0711 02:28:53.648829 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333\n",
            "I0711 02:28:53.650960 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32801\n",
            "I0711 02:28:56.686987 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329141\n",
            "I0711 02:28:56.688395 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26626\n",
            "I0711 02:28:59.674789 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334689\n",
            "I0711 02:28:59.676384 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35502\n",
            "I0711 02:29:02.655291 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335532\n",
            "I0711 02:29:02.656917 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36851\n",
            "I0711 02:29:05.598177 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.5440059 (29.768 sec)\n",
            "I0711 02:29:05.605903 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338902\n",
            "I0711 02:29:05.607113 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42243\n",
            "I0711 02:29:08.531834 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341773\n",
            "I0711 02:29:08.533715 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46836\n",
            "I0711 02:29:11.497767 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337155\n",
            "I0711 02:29:11.499457 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39448\n",
            "I0711 02:29:14.426436 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341456\n",
            "I0711 02:29:14.428169 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4633\n",
            "I0711 02:29:17.393156 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337071\n",
            "I0711 02:29:17.394620 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39314\n",
            "I0711 02:29:20.374348 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335431\n",
            "I0711 02:29:20.376524 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36689\n",
            "I0711 02:29:23.357496 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335225\n",
            "I0711 02:29:23.358776 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36359\n",
            "I0711 02:29:26.336241 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33571\n",
            "I0711 02:29:26.337622 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37135\n",
            "I0711 02:29:29.308615 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336433\n",
            "I0711 02:29:29.310364 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38292\n",
            "I0711 02:29:32.289931 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335423\n",
            "I0711 02:29:32.291583 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36677\n",
            "I0711 02:29:35.220330 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.5371666 (29.622 sec)\n",
            "I0711 02:29:35.229226 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340214\n",
            "I0711 02:29:35.230714 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44343\n",
            "I0711 02:29:38.205628 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335977\n",
            "I0711 02:29:38.206998 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37564\n",
            "I0711 02:29:41.212241 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332604\n",
            "I0711 02:29:41.213920 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32166\n",
            "I0711 02:29:44.203296 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33433\n",
            "I0711 02:29:44.205162 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34927\n",
            "I0711 02:29:47.189368 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334881\n",
            "I0711 02:29:47.190910 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3581\n",
            "I0711 02:29:50.201440 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332003\n",
            "I0711 02:29:50.206156 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31206\n",
            "I0711 02:29:53.194579 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334109\n",
            "I0711 02:29:53.197754 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34574\n",
            "I0711 02:29:56.186541 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334215\n",
            "I0711 02:29:56.191572 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34743\n",
            "I0711 02:29:59.153033 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337102\n",
            "I0711 02:29:59.154752 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39364\n",
            "I0711 02:30:02.148810 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333798\n",
            "I0711 02:30:02.150256 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34078\n",
            "I0711 02:30:05.114131 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0005345 (29.894 sec)\n",
            "I0711 02:30:05.126455 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335841\n",
            "I0711 02:30:05.128342 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37346\n",
            "I0711 02:30:08.083790 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338141\n",
            "I0711 02:30:08.085198 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41026\n",
            "I0711 02:30:11.062660 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335691\n",
            "I0711 02:30:11.064289 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37106\n",
            "I0711 02:30:14.050167 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33473\n",
            "I0711 02:30:14.051497 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35568\n",
            "I0711 02:30:17.011764 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337655\n",
            "I0711 02:30:17.013282 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40249\n",
            "I0711 02:30:19.993083 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335428\n",
            "I0711 02:30:19.994871 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36684\n",
            "I0711 02:30:23.027965 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329494\n",
            "I0711 02:30:23.029437 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27191\n",
            "I0711 02:30:26.010845 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335252\n",
            "I0711 02:30:26.012577 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36403\n",
            "I0711 02:30:28.985090 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336215\n",
            "I0711 02:30:28.986585 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37944\n",
            "I0711 02:30:31.965630 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335509\n",
            "I0711 02:30:31.967021 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36815\n",
            "I0711 02:30:34.942653 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.428694 (29.829 sec)\n",
            "I0711 02:30:34.945053 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335634\n",
            "I0711 02:30:34.947840 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37015\n",
            "I0711 02:30:37.932198 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334772\n",
            "I0711 02:30:37.933565 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35635\n",
            "I0711 02:30:40.934923 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333036\n",
            "I0711 02:30:40.936569 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32857\n",
            "I0711 02:30:43.940718 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332683\n",
            "I0711 02:30:43.942059 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32292\n",
            "I0711 02:30:46.939704 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333445\n",
            "I0711 02:30:46.941268 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33513\n",
            "I0711 02:30:49.917175 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335863\n",
            "I0711 02:30:49.918810 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37381\n",
            "I0711 02:30:52.937028 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331137\n",
            "I0711 02:30:52.938595 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2982\n",
            "I0711 02:30:55.928784 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334254\n",
            "I0711 02:30:55.930673 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34806\n",
            "I0711 02:30:58.926942 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333535\n",
            "I0711 02:30:58.928550 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33656\n",
            "I0711 02:31:02.080160 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.317138\n",
            "I0711 02:31:02.081546 140338823006080 tpu_estimator.py:2160] examples/sec: 5.0742\n",
            "I0711 02:31:05.085059 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.7322965 (30.142 sec)\n",
            "I0711 02:31:05.092818 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332038\n",
            "I0711 02:31:05.096210 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3126\n",
            "I0711 02:31:08.075222 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335196\n",
            "I0711 02:31:08.077138 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36314\n",
            "I0711 02:31:11.106287 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329914\n",
            "I0711 02:31:11.107982 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27863\n",
            "I0711 02:31:14.062371 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338285\n",
            "I0711 02:31:14.063914 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41256\n",
            "I0711 02:31:17.033309 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336598\n",
            "I0711 02:31:17.035034 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38557\n",
            "I0711 02:31:20.008935 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336063\n",
            "I0711 02:31:20.010333 140338823006080 tpu_estimator.py:2160] examples/sec: 5.377\n",
            "I0711 02:31:23.050648 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328756\n",
            "I0711 02:31:23.052823 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26009\n",
            "I0711 02:31:26.062127 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33207\n",
            "I0711 02:31:26.063766 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31313\n",
            "I0711 02:31:29.024812 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337528\n",
            "I0711 02:31:29.026278 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40045\n",
            "I0711 02:31:32.011132 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334862\n",
            "I0711 02:31:32.012459 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3578\n",
            "I0711 02:31:34.986378 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.5372224 (29.901 sec)\n",
            "I0711 02:31:34.995643 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335057\n",
            "I0711 02:31:34.998784 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36091\n",
            "I0711 02:31:37.988977 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334077\n",
            "I0711 02:31:37.990351 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34524\n",
            "I0711 02:31:40.954386 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337223\n",
            "I0711 02:31:40.956253 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39557\n",
            "I0711 02:31:43.904824 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338932\n",
            "I0711 02:31:43.906121 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42291\n",
            "I0711 02:31:46.933968 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330126\n",
            "I0711 02:31:46.935574 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28202\n",
            "I0711 02:31:49.895235 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337695\n",
            "I0711 02:31:49.896603 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40313\n",
            "I0711 02:31:52.965572 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.325697\n",
            "I0711 02:31:52.967171 140338823006080 tpu_estimator.py:2160] examples/sec: 5.21115\n",
            "I0711 02:31:55.930259 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337311\n",
            "I0711 02:31:55.931912 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39698\n",
            "I0711 02:31:58.904017 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336265\n",
            "I0711 02:31:58.905360 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38024\n",
            "I0711 02:32:01.868891 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337283\n",
            "I0711 02:32:01.870542 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39653\n",
            "I0711 02:32:04.813622 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.14655757 (29.827 sec)\n",
            "I0711 02:32:04.823290 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338489\n",
            "I0711 02:32:04.824684 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41583\n",
            "I0711 02:32:07.836416 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331873\n",
            "I0711 02:32:07.838190 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30997\n",
            "I0711 02:32:10.810133 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336281\n",
            "I0711 02:32:10.811890 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3805\n",
            "I0711 02:32:13.759797 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339017\n",
            "I0711 02:32:13.761179 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42428\n",
            "I0711 02:32:16.778150 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331313\n",
            "I0711 02:32:16.779579 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30101\n",
            "I0711 02:32:19.764403 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334863\n",
            "I0711 02:32:19.765698 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35781\n",
            "I0711 02:32:22.759804 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333845\n",
            "I0711 02:32:22.761276 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34153\n",
            "I0711 02:32:25.761976 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333094\n",
            "I0711 02:32:25.763385 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32951\n",
            "I0711 02:32:28.770987 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33233\n",
            "I0711 02:32:28.772385 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31729\n",
            "I0711 02:32:31.746703 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336055\n",
            "I0711 02:32:31.748301 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37688\n",
            "I0711 02:32:34.722960 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0756955 (29.909 sec)\n",
            "I0711 02:32:34.727359 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335494\n",
            "I0711 02:32:34.732850 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3679\n",
            "I0711 02:32:37.734473 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332546\n",
            "I0711 02:32:37.736264 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32073\n",
            "I0711 02:32:40.744503 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332223\n",
            "I0711 02:32:40.746105 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31557\n",
            "I0711 02:32:43.741934 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333623\n",
            "I0711 02:32:43.743578 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33796\n",
            "I0711 02:32:46.701592 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337876\n",
            "I0711 02:32:46.703125 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40602\n",
            "I0711 02:32:49.697723 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333772\n",
            "I0711 02:32:49.701516 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34034\n",
            "I0711 02:32:52.662820 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337249\n",
            "I0711 02:32:52.664378 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39598\n",
            "I0711 02:32:55.656129 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334079\n",
            "I0711 02:32:55.657710 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34526\n",
            "I0711 02:32:58.649571 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334061\n",
            "I0711 02:32:58.650904 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34498\n",
            "I0711 02:33:01.622025 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336425\n",
            "I0711 02:33:01.623574 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3828\n",
            "I0711 02:33:04.654372 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.6292076 (29.931 sec)\n",
            "I0711 02:33:04.656980 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329491\n",
            "I0711 02:33:04.659784 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27186\n",
            "I0711 02:33:07.609716 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338671\n",
            "I0711 02:33:07.617719 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41874\n",
            "I0711 02:33:10.587289 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335845\n",
            "I0711 02:33:10.588764 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37352\n",
            "I0711 02:33:13.553000 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337187\n",
            "I0711 02:33:13.554709 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39499\n",
            "I0711 02:33:16.541722 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334592\n",
            "I0711 02:33:16.545644 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35348\n",
            "I0711 02:33:19.542489 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333245\n",
            "I0711 02:33:19.544020 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33191\n",
            "I0711 02:33:22.528113 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334952\n",
            "I0711 02:33:22.529545 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35924\n",
            "I0711 02:33:25.582959 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.327334\n",
            "I0711 02:33:25.585179 140338823006080 tpu_estimator.py:2160] examples/sec: 5.23734\n",
            "I0711 02:33:28.538814 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338313\n",
            "I0711 02:33:28.540172 140338823006080 tpu_estimator.py:2160] examples/sec: 5.413\n",
            "I0711 02:33:31.525948 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334795\n",
            "I0711 02:33:31.527878 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35672\n",
            "I0711 02:33:34.495223 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.8557751 (29.841 sec)\n",
            "I0711 02:33:34.498491 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336391\n",
            "I0711 02:33:34.505337 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38226\n",
            "I0711 02:33:37.487801 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33452\n",
            "I0711 02:33:37.489169 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35231\n",
            "I0711 02:33:40.429228 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339976\n",
            "I0711 02:33:40.430912 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43961\n",
            "I0711 02:33:43.383736 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338465\n",
            "I0711 02:33:43.385186 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41544\n",
            "I0711 02:33:46.372867 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334541\n",
            "I0711 02:33:46.374348 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35266\n",
            "I0711 02:33:49.360958 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334671\n",
            "I0711 02:33:49.362596 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35474\n",
            "I0711 02:33:52.343781 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335249\n",
            "I0711 02:33:52.345385 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36398\n",
            "I0711 02:33:55.297943 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338507\n",
            "I0711 02:33:55.299403 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41612\n",
            "I0711 02:33:58.228712 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341202\n",
            "I0711 02:33:58.230352 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45924\n",
            "I0711 02:34:01.183440 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338446\n",
            "I0711 02:34:01.184882 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41514\n",
            "I0711 02:34:04.104183 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2604588 (29.609 sec)\n",
            "I0711 02:34:04.108601 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341855\n",
            "I0711 02:34:04.113957 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46969\n",
            "I0711 02:34:07.077430 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336834\n",
            "I0711 02:34:07.078860 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38935\n",
            "I0711 02:34:10.025000 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339262\n",
            "I0711 02:34:10.026421 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42819\n",
            "I0711 02:34:12.993620 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336858\n",
            "I0711 02:34:12.995000 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38973\n",
            "I0711 02:34:15.979606 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334899\n",
            "I0711 02:34:15.981119 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35839\n",
            "I0711 02:34:18.954914 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336097\n",
            "I0711 02:34:18.956653 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37755\n",
            "I0711 02:34:21.882097 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341626\n",
            "I0711 02:34:21.883751 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46602\n",
            "I0711 02:34:24.836888 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338434\n",
            "I0711 02:34:24.838350 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41494\n",
            "I0711 02:34:27.837685 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33325\n",
            "I0711 02:34:27.839327 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33201\n",
            "I0711 02:34:30.831896 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333971\n",
            "I0711 02:34:30.833280 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34354\n",
            "I0711 02:34:33.814690 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.2567348 (29.711 sec)\n",
            "I0711 02:34:33.823929 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33423\n",
            "I0711 02:34:33.826358 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34767\n",
            "I0711 02:34:36.795891 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33647\n",
            "I0711 02:34:36.797374 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38352\n",
            "I0711 02:34:39.757375 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33767\n",
            "I0711 02:34:39.759142 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40272\n",
            "I0711 02:34:42.720549 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337471\n",
            "I0711 02:34:42.721999 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39954\n",
            "I0711 02:34:45.649502 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341421\n",
            "I0711 02:34:45.651021 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46274\n",
            "I0711 02:34:48.590351 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340044\n",
            "I0711 02:34:48.591939 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4407\n",
            "I0711 02:34:51.571777 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335405\n",
            "I0711 02:34:51.573658 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36648\n",
            "I0711 02:34:54.542928 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336568\n",
            "I0711 02:34:54.544389 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38508\n",
            "I0711 02:34:57.530705 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334699\n",
            "I0711 02:34:57.532269 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35518\n",
            "I0711 02:35:00.504503 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336274\n",
            "I0711 02:35:00.506305 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38038\n",
            "I0711 02:35:03.493355 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.8314698 (29.679 sec)\n",
            "I0711 02:35:03.496150 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334262\n",
            "I0711 02:35:03.502781 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3482\n",
            "I0711 02:35:06.481038 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335023\n",
            "I0711 02:35:06.483144 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36037\n",
            "I0711 02:35:09.428058 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339331\n",
            "I0711 02:35:09.429846 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42929\n",
            "I0711 02:35:12.379994 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33875\n",
            "I0711 02:35:12.381496 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41999\n",
            "I0711 02:35:15.338347 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33803\n",
            "I0711 02:35:15.339777 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40847\n",
            "I0711 02:35:18.303526 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337247\n",
            "I0711 02:35:18.304943 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39595\n",
            "I0711 02:35:21.294037 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334388\n",
            "I0711 02:35:21.295642 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3502\n",
            "I0711 02:35:24.255875 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337628\n",
            "I0711 02:35:24.257573 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40205\n",
            "I0711 02:35:27.219517 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33742\n",
            "I0711 02:35:27.221917 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39873\n",
            "I0711 02:35:30.195846 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336002\n",
            "I0711 02:35:30.197462 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37603\n",
            "I0711 02:35:33.163057 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.1378169 (29.670 sec)\n",
            "I0711 02:35:33.171416 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336055\n",
            "I0711 02:35:33.172843 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37688\n",
            "I0711 02:35:36.166679 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333864\n",
            "I0711 02:35:36.168026 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34183\n",
            "I0711 02:35:39.088865 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342217\n",
            "I0711 02:35:39.090600 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47547\n",
            "I0711 02:35:42.064415 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336065\n",
            "I0711 02:35:42.065953 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37704\n",
            "I0711 02:35:45.001406 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34049\n",
            "I0711 02:35:45.003628 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44784\n",
            "I0711 02:35:47.920933 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342517\n",
            "I0711 02:35:47.922368 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48028\n",
            "I0711 02:35:50.861691 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340048\n",
            "I0711 02:35:50.863226 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44077\n",
            "I0711 02:35:53.781734 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342461\n",
            "I0711 02:35:53.783138 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47938\n",
            "I0711 02:35:56.746152 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337331\n",
            "I0711 02:35:56.747885 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3973\n",
            "I0711 02:35:59.699517 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338597\n",
            "I0711 02:35:59.701092 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41756\n",
            "I0711 02:36:02.644372 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.78348684 (29.481 sec)\n",
            "I0711 02:36:02.654458 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338422\n",
            "I0711 02:36:02.657337 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41475\n",
            "I0711 02:36:05.623178 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336852\n",
            "I0711 02:36:05.624734 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38963\n",
            "I0711 02:36:08.544338 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342323\n",
            "I0711 02:36:08.546326 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47717\n",
            "I0711 02:36:11.458310 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343173\n",
            "I0711 02:36:11.459972 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49077\n",
            "I0711 02:36:14.391908 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340873\n",
            "I0711 02:36:14.393528 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45397\n",
            "I0711 02:36:17.347686 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338323\n",
            "I0711 02:36:17.349123 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41317\n",
            "I0711 02:36:20.313844 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337137\n",
            "I0711 02:36:20.315678 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3942\n",
            "I0711 02:36:23.250149 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340562\n",
            "I0711 02:36:23.251484 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44899\n",
            "I0711 02:36:26.206400 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338265\n",
            "I0711 02:36:26.207779 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41224\n",
            "I0711 02:36:29.120651 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343139\n",
            "I0711 02:36:29.121888 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49023\n",
            "I0711 02:36:32.028512 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.239797 (29.384 sec)\n",
            "I0711 02:36:32.031179 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343577\n",
            "I0711 02:36:32.032984 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49724\n",
            "I0711 02:36:34.980154 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339107\n",
            "I0711 02:36:34.981552 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42572\n",
            "I0711 02:36:37.959815 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335609\n",
            "I0711 02:36:37.961632 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36974\n",
            "I0711 02:36:40.931481 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336511\n",
            "I0711 02:36:40.933269 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38418\n",
            "I0711 02:36:43.871293 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340157\n",
            "I0711 02:36:43.873147 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44251\n",
            "I0711 02:36:46.841481 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336678\n",
            "I0711 02:36:46.842893 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38685\n",
            "I0711 02:36:49.798833 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338151\n",
            "I0711 02:36:49.800462 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41041\n",
            "I0711 02:36:52.759178 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337797\n",
            "I0711 02:36:52.760740 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40475\n",
            "I0711 02:36:55.695492 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340591\n",
            "I0711 02:36:55.697690 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44946\n",
            "I0711 02:36:58.647695 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338692\n",
            "I0711 02:36:58.649475 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41907\n",
            "I0711 02:37:01.601216 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.3610232 (29.573 sec)\n",
            "I0711 02:37:01.608975 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337758\n",
            "I0711 02:37:01.611748 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40413\n",
            "I0711 02:37:04.640262 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329836\n",
            "I0711 02:37:04.641949 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27737\n",
            "I0711 02:37:07.577717 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340426\n",
            "I0711 02:37:07.579530 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44682\n",
            "I0711 02:37:10.511062 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340906\n",
            "I0711 02:37:10.512775 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4545\n",
            "I0711 02:37:13.455527 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339625\n",
            "I0711 02:37:13.457020 140338823006080 tpu_estimator.py:2160] examples/sec: 5.434\n",
            "I0711 02:37:16.435435 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33559\n",
            "I0711 02:37:16.436900 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36945\n",
            "I0711 02:37:19.414308 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335686\n",
            "I0711 02:37:19.415801 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37098\n",
            "I0711 02:37:22.369446 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338394\n",
            "I0711 02:37:22.371131 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4143\n",
            "I0711 02:37:25.349730 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335545\n",
            "I0711 02:37:25.351155 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36872\n",
            "I0711 02:37:28.319540 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336714\n",
            "I0711 02:37:28.321666 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38742\n",
            "I0711 02:37:31.286387 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.8370733 (29.685 sec)\n",
            "I0711 02:37:31.293742 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336252\n",
            "I0711 02:37:31.295576 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38003\n",
            "I0711 02:37:34.284158 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334381\n",
            "I0711 02:37:34.285881 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3501\n",
            "I0711 02:37:37.270475 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33486\n",
            "I0711 02:37:37.272027 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35777\n",
            "I0711 02:37:40.239228 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336839\n",
            "I0711 02:37:40.240711 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38942\n",
            "I0711 02:37:43.219484 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33554\n",
            "I0711 02:37:43.221184 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36865\n",
            "I0711 02:37:46.233527 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331783\n",
            "I0711 02:37:46.235060 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30853\n",
            "I0711 02:37:49.198551 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337261\n",
            "I0711 02:37:49.200199 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39617\n",
            "I0711 02:37:52.182467 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335132\n",
            "I0711 02:37:52.184190 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36211\n",
            "I0711 02:37:55.180472 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333556\n",
            "I0711 02:37:55.182105 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33689\n",
            "I0711 02:37:58.163121 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335279\n",
            "I0711 02:37:58.164550 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36446\n",
            "I0711 02:38:01.122978 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.7326994 (29.837 sec)\n",
            "I0711 02:38:01.129860 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337064\n",
            "I0711 02:38:01.133815 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39302\n",
            "I0711 02:38:04.093404 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337449\n",
            "I0711 02:38:04.095416 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39918\n",
            "I0711 02:38:07.067368 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336236\n",
            "I0711 02:38:07.068887 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37977\n",
            "I0711 02:38:10.051707 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335084\n",
            "I0711 02:38:10.052996 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36134\n",
            "I0711 02:38:12.988342 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34053\n",
            "I0711 02:38:12.990223 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44847\n",
            "I0711 02:38:15.946850 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338001\n",
            "I0711 02:38:15.948498 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40802\n",
            "I0711 02:38:18.930322 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33518\n",
            "I0711 02:38:18.931888 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36288\n",
            "I0711 02:38:21.887187 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338198\n",
            "I0711 02:38:21.888628 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41117\n",
            "I0711 02:38:24.930542 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328584\n",
            "I0711 02:38:24.931820 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25735\n",
            "I0711 02:38:27.890058 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337896\n",
            "I0711 02:38:27.891739 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40634\n",
            "I0711 02:38:30.848215 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.5414107 (29.725 sec)\n",
            "I0711 02:38:30.851417 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337679\n",
            "I0711 02:38:30.852694 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40286\n",
            "I0711 02:38:33.802640 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338847\n",
            "I0711 02:38:33.804151 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42155\n",
            "I0711 02:38:36.772012 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336771\n",
            "I0711 02:38:36.773743 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38834\n",
            "I0711 02:38:39.751192 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335658\n",
            "I0711 02:38:39.752856 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37053\n",
            "I0711 02:38:42.732141 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335468\n",
            "I0711 02:38:42.733567 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36749\n",
            "I0711 02:38:45.729647 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333609\n",
            "I0711 02:38:45.731262 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33774\n",
            "I0711 02:38:48.737146 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3325\n",
            "I0711 02:38:48.738869 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32\n",
            "I0711 02:38:51.717384 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335547\n",
            "I0711 02:38:51.718697 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36875\n",
            "I0711 02:38:54.700603 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335209\n",
            "I0711 02:38:54.701840 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36334\n",
            "I0711 02:38:57.687016 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334854\n",
            "I0711 02:38:57.688857 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35767\n",
            "I0711 02:39:00.641582 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.13279486 (29.793 sec)\n",
            "I0711 02:39:00.645591 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337991\n",
            "I0711 02:39:00.651271 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40786\n",
            "I0711 02:39:03.668526 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330812\n",
            "I0711 02:39:03.670302 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29299\n",
            "I0711 02:39:06.617651 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33908\n",
            "I0711 02:39:06.618983 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42527\n",
            "I0711 02:39:09.555706 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340359\n",
            "I0711 02:39:09.557121 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44575\n",
            "I0711 02:39:12.561259 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332723\n",
            "I0711 02:39:12.562888 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32356\n",
            "I0711 02:39:15.535961 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336161\n",
            "I0711 02:39:15.537392 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37858\n",
            "I0711 02:39:18.514319 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335756\n",
            "I0711 02:39:18.515975 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3721\n",
            "I0711 02:39:21.489058 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336162\n",
            "I0711 02:39:21.490732 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3786\n",
            "I0711 02:39:24.479844 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33437\n",
            "I0711 02:39:24.481498 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34993\n",
            "I0711 02:39:27.449953 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336679\n",
            "I0711 02:39:27.451394 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38687\n",
            "I0711 02:39:30.489713 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0234709 (29.848 sec)\n",
            "I0711 02:39:30.500865 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.32777\n",
            "I0711 02:39:30.502901 140338823006080 tpu_estimator.py:2160] examples/sec: 5.24433\n",
            "I0711 02:39:33.508678 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332467\n",
            "I0711 02:39:33.510101 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31948\n",
            "I0711 02:39:36.493682 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335014\n",
            "I0711 02:39:36.495404 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36022\n",
            "I0711 02:39:39.466602 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336373\n",
            "I0711 02:39:39.468211 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38197\n",
            "I0711 02:39:42.425169 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338\n",
            "I0711 02:39:42.426700 140338823006080 tpu_estimator.py:2160] examples/sec: 5.408\n",
            "I0711 02:39:45.401434 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335985\n",
            "I0711 02:39:45.402990 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37575\n",
            "I0711 02:39:48.421019 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331169\n",
            "I0711 02:39:48.422439 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29871\n",
            "I0711 02:39:51.380462 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337916\n",
            "I0711 02:39:51.382170 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40665\n",
            "I0711 02:39:54.347839 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336988\n",
            "I0711 02:39:54.349474 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39181\n",
            "I0711 02:39:57.322028 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336223\n",
            "I0711 02:39:57.324141 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37956\n",
            "I0711 02:40:00.309034 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.17800498 (29.819 sec)\n",
            "I0711 02:40:00.311566 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334512\n",
            "I0711 02:40:00.314590 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35219\n",
            "I0711 02:40:03.312125 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333261\n",
            "I0711 02:40:03.313730 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33218\n",
            "I0711 02:40:06.286817 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336171\n",
            "I0711 02:40:06.288428 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37873\n",
            "I0711 02:40:09.256415 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336741\n",
            "I0711 02:40:09.257900 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38785\n",
            "I0711 02:40:12.230572 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33624\n",
            "I0711 02:40:12.232135 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37983\n",
            "I0711 02:40:15.222717 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334203\n",
            "I0711 02:40:15.224620 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34724\n",
            "I0711 02:40:18.173294 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338916\n",
            "I0711 02:40:18.175425 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42266\n",
            "I0711 02:40:21.180845 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332495\n",
            "I0711 02:40:21.182286 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31992\n",
            "I0711 02:40:24.216392 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329429\n",
            "I0711 02:40:24.218095 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27087\n",
            "I0711 02:40:27.168086 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338789\n",
            "I0711 02:40:27.169871 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42063\n",
            "I0711 02:40:30.109178 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.524809 (29.800 sec)\n",
            "I0711 02:40:30.117741 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339022\n",
            "I0711 02:40:30.119379 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42435\n",
            "I0711 02:40:33.038178 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342418\n",
            "I0711 02:40:33.039915 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47869\n",
            "I0711 02:40:35.982844 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339594\n",
            "I0711 02:40:35.984242 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4335\n",
            "I0711 02:40:38.990260 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332518\n",
            "I0711 02:40:38.991972 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3203\n",
            "I0711 02:40:41.918117 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341544\n",
            "I0711 02:40:41.919636 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46471\n",
            "I0711 02:40:44.935547 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331411\n",
            "I0711 02:40:44.941461 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30257\n",
            "I0711 02:40:47.900440 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337275\n",
            "I0711 02:40:47.901912 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39641\n",
            "I0711 02:40:50.882465 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33535\n",
            "I0711 02:40:50.883927 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36561\n",
            "I0711 02:40:53.861562 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335663\n",
            "I0711 02:40:53.862846 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37061\n",
            "I0711 02:40:56.832499 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336604\n",
            "I0711 02:40:56.833954 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38566\n",
            "I0711 02:40:59.780951 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.65228844 (29.672 sec)\n",
            "I0711 02:40:59.786380 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338527\n",
            "I0711 02:40:59.788099 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41643\n",
            "I0711 02:41:02.766237 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335586\n",
            "I0711 02:41:02.768043 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36938\n",
            "I0711 02:41:05.727280 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337727\n",
            "I0711 02:41:05.728981 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40364\n",
            "I0711 02:41:08.709444 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335325\n",
            "I0711 02:41:08.710948 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36519\n",
            "I0711 02:41:11.660233 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33889\n",
            "I0711 02:41:11.661850 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42224\n",
            "I0711 02:41:14.653491 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334085\n",
            "I0711 02:41:14.655148 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34536\n",
            "I0711 02:41:17.641173 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334718\n",
            "I0711 02:41:17.643729 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35548\n",
            "I0711 02:41:20.613305 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336451\n",
            "I0711 02:41:20.614796 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38321\n",
            "I0711 02:41:23.557908 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339598\n",
            "I0711 02:41:23.559726 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43356\n",
            "I0711 02:41:26.528889 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336594\n",
            "I0711 02:41:26.530566 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38551\n",
            "I0711 02:41:29.480675 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.10040283 (29.700 sec)\n",
            "I0711 02:41:29.488789 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337892\n",
            "I0711 02:41:29.490431 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40627\n",
            "I0711 02:41:32.444705 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338261\n",
            "I0711 02:41:32.446825 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41218\n",
            "I0711 02:41:35.362447 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34273\n",
            "I0711 02:41:35.364007 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48367\n",
            "I0711 02:41:38.395903 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329658\n",
            "I0711 02:41:38.397679 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27453\n",
            "I0711 02:41:41.377373 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335401\n",
            "I0711 02:41:41.378974 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36642\n",
            "I0711 02:41:44.301491 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341983\n",
            "I0711 02:41:44.303286 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47173\n",
            "I0711 02:41:47.238694 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340463\n",
            "I0711 02:41:47.240037 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44741\n",
            "I0711 02:41:50.148286 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343693\n",
            "I0711 02:41:50.149857 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49909\n",
            "I0711 02:41:53.152635 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332845\n",
            "I0711 02:41:53.154059 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32552\n",
            "I0711 02:41:56.086550 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340846\n",
            "I0711 02:41:56.088163 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45354\n",
            "I0711 02:41:59.055674 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.5339525 (29.575 sec)\n",
            "I0711 02:41:59.058378 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336489\n",
            "I0711 02:41:59.064131 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38383\n",
            "I0711 02:42:02.029266 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336601\n",
            "I0711 02:42:02.030800 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38561\n",
            "I0711 02:42:05.022745 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334063\n",
            "I0711 02:42:05.024393 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34501\n",
            "I0711 02:42:08.008516 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33492\n",
            "I0711 02:42:08.010105 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35872\n",
            "I0711 02:42:10.992889 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335076\n",
            "I0711 02:42:10.994276 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36122\n",
            "I0711 02:42:13.934133 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339994\n",
            "I0711 02:42:13.935616 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43991\n",
            "I0711 02:42:16.914621 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335513\n",
            "I0711 02:42:16.916249 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36821\n",
            "I0711 02:42:19.895714 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335448\n",
            "I0711 02:42:19.896896 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36717\n",
            "I0711 02:42:22.927408 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.32985\n",
            "I0711 02:42:22.928808 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2776\n",
            "I0711 02:42:25.914364 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334785\n",
            "I0711 02:42:25.915918 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35655\n",
            "I0711 02:42:28.941671 140338823006080 basic_session_run_hooks.py:260] batch_loss = 8.665791 (29.886 sec)\n",
            "I0711 02:42:28.950903 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329324\n",
            "I0711 02:42:28.952089 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26919\n",
            "I0711 02:42:31.900762 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338998\n",
            "I0711 02:42:31.902176 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42397\n",
            "I0711 02:42:34.873369 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336419\n",
            "I0711 02:42:34.874892 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3827\n",
            "I0711 02:42:37.842885 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336743\n",
            "I0711 02:42:37.844592 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38789\n",
            "I0711 02:42:40.807798 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337278\n",
            "I0711 02:42:40.809462 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39644\n",
            "I0711 02:42:43.801543 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334035\n",
            "I0711 02:42:43.803344 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34456\n",
            "I0711 02:42:46.782850 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335418\n",
            "I0711 02:42:46.784284 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36669\n",
            "I0711 02:42:49.760517 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335831\n",
            "I0711 02:42:49.761861 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37329\n",
            "I0711 02:42:52.774138 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331849\n",
            "I0711 02:42:52.775766 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30958\n",
            "I0711 02:42:55.781407 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332508\n",
            "I0711 02:42:55.782908 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32013\n",
            "I0711 02:42:58.751383 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.7872195 (29.810 sec)\n",
            "I0711 02:42:58.758988 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336108\n",
            "I0711 02:42:58.761605 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37772\n",
            "I0711 02:43:01.766563 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332234\n",
            "I0711 02:43:01.768150 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31574\n",
            "I0711 02:43:04.772274 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332704\n",
            "I0711 02:43:04.773865 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32327\n",
            "I0711 02:43:07.735438 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337475\n",
            "I0711 02:43:07.737231 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3996\n",
            "I0711 02:43:10.723526 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334669\n",
            "I0711 02:43:10.725162 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3547\n",
            "I0711 02:43:13.712131 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334595\n",
            "I0711 02:43:13.713708 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35352\n",
            "I0711 02:43:16.697468 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334981\n",
            "I0711 02:43:16.698922 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35969\n",
            "I0711 02:43:19.701309 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332901\n",
            "I0711 02:43:19.703130 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32642\n",
            "I0711 02:43:22.671450 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336688\n",
            "I0711 02:43:22.672870 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38701\n",
            "I0711 02:43:25.676578 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33276\n",
            "I0711 02:43:25.678250 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32415\n",
            "I0711 02:43:28.655903 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.5673084 (29.905 sec)\n",
            "I0711 02:43:28.663982 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334757\n",
            "I0711 02:43:28.666014 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35611\n",
            "I0711 02:43:31.676604 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331923\n",
            "I0711 02:43:31.678403 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31076\n",
            "I0711 02:43:34.694166 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331385\n",
            "I0711 02:43:34.696488 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30217\n",
            "I0711 02:43:37.697765 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33294\n",
            "I0711 02:43:37.699138 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32704\n",
            "I0711 02:43:40.701908 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332879\n",
            "I0711 02:43:40.703325 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32606\n",
            "I0711 02:43:43.706712 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3328\n",
            "I0711 02:43:43.708290 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32479\n",
            "I0711 02:43:46.724319 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331387\n",
            "I0711 02:43:46.725705 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30218\n",
            "I0711 02:43:49.710582 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334862\n",
            "I0711 02:43:49.712516 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35778\n",
            "I0711 02:43:52.704513 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334006\n",
            "I0711 02:43:52.706008 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3441\n",
            "I0711 02:43:55.655415 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338887\n",
            "I0711 02:43:55.657213 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42219\n",
            "I0711 02:43:58.644903 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.63651824 (29.989 sec)\n",
            "I0711 02:43:58.648687 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334078\n",
            "I0711 02:43:58.653487 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34524\n",
            "I0711 02:44:01.643913 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33387\n",
            "I0711 02:44:01.645587 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34192\n",
            "I0711 02:44:04.615286 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336541\n",
            "I0711 02:44:04.616609 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38466\n",
            "I0711 02:44:07.610889 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333833\n",
            "I0711 02:44:07.612292 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34133\n",
            "I0711 02:44:10.566768 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3383\n",
            "I0711 02:44:10.568173 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4128\n",
            "I0711 02:44:13.571580 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332795\n",
            "I0711 02:44:13.572896 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32472\n",
            "I0711 02:44:16.525902 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338486\n",
            "I0711 02:44:16.527463 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41578\n",
            "I0711 02:44:19.506036 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335561\n",
            "I0711 02:44:19.507433 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36897\n",
            "I0711 02:44:22.502138 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333772\n",
            "I0711 02:44:22.503960 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34035\n",
            "I0711 02:44:25.491960 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334457\n",
            "I0711 02:44:25.493679 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35131\n",
            "I0711 02:44:28.478175 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.80709624 (29.833 sec)\n",
            "I0711 02:44:28.486717 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333918\n",
            "I0711 02:44:28.487953 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34269\n",
            "I0711 02:44:31.478360 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334274\n",
            "I0711 02:44:31.480024 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34838\n",
            "I0711 02:44:34.427645 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339058\n",
            "I0711 02:44:34.428979 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42493\n",
            "I0711 02:44:37.394054 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337108\n",
            "I0711 02:44:37.397846 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39372\n",
            "I0711 02:44:40.386057 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334221\n",
            "I0711 02:44:40.387523 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34754\n",
            "I0711 02:44:43.356445 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336657\n",
            "I0711 02:44:43.358001 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38652\n",
            "I0711 02:44:46.319516 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337488\n",
            "I0711 02:44:46.320862 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39981\n",
            "I0711 02:44:49.248579 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341411\n",
            "I0711 02:44:49.250200 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46258\n",
            "I0711 02:44:52.199245 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338905\n",
            "I0711 02:44:52.200628 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42248\n",
            "I0711 02:44:55.141415 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339883\n",
            "I0711 02:44:55.142965 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43812\n",
            "I0711 02:44:58.108933 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.3987589 (29.631 sec)\n",
            "I0711 02:44:58.115201 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336273\n",
            "I0711 02:44:58.118365 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38036\n",
            "I0711 02:45:01.085124 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33671\n",
            "I0711 02:45:01.086521 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38736\n",
            "I0711 02:45:04.031301 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339429\n",
            "I0711 02:45:04.032922 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43086\n",
            "I0711 02:45:06.981580 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338947\n",
            "I0711 02:45:06.982857 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42315\n",
            "I0711 02:45:09.924385 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339814\n",
            "I0711 02:45:09.925776 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43702\n",
            "I0711 02:45:12.893894 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336757\n",
            "I0711 02:45:12.895793 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3881\n",
            "I0711 02:45:15.867824 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336261\n",
            "I0711 02:45:15.869653 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38018\n",
            "I0711 02:45:18.809254 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339972\n",
            "I0711 02:45:18.810587 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43954\n",
            "I0711 02:45:21.765743 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338229\n",
            "I0711 02:45:21.767149 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41166\n",
            "I0711 02:45:24.736173 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336651\n",
            "I0711 02:45:24.737747 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38642\n",
            "I0711 02:45:27.721156 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.051657677 (29.612 sec)\n",
            "I0711 02:45:27.723904 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334702\n",
            "I0711 02:45:27.732246 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35524\n",
            "I0711 02:45:30.673094 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339085\n",
            "I0711 02:45:30.678694 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42535\n",
            "I0711 02:45:33.657262 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335099\n",
            "I0711 02:45:33.659094 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36158\n",
            "I0711 02:45:36.609392 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338728\n",
            "I0711 02:45:36.611252 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41965\n",
            "I0711 02:45:39.557950 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339152\n",
            "I0711 02:45:39.559631 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42644\n",
            "I0711 02:45:42.499565 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339948\n",
            "I0711 02:45:42.500900 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43917\n",
            "I0711 02:45:45.439735 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340118\n",
            "I0711 02:45:45.441257 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44189\n",
            "I0711 02:45:48.363476 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342024\n",
            "I0711 02:45:48.364929 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47238\n",
            "I0711 02:45:51.303000 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340197\n",
            "I0711 02:45:51.304476 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44315\n",
            "I0711 02:45:54.208696 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344152\n",
            "I0711 02:45:54.210384 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50643\n",
            "I0711 02:45:57.113635 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.7707043 (29.392 sec)\n",
            "I0711 02:45:57.121734 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343287\n",
            "I0711 02:45:57.123313 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49259\n",
            "I0711 02:46:00.116944 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33386\n",
            "I0711 02:46:00.118726 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34177\n",
            "I0711 02:46:03.020419 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344423\n",
            "I0711 02:46:03.022037 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51076\n",
            "I0711 02:46:05.967162 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339353\n",
            "I0711 02:46:05.968677 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42965\n",
            "I0711 02:46:08.956676 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334501\n",
            "I0711 02:46:08.958251 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35201\n",
            "I0711 02:46:11.879410 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342149\n",
            "I0711 02:46:11.881011 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47438\n",
            "I0711 02:46:14.792605 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343264\n",
            "I0711 02:46:14.794222 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49222\n",
            "I0711 02:46:17.700347 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343924\n",
            "I0711 02:46:17.701834 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50279\n",
            "I0711 02:46:20.667461 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337015\n",
            "I0711 02:46:20.669189 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39225\n",
            "I0711 02:46:23.577529 140338823006080 basic_session_run_hooks.py:606] Saving checkpoints for 2000 into /content/drive/My Drive/ner_checkpoints/model.ckpt.\n",
            "I0711 02:46:53.911048 140338823006080 estimator.py:1145] Calling model_fn.\n",
            "I0711 02:46:53.912420 140338823006080 tpu_estimator.py:2965] Running eval on CPU\n",
            "I0711 02:46:53.917379 140338823006080 <ipython-input-9-2f5db7366c31>:282] *** Features ***\n",
            "I0711 02:46:53.921310 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_ids, shape = (?, 128)\n",
            "I0711 02:46:53.923958 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_mask, shape = (?, 128)\n",
            "I0711 02:46:53.926148 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = label_ids, shape = (?, 128)\n",
            "I0711 02:46:53.928794 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = segment_ids, shape = (?, 128)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "shape of input_ids (?, 128)\n",
            "shape of label_ids (?, 128)\n",
            "seq_length 128\n",
            "lengths Tensor(\"Sum:0\", shape=(?,), dtype=int32)\n",
            "####################\n",
            "shape of output_layer: (?, 128, 1024)\n",
            "embedding_size:1024\n",
            "seq_length:128\n",
            "shape of logit (?, 128, 11)\n",
            "shape of loss ()\n",
            "shape of per_example_loss (?,)\n",
            "num labels:11\n",
            "####################\n",
            "shape of pred_ids (?, 128)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "I0711 02:47:01.576615 140338823006080 estimator.py:1147] Done calling model_fn.\n",
            "I0711 02:47:01.607501 140338823006080 evaluation.py:255] Starting evaluation at 2019-07-11T02:47:01Z\n",
            "I0711 02:47:02.845706 140338823006080 monitored_session.py:240] Graph was finalized.\n",
            "I0711 02:47:02.876128 140338823006080 saver.py:1280] Restoring parameters from /content/drive/My Drive/ner_checkpoints/model.ckpt-2000\n",
            "I0711 02:47:11.779714 140338823006080 session_manager.py:500] Running local_init_op.\n",
            "I0711 02:47:11.889309 140338823006080 session_manager.py:502] Done running local_init_op.\n",
            "I0711 02:47:18.935450 140338823006080 evaluation.py:167] Evaluation [10/100]\n",
            "I0711 02:47:24.136502 140338823006080 evaluation.py:167] Evaluation [20/100]\n",
            "I0711 02:47:29.295559 140338823006080 evaluation.py:167] Evaluation [30/100]\n",
            "I0711 02:47:34.459670 140338823006080 evaluation.py:167] Evaluation [40/100]\n",
            "I0711 02:47:39.820130 140338823006080 evaluation.py:167] Evaluation [50/100]\n",
            "I0711 02:47:44.664634 140338823006080 evaluation.py:167] Evaluation [60/100]\n",
            "I0711 02:47:49.535715 140338823006080 evaluation.py:167] Evaluation [70/100]\n",
            "I0711 02:47:54.833688 140338823006080 evaluation.py:167] Evaluation [80/100]\n",
            "I0711 02:48:00.154877 140338823006080 evaluation.py:167] Evaluation [90/100]\n",
            "I0711 02:48:05.828824 140338823006080 evaluation.py:167] Evaluation [100/100]\n",
            "I0711 02:48:06.052124 140338823006080 evaluation.py:275] Finished evaluation at 2019-07-11-02:48:06\n",
            "I0711 02:48:06.053911 140338823006080 estimator.py:2039] Saving dict for global step 2000: eval_accuracy = 0.9868324, eval_f = 0.9341754, eval_loss = 1.4451218, eval_precision = 0.93667805, eval_recall = 0.93168604, global_step = 2000, loss = 1.4451218\n",
            "I0711 02:48:06.058482 140338823006080 estimator.py:2099] Saving 'checkpoint_path' summary for global step 2000: /content/drive/My Drive/ner_checkpoints/model.ckpt-2000\n",
            "I0711 02:48:06.061501 140338823006080 error_handling.py:96] evaluation_loop marked as finished\n",
            "I0711 02:48:06.098533 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.00948487\n",
            "I0711 02:48:06.100542 140338823006080 tpu_estimator.py:2160] examples/sec: 0.151758\n",
            "I0711 02:48:09.045662 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.5069947 (131.932 sec)\n",
            "I0711 02:48:09.050594 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338749\n",
            "I0711 02:48:09.056498 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41998\n",
            "I0711 02:48:11.997595 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339318\n",
            "I0711 02:48:11.999083 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42909\n",
            "I0711 02:48:14.907631 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343642\n",
            "I0711 02:48:14.909357 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49827\n",
            "I0711 02:48:17.818925 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343485\n",
            "I0711 02:48:17.820576 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49576\n",
            "I0711 02:48:20.729232 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343615\n",
            "I0711 02:48:20.730816 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49785\n",
            "I0711 02:48:23.652315 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342106\n",
            "I0711 02:48:23.654157 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4737\n",
            "I0711 02:48:26.581539 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34138\n",
            "I0711 02:48:26.583688 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46208\n",
            "I0711 02:48:29.553379 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336503\n",
            "I0711 02:48:29.555629 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38405\n",
            "I0711 02:48:32.490113 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340504\n",
            "I0711 02:48:32.491636 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44807\n",
            "I0711 02:48:35.462444 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336436\n",
            "I0711 02:48:35.464127 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38298\n",
            "I0711 02:48:38.377001 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.044816494 (29.331 sec)\n",
            "I0711 02:48:38.384592 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342214\n",
            "I0711 02:48:38.386805 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47543\n",
            "I0711 02:48:41.314237 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341344\n",
            "I0711 02:48:41.315802 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46151\n",
            "I0711 02:48:44.281969 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336951\n",
            "I0711 02:48:44.283524 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39122\n",
            "I0711 02:48:47.237756 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338317\n",
            "I0711 02:48:47.239197 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41308\n",
            "I0711 02:48:50.274715 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329276\n",
            "I0711 02:48:50.276262 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26841\n",
            "I0711 02:48:53.269005 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333975\n",
            "I0711 02:48:53.270626 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34361\n",
            "I0711 02:48:56.266856 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333569\n",
            "I0711 02:48:56.268496 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33711\n",
            "I0711 02:48:59.235597 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336846\n",
            "I0711 02:48:59.237697 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38953\n",
            "I0711 02:49:02.218405 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335251\n",
            "I0711 02:49:02.219881 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36402\n",
            "I0711 02:49:05.189245 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336612\n",
            "I0711 02:49:05.190837 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3858\n",
            "I0711 02:49:08.151457 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.25569582 (29.774 sec)\n",
            "I0711 02:49:08.154211 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337272\n",
            "I0711 02:49:08.156031 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39636\n",
            "I0711 02:49:11.138368 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335104\n",
            "I0711 02:49:11.139743 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36166\n",
            "I0711 02:49:14.131185 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334127\n",
            "I0711 02:49:14.132636 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34603\n",
            "I0711 02:49:17.114270 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335221\n",
            "I0711 02:49:17.115576 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36353\n",
            "I0711 02:49:20.091193 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335919\n",
            "I0711 02:49:20.092708 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37471\n",
            "I0711 02:49:23.076136 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335016\n",
            "I0711 02:49:23.077887 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36025\n",
            "I0711 02:49:26.100269 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330679\n",
            "I0711 02:49:26.101944 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29087\n",
            "I0711 02:49:29.085113 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335016\n",
            "I0711 02:49:29.086546 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36025\n",
            "I0711 02:49:32.058156 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336365\n",
            "I0711 02:49:32.059895 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38185\n",
            "I0711 02:49:35.099650 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328785\n",
            "I0711 02:49:35.101050 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26056\n",
            "I0711 02:49:38.104285 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.6925716 (29.953 sec)\n",
            "I0711 02:49:38.110990 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332073\n",
            "I0711 02:49:38.113845 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31317\n",
            "I0711 02:49:41.122701 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332039\n",
            "I0711 02:49:41.124119 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31263\n",
            "I0711 02:49:44.111494 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334587\n",
            "I0711 02:49:44.113420 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35339\n",
            "I0711 02:49:47.081294 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336721\n",
            "I0711 02:49:47.083012 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38754\n",
            "I0711 02:49:50.053943 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336391\n",
            "I0711 02:49:50.055806 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38226\n",
            "I0711 02:49:53.033449 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335631\n",
            "I0711 02:49:53.034873 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3701\n",
            "I0711 02:49:55.958657 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341851\n",
            "I0711 02:49:55.960338 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46961\n",
            "I0711 02:49:58.908826 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338969\n",
            "I0711 02:49:58.910305 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4235\n",
            "I0711 02:50:01.856219 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33929\n",
            "I0711 02:50:01.857951 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42865\n",
            "I0711 02:50:04.820492 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337342\n",
            "I0711 02:50:04.822042 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39747\n",
            "I0711 02:50:07.753802 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.4197743 (29.650 sec)\n",
            "I0711 02:50:07.757007 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340532\n",
            "I0711 02:50:07.764374 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44851\n",
            "I0711 02:50:10.718929 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337624\n",
            "I0711 02:50:10.720545 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40199\n",
            "I0711 02:50:13.660830 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339922\n",
            "I0711 02:50:13.662372 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43875\n",
            "I0711 02:50:16.592573 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341098\n",
            "I0711 02:50:16.594144 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45756\n",
            "I0711 02:50:19.518627 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341754\n",
            "I0711 02:50:19.520278 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46806\n",
            "I0711 02:50:22.408947 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345976\n",
            "I0711 02:50:22.410315 140338823006080 tpu_estimator.py:2160] examples/sec: 5.53562\n",
            "I0711 02:50:25.331213 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342204\n",
            "I0711 02:50:25.332607 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47527\n",
            "I0711 02:50:28.231827 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34475\n",
            "I0711 02:50:28.233663 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51601\n",
            "I0711 02:50:31.169434 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340413\n",
            "I0711 02:50:31.170806 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44661\n",
            "I0711 02:50:34.078628 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34374\n",
            "I0711 02:50:34.080147 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49984\n",
            "I0711 02:50:37.030559 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.3463883 (29.277 sec)\n",
            "I0711 02:50:37.034408 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338321\n",
            "I0711 02:50:37.036859 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41314\n",
            "I0711 02:50:39.965933 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341121\n",
            "I0711 02:50:39.970823 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45793\n",
            "I0711 02:50:42.871113 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344219\n",
            "I0711 02:50:42.872845 140338823006080 tpu_estimator.py:2160] examples/sec: 5.5075\n",
            "I0711 02:50:45.804288 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340926\n",
            "I0711 02:50:45.805910 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45481\n",
            "I0711 02:50:48.763704 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3379\n",
            "I0711 02:50:48.765099 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4064\n",
            "I0711 02:50:51.659362 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34534\n",
            "I0711 02:50:51.660723 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52543\n",
            "I0711 02:50:54.625408 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337152\n",
            "I0711 02:50:54.626873 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39443\n",
            "I0711 02:50:57.544893 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342523\n",
            "I0711 02:50:57.546626 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48036\n",
            "I0711 02:51:00.495307 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338943\n",
            "I0711 02:51:00.496967 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42309\n",
            "I0711 02:51:03.431042 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340621\n",
            "I0711 02:51:03.432503 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44994\n",
            "I0711 02:51:06.334846 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.023500204 (29.304 sec)\n",
            "I0711 02:51:06.337487 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344067\n",
            "I0711 02:51:06.342351 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50507\n",
            "I0711 02:51:09.281135 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339716\n",
            "I0711 02:51:09.282589 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43545\n",
            "I0711 02:51:12.198348 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342789\n",
            "I0711 02:51:12.199720 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48462\n",
            "I0711 02:51:15.132135 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34086\n",
            "I0711 02:51:15.133800 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45376\n",
            "I0711 02:51:18.071835 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340168\n",
            "I0711 02:51:18.073444 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44268\n",
            "I0711 02:51:21.027851 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338294\n",
            "I0711 02:51:21.029432 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4127\n",
            "I0711 02:51:23.970144 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339879\n",
            "I0711 02:51:23.972047 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43806\n",
            "I0711 02:51:26.924294 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338501\n",
            "I0711 02:51:26.925974 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41601\n",
            "I0711 02:51:29.868874 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339605\n",
            "I0711 02:51:29.870152 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43368\n",
            "I0711 02:51:32.837967 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336802\n",
            "I0711 02:51:32.839348 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38884\n",
            "I0711 02:51:35.776176 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.27477777 (29.441 sec)\n",
            "I0711 02:51:35.779131 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340005\n",
            "I0711 02:51:35.781495 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44008\n",
            "I0711 02:51:38.692169 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343282\n",
            "I0711 02:51:38.693688 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49252\n",
            "I0711 02:51:41.597970 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344139\n",
            "I0711 02:51:41.599649 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50622\n",
            "I0711 02:51:44.579712 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335374\n",
            "I0711 02:51:44.581111 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36599\n",
            "I0711 02:51:47.502741 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34211\n",
            "I0711 02:51:47.504110 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47377\n",
            "I0711 02:51:50.446145 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339744\n",
            "I0711 02:51:50.447951 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43591\n",
            "I0711 02:51:53.376502 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341256\n",
            "I0711 02:51:53.378170 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46009\n",
            "I0711 02:51:56.305990 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341354\n",
            "I0711 02:51:56.307476 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46166\n",
            "I0711 02:51:59.211292 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344203\n",
            "I0711 02:51:59.212967 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50726\n",
            "I0711 02:52:02.126231 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343063\n",
            "I0711 02:52:02.127802 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48901\n",
            "I0711 02:52:05.037581 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.036274195 (29.261 sec)\n",
            "I0711 02:52:05.045423 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342572\n",
            "I0711 02:52:05.047995 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48114\n",
            "I0711 02:52:08.012289 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337042\n",
            "I0711 02:52:08.014454 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39267\n",
            "I0711 02:52:10.928258 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342938\n",
            "I0711 02:52:10.930083 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48701\n",
            "I0711 02:52:13.870279 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339902\n",
            "I0711 02:52:13.871901 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43844\n",
            "I0711 02:52:16.775526 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344208\n",
            "I0711 02:52:16.776752 140338823006080 tpu_estimator.py:2160] examples/sec: 5.50732\n",
            "I0711 02:52:19.674529 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.344943\n",
            "I0711 02:52:19.676003 140338823006080 tpu_estimator.py:2160] examples/sec: 5.51908\n",
            "I0711 02:52:22.604709 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341275\n",
            "I0711 02:52:22.606137 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4604\n",
            "I0711 02:52:25.596952 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334198\n",
            "I0711 02:52:25.598322 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34716\n",
            "I0711 02:52:28.508686 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343437\n",
            "I0711 02:52:28.510016 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49499\n",
            "I0711 02:52:31.445210 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340545\n",
            "I0711 02:52:31.446557 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44871\n",
            "I0711 02:52:34.433448 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.5284407 (29.396 sec)\n",
            "I0711 02:52:34.439062 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334011\n",
            "I0711 02:52:34.440735 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34417\n",
            "I0711 02:52:37.436887 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333581\n",
            "I0711 02:52:37.438303 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33729\n",
            "I0711 02:52:40.446226 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3323\n",
            "I0711 02:52:40.447696 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3168\n",
            "I0711 02:52:43.435051 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334577\n",
            "I0711 02:52:43.436505 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35324\n",
            "I0711 02:52:46.432617 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333607\n",
            "I0711 02:52:46.434120 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33771\n",
            "I0711 02:52:49.382788 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338971\n",
            "I0711 02:52:49.384198 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42354\n",
            "I0711 02:52:52.380954 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333528\n",
            "I0711 02:52:52.382572 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33646\n",
            "I0711 02:52:55.317895 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340483\n",
            "I0711 02:52:55.319506 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44773\n",
            "I0711 02:52:58.253129 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340693\n",
            "I0711 02:52:58.254514 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45109\n",
            "I0711 02:53:01.219017 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337165\n",
            "I0711 02:53:01.220467 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39464\n",
            "I0711 02:53:04.149224 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.25223327 (29.716 sec)\n",
            "I0711 02:53:04.151256 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341032\n",
            "I0711 02:53:04.152794 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45651\n",
            "I0711 02:53:07.139394 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334661\n",
            "I0711 02:53:07.140745 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35457\n",
            "I0711 02:53:10.109113 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336734\n",
            "I0711 02:53:10.110554 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38775\n",
            "I0711 02:53:13.034303 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341864\n",
            "I0711 02:53:13.035722 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46982\n",
            "I0711 02:53:15.982131 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339231\n",
            "I0711 02:53:15.983474 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4277\n",
            "I0711 02:53:18.975687 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334049\n",
            "I0711 02:53:18.977193 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34479\n",
            "I0711 02:53:21.955137 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335638\n",
            "I0711 02:53:21.956697 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3702\n",
            "I0711 02:53:24.934455 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335637\n",
            "I0711 02:53:24.935799 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37019\n",
            "I0711 02:53:27.844166 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343682\n",
            "I0711 02:53:27.845734 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49891\n",
            "I0711 02:53:30.787705 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339732\n",
            "I0711 02:53:30.789171 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43572\n",
            "I0711 02:53:33.709694 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.839154 (29.560 sec)\n",
            "I0711 02:53:33.714118 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34173\n",
            "I0711 02:53:33.717205 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46767\n",
            "I0711 02:53:36.637870 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342\n",
            "I0711 02:53:36.639529 140338823006080 tpu_estimator.py:2160] examples/sec: 5.472\n",
            "I0711 02:53:39.590871 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33864\n",
            "I0711 02:53:39.592700 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41823\n",
            "I0711 02:53:42.525398 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340775\n",
            "I0711 02:53:42.527146 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4524\n",
            "I0711 02:53:45.505153 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335604\n",
            "I0711 02:53:45.506813 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36966\n",
            "I0711 02:53:48.422947 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342721\n",
            "I0711 02:53:48.424518 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48354\n",
            "I0711 02:53:51.397353 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336196\n",
            "I0711 02:53:51.398553 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37914\n",
            "I0711 02:53:54.391988 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333932\n",
            "I0711 02:53:54.393462 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34292\n",
            "I0711 02:53:57.327178 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340705\n",
            "I0711 02:53:57.328908 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45129\n",
            "I0711 02:54:00.296431 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33677\n",
            "I0711 02:54:00.298330 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38833\n",
            "I0711 02:54:03.270371 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.062436104 (29.561 sec)\n",
            "I0711 02:54:03.279886 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335211\n",
            "I0711 02:54:03.281376 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36338\n",
            "I0711 02:54:06.256892 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335881\n",
            "I0711 02:54:06.258280 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37409\n",
            "I0711 02:54:09.275779 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331253\n",
            "I0711 02:54:09.277816 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30004\n",
            "I0711 02:54:12.278031 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333088\n",
            "I0711 02:54:12.279525 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3294\n",
            "I0711 02:54:15.267422 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33451\n",
            "I0711 02:54:15.269210 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35216\n",
            "I0711 02:54:18.232402 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337271\n",
            "I0711 02:54:18.233766 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39634\n",
            "I0711 02:54:21.248374 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331566\n",
            "I0711 02:54:21.249785 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30506\n",
            "I0711 02:54:24.176533 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341505\n",
            "I0711 02:54:24.177917 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46408\n",
            "I0711 02:54:27.194588 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331342\n",
            "I0711 02:54:27.196151 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30147\n",
            "I0711 02:54:30.198092 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332949\n",
            "I0711 02:54:30.199390 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32719\n",
            "I0711 02:54:33.170703 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.100694895 (29.900 sec)\n",
            "I0711 02:54:33.178820 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335525\n",
            "I0711 02:54:33.180642 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36839\n",
            "I0711 02:54:36.157190 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335714\n",
            "I0711 02:54:36.158626 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37143\n",
            "I0711 02:54:39.126557 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336773\n",
            "I0711 02:54:39.127897 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38837\n",
            "I0711 02:54:42.107013 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335517\n",
            "I0711 02:54:42.108601 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36827\n",
            "I0711 02:54:45.073423 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337112\n",
            "I0711 02:54:45.075047 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3938\n",
            "I0711 02:54:48.064503 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334336\n",
            "I0711 02:54:48.065980 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34938\n",
            "I0711 02:54:51.044358 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335584\n",
            "I0711 02:54:51.045861 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36934\n",
            "I0711 02:54:54.088183 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328529\n",
            "I0711 02:54:54.089872 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25647\n",
            "I0711 02:54:57.085039 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333683\n",
            "I0711 02:54:57.088418 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33893\n",
            "I0711 02:55:00.118669 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329633\n",
            "I0711 02:55:00.120336 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27413\n",
            "I0711 02:55:03.124467 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.6196871 (29.954 sec)\n",
            "I0711 02:55:03.131305 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331939\n",
            "I0711 02:55:03.134006 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31103\n",
            "I0711 02:55:06.145961 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331711\n",
            "I0711 02:55:06.147658 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30737\n",
            "I0711 02:55:09.167588 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330951\n",
            "I0711 02:55:09.169343 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29522\n",
            "I0711 02:55:12.162038 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333955\n",
            "I0711 02:55:12.163728 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34328\n",
            "I0711 02:55:15.168327 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332638\n",
            "I0711 02:55:15.169897 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32221\n",
            "I0711 02:55:18.130794 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33756\n",
            "I0711 02:55:18.132389 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40096\n",
            "I0711 02:55:21.118000 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334748\n",
            "I0711 02:55:21.119559 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35596\n",
            "I0711 02:55:24.079197 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337709\n",
            "I0711 02:55:24.080823 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40335\n",
            "I0711 02:55:27.061729 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335279\n",
            "I0711 02:55:27.063198 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36446\n",
            "I0711 02:55:30.042088 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335527\n",
            "I0711 02:55:30.043550 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36844\n",
            "I0711 02:55:33.012919 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2398851 (29.888 sec)\n",
            "I0711 02:55:33.020797 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335717\n",
            "I0711 02:55:33.023456 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37147\n",
            "I0711 02:55:35.957485 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340516\n",
            "I0711 02:55:35.959042 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44826\n",
            "I0711 02:55:38.939678 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335325\n",
            "I0711 02:55:38.941222 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3652\n",
            "I0711 02:55:41.893013 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338598\n",
            "I0711 02:55:41.894455 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41757\n",
            "I0711 02:55:44.802820 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343671\n",
            "I0711 02:55:44.804222 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49873\n",
            "I0711 02:55:47.757588 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338436\n",
            "I0711 02:55:47.758944 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41497\n",
            "I0711 02:55:50.742953 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334965\n",
            "I0711 02:55:50.744465 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35943\n",
            "I0711 02:55:53.670003 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341655\n",
            "I0711 02:55:53.671532 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46648\n",
            "I0711 02:55:56.632648 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337529\n",
            "I0711 02:55:56.634274 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40046\n",
            "I0711 02:55:59.578438 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33946\n",
            "I0711 02:55:59.579888 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43136\n",
            "I0711 02:56:02.585605 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.88696766 (29.573 sec)\n",
            "I0711 02:56:02.593939 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33171\n",
            "I0711 02:56:02.595923 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30736\n",
            "I0711 02:56:05.579041 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334906\n",
            "I0711 02:56:05.580798 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35849\n",
            "I0711 02:56:08.523370 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339635\n",
            "I0711 02:56:08.524747 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43415\n",
            "I0711 02:56:11.517163 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334035\n",
            "I0711 02:56:11.519100 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34456\n",
            "I0711 02:56:14.433784 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342854\n",
            "I0711 02:56:14.435137 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48566\n",
            "I0711 02:56:17.378399 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339604\n",
            "I0711 02:56:17.380316 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43367\n",
            "I0711 02:56:20.350798 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336425\n",
            "I0711 02:56:20.352332 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38281\n",
            "I0711 02:56:23.330619 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335597\n",
            "I0711 02:56:23.332167 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36956\n",
            "I0711 02:56:26.269575 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340252\n",
            "I0711 02:56:26.271124 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44403\n",
            "I0711 02:56:29.178925 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343724\n",
            "I0711 02:56:29.180515 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49959\n",
            "I0711 02:56:32.114779 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.48946404 (29.529 sec)\n",
            "I0711 02:56:32.121902 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339796\n",
            "I0711 02:56:32.124224 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43674\n",
            "I0711 02:56:35.041465 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342505\n",
            "I0711 02:56:35.043124 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48008\n",
            "I0711 02:56:37.979257 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340398\n",
            "I0711 02:56:37.980778 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44636\n",
            "I0711 02:56:40.944781 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337206\n",
            "I0711 02:56:40.946336 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39529\n",
            "I0711 02:56:43.886389 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339948\n",
            "I0711 02:56:43.888031 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43917\n",
            "I0711 02:56:46.828611 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339881\n",
            "I0711 02:56:46.830060 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43809\n",
            "I0711 02:56:49.753908 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34185\n",
            "I0711 02:56:49.755347 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46959\n",
            "I0711 02:56:52.674203 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342428\n",
            "I0711 02:56:52.675528 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47884\n",
            "I0711 02:56:55.658497 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335089\n",
            "I0711 02:56:55.660118 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36142\n",
            "I0711 02:56:58.613843 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338368\n",
            "I0711 02:56:58.615276 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41389\n",
            "I0711 02:57:01.550120 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0039527 (29.435 sec)\n",
            "I0711 02:57:01.560536 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339365\n",
            "I0711 02:57:01.563615 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42984\n",
            "I0711 02:57:04.542504 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335375\n",
            "I0711 02:57:04.544463 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36601\n",
            "I0711 02:57:07.477244 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340739\n",
            "I0711 02:57:07.478742 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45182\n",
            "I0711 02:57:10.502344 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330564\n",
            "I0711 02:57:10.503868 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28903\n",
            "I0711 02:57:13.481014 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335703\n",
            "I0711 02:57:13.482520 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37125\n",
            "I0711 02:57:16.447984 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337042\n",
            "I0711 02:57:16.449687 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39268\n",
            "I0711 02:57:19.432824 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33503\n",
            "I0711 02:57:19.434460 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36048\n",
            "I0711 02:57:22.421144 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334639\n",
            "I0711 02:57:22.422588 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35423\n",
            "I0711 02:57:25.376843 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338323\n",
            "I0711 02:57:25.378306 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41317\n",
            "I0711 02:57:28.395187 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331314\n",
            "I0711 02:57:28.396714 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30102\n",
            "I0711 02:57:31.402942 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.33183 (29.853 sec)\n",
            "I0711 02:57:31.405369 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332209\n",
            "I0711 02:57:31.407136 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31534\n",
            "I0711 02:57:34.370142 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337288\n",
            "I0711 02:57:34.376437 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3966\n",
            "I0711 02:57:37.373268 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332992\n",
            "I0711 02:57:37.374772 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32787\n",
            "I0711 02:57:40.371643 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333507\n",
            "I0711 02:57:40.373044 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33611\n",
            "I0711 02:57:43.376667 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332777\n",
            "I0711 02:57:43.378405 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32444\n",
            "I0711 02:57:46.327923 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338845\n",
            "I0711 02:57:46.329456 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42152\n",
            "I0711 02:57:49.304627 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335931\n",
            "I0711 02:57:49.306356 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3749\n",
            "I0711 02:57:52.285966 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335423\n",
            "I0711 02:57:52.287588 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36677\n",
            "I0711 02:57:55.226410 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340086\n",
            "I0711 02:57:55.228130 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44138\n",
            "I0711 02:57:58.156113 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341329\n",
            "I0711 02:57:58.157753 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46126\n",
            "I0711 02:58:01.148599 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.1550179 (29.746 sec)\n",
            "I0711 02:58:01.156051 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333339\n",
            "I0711 02:58:01.158814 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33342\n",
            "I0711 02:58:04.111121 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338405\n",
            "I0711 02:58:04.112651 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41449\n",
            "I0711 02:58:07.094050 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335242\n",
            "I0711 02:58:07.095553 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36388\n",
            "I0711 02:58:10.040052 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33944\n",
            "I0711 02:58:10.041473 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43103\n",
            "I0711 02:58:13.005241 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337248\n",
            "I0711 02:58:13.007430 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39596\n",
            "I0711 02:58:16.045978 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328866\n",
            "I0711 02:58:16.047398 140338823006080 tpu_estimator.py:2160] examples/sec: 5.26186\n",
            "I0711 02:58:19.107429 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.326675\n",
            "I0711 02:58:19.110036 140338823006080 tpu_estimator.py:2160] examples/sec: 5.22679\n",
            "I0711 02:58:22.098136 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334343\n",
            "I0711 02:58:22.099608 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34948\n",
            "I0711 02:58:25.091881 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334024\n",
            "I0711 02:58:25.093618 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34438\n",
            "I0711 02:58:28.072455 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335506\n",
            "I0711 02:58:28.073787 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36809\n",
            "I0711 02:58:31.062017 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.46167326 (29.913 sec)\n",
            "I0711 02:58:31.065650 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334087\n",
            "I0711 02:58:31.071486 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3454\n",
            "I0711 02:58:34.062276 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333714\n",
            "I0711 02:58:34.063990 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33943\n",
            "I0711 02:58:37.070789 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332391\n",
            "I0711 02:58:37.072635 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31826\n",
            "I0711 02:58:40.065671 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333903\n",
            "I0711 02:58:40.067159 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34244\n",
            "I0711 02:58:43.020250 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338456\n",
            "I0711 02:58:43.021628 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41529\n",
            "I0711 02:58:46.002144 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335361\n",
            "I0711 02:58:46.003678 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36578\n",
            "I0711 02:58:48.991321 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334532\n",
            "I0711 02:58:48.992836 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35252\n",
            "I0711 02:58:51.991038 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333367\n",
            "I0711 02:58:51.992754 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33387\n",
            "I0711 02:58:54.972568 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335397\n",
            "I0711 02:58:54.974104 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36635\n",
            "I0711 02:58:57.947468 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336156\n",
            "I0711 02:58:57.948874 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37849\n",
            "I0711 02:59:00.946511 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.6034226 (29.884 sec)\n",
            "I0711 02:59:00.955590 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332424\n",
            "I0711 02:59:00.956855 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31879\n",
            "I0711 02:59:03.930881 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336102\n",
            "I0711 02:59:03.932519 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37764\n",
            "I0711 02:59:06.872437 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339953\n",
            "I0711 02:59:06.874151 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43925\n",
            "I0711 02:59:09.827007 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338464\n",
            "I0711 02:59:09.828733 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41543\n",
            "I0711 02:59:12.837985 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332115\n",
            "I0711 02:59:12.839838 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31383\n",
            "I0711 02:59:15.854524 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331511\n",
            "I0711 02:59:15.856096 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30418\n",
            "I0711 02:59:18.832823 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335759\n",
            "I0711 02:59:18.834322 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37215\n",
            "I0711 02:59:21.785949 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338624\n",
            "I0711 02:59:21.787963 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41798\n",
            "I0711 02:59:24.761949 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336019\n",
            "I0711 02:59:24.763432 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37631\n",
            "I0711 02:59:27.726544 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33732\n",
            "I0711 02:59:27.728114 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39712\n",
            "I0711 02:59:30.683112 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.3238442 (29.737 sec)\n",
            "I0711 02:59:30.693167 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337082\n",
            "I0711 02:59:30.696570 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39331\n",
            "I0711 02:59:33.679270 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334883\n",
            "I0711 02:59:33.680643 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35813\n",
            "I0711 02:59:36.665722 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334844\n",
            "I0711 02:59:36.667203 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3575\n",
            "I0711 02:59:39.631100 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337227\n",
            "I0711 02:59:39.632480 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39563\n",
            "I0711 02:59:42.621455 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334404\n",
            "I0711 02:59:42.622906 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35046\n",
            "I0711 02:59:45.579833 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338034\n",
            "I0711 02:59:45.581219 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40854\n",
            "I0711 02:59:48.553424 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33629\n",
            "I0711 02:59:48.555641 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38064\n",
            "I0711 02:59:51.525279 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336498\n",
            "I0711 02:59:51.526721 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38396\n",
            "I0711 02:59:54.510394 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334985\n",
            "I0711 02:59:54.511666 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35975\n",
            "I0711 02:59:57.475936 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337203\n",
            "I0711 02:59:57.477557 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39524\n",
            "I0711 03:00:00.450167 140338823006080 basic_session_run_hooks.py:260] batch_loss = 2.2519379 (29.767 sec)\n",
            "I0711 03:00:00.459663 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33515\n",
            "I0711 03:00:00.461098 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3624\n",
            "I0711 03:00:03.427347 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336973\n",
            "I0711 03:00:03.428891 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39156\n",
            "I0711 03:00:06.378232 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338877\n",
            "I0711 03:00:06.379739 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42203\n",
            "I0711 03:00:09.387020 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332356\n",
            "I0711 03:00:09.388774 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3177\n",
            "I0711 03:00:12.390788 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332924\n",
            "I0711 03:00:12.392123 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32678\n",
            "I0711 03:00:15.361685 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336595\n",
            "I0711 03:00:15.363732 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38551\n",
            "I0711 03:00:18.303531 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339933\n",
            "I0711 03:00:18.305036 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43893\n",
            "I0711 03:00:21.291193 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3347\n",
            "I0711 03:00:21.292586 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35521\n",
            "I0711 03:00:24.253118 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33762\n",
            "I0711 03:00:24.254640 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40193\n",
            "I0711 03:00:27.213957 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337738\n",
            "I0711 03:00:27.215573 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40381\n",
            "I0711 03:00:30.160758 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.6719074 (29.711 sec)\n",
            "I0711 03:00:30.169672 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33834\n",
            "I0711 03:00:30.171311 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41344\n",
            "I0711 03:00:33.153485 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335146\n",
            "I0711 03:00:33.154914 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36233\n",
            "I0711 03:00:36.158576 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332752\n",
            "I0711 03:00:36.159782 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32404\n",
            "I0711 03:00:39.119295 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337757\n",
            "I0711 03:00:39.120906 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40412\n",
            "I0711 03:00:42.111747 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334176\n",
            "I0711 03:00:42.113577 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34682\n",
            "I0711 03:00:45.091919 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335548\n",
            "I0711 03:00:45.093540 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36877\n",
            "I0711 03:00:48.091529 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333377\n",
            "I0711 03:00:48.093124 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33404\n",
            "I0711 03:00:51.068235 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335948\n",
            "I0711 03:00:51.069681 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37517\n",
            "I0711 03:00:54.031270 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337486\n",
            "I0711 03:00:54.033360 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39977\n",
            "I0711 03:00:57.002953 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336514\n",
            "I0711 03:00:57.004503 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38422\n",
            "I0711 03:01:00.007046 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.8926885 (29.846 sec)\n",
            "I0711 03:01:00.014791 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332051\n",
            "I0711 03:01:00.016694 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31281\n",
            "I0711 03:01:02.995928 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33541\n",
            "I0711 03:01:03.002220 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36656\n",
            "I0711 03:01:05.999456 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332938\n",
            "I0711 03:01:06.001415 140338823006080 tpu_estimator.py:2160] examples/sec: 5.327\n",
            "I0711 03:01:08.955424 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338309\n",
            "I0711 03:01:08.957315 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41294\n",
            "I0711 03:01:11.900458 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339544\n",
            "I0711 03:01:11.902105 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4327\n",
            "I0711 03:01:14.794701 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.345517\n",
            "I0711 03:01:14.796542 140338823006080 tpu_estimator.py:2160] examples/sec: 5.52827\n",
            "I0711 03:01:17.752815 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33805\n",
            "I0711 03:01:17.754637 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4088\n",
            "I0711 03:01:20.713029 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337818\n",
            "I0711 03:01:20.714677 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40509\n",
            "I0711 03:01:23.647834 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340737\n",
            "I0711 03:01:23.649600 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45179\n",
            "I0711 03:01:26.534517 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.346417\n",
            "I0711 03:01:26.536308 140338823006080 tpu_estimator.py:2160] examples/sec: 5.54268\n",
            "I0711 03:01:29.499825 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.07774186 (29.493 sec)\n",
            "I0711 03:01:29.508179 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336353\n",
            "I0711 03:01:29.509712 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38164\n",
            "I0711 03:01:32.492351 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335032\n",
            "I0711 03:01:32.493974 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36052\n",
            "I0711 03:01:35.447583 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338382\n",
            "I0711 03:01:35.448768 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4141\n",
            "I0711 03:01:38.385904 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340332\n",
            "I0711 03:01:38.387387 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4453\n",
            "I0711 03:01:41.335288 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339059\n",
            "I0711 03:01:41.336792 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42494\n",
            "I0711 03:01:44.312750 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335864\n",
            "I0711 03:01:44.314365 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37382\n",
            "I0711 03:01:47.273698 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337723\n",
            "I0711 03:01:47.275230 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40356\n",
            "I0711 03:01:50.306520 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329731\n",
            "I0711 03:01:50.307894 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27569\n",
            "I0711 03:01:53.302887 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333728\n",
            "I0711 03:01:53.304258 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33965\n",
            "I0711 03:01:56.260358 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338129\n",
            "I0711 03:01:56.261963 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41006\n",
            "I0711 03:01:59.174414 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.843711 (29.675 sec)\n",
            "I0711 03:01:59.177620 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342793\n",
            "I0711 03:01:59.179511 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48469\n",
            "I0711 03:02:02.141174 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337426\n",
            "I0711 03:02:02.142560 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39881\n",
            "I0711 03:02:05.133033 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33424\n",
            "I0711 03:02:05.134495 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34784\n",
            "I0711 03:02:08.065145 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341053\n",
            "I0711 03:02:08.066737 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45685\n",
            "I0711 03:02:10.996519 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341136\n",
            "I0711 03:02:10.998060 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45818\n",
            "I0711 03:02:13.970149 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336292\n",
            "I0711 03:02:13.971725 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38067\n",
            "I0711 03:02:16.959749 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334486\n",
            "I0711 03:02:16.961746 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35178\n",
            "I0711 03:02:19.948388 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334617\n",
            "I0711 03:02:19.949987 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35388\n",
            "I0711 03:02:22.934317 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334894\n",
            "I0711 03:02:22.935912 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3583\n",
            "I0711 03:02:25.908905 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336177\n",
            "I0711 03:02:25.910693 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37884\n",
            "I0711 03:02:28.879904 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.91652775 (29.705 sec)\n",
            "I0711 03:02:28.887979 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335726\n",
            "I0711 03:02:28.890185 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37162\n",
            "I0711 03:02:31.853357 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337179\n",
            "I0711 03:02:31.854903 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39487\n",
            "I0711 03:02:34.792004 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340288\n",
            "I0711 03:02:34.793715 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44461\n",
            "I0711 03:02:37.773294 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335434\n",
            "I0711 03:02:37.774907 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36694\n",
            "I0711 03:02:40.715132 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339912\n",
            "I0711 03:02:40.716542 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43859\n",
            "I0711 03:02:43.684422 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336785\n",
            "I0711 03:02:43.686040 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38856\n",
            "I0711 03:02:46.640563 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338306\n",
            "I0711 03:02:46.642294 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41289\n",
            "I0711 03:02:49.614875 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336186\n",
            "I0711 03:02:49.616265 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37897\n",
            "I0711 03:02:52.544214 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341374\n",
            "I0711 03:02:52.545859 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46198\n",
            "I0711 03:02:55.500426 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33827\n",
            "I0711 03:02:55.502055 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41231\n",
            "I0711 03:02:58.458573 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.79709625 (29.579 sec)\n",
            "I0711 03:02:58.465683 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337279\n",
            "I0711 03:02:58.469115 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39646\n",
            "I0711 03:03:01.492411 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330352\n",
            "I0711 03:03:01.494032 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28563\n",
            "I0711 03:03:04.415725 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342077\n",
            "I0711 03:03:04.417023 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47322\n",
            "I0711 03:03:07.393955 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335775\n",
            "I0711 03:03:07.395377 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3724\n",
            "I0711 03:03:10.348104 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338503\n",
            "I0711 03:03:10.349871 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41605\n",
            "I0711 03:03:13.300295 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338736\n",
            "I0711 03:03:13.301772 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41977\n",
            "I0711 03:03:16.274741 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33619\n",
            "I0711 03:03:16.276399 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37904\n",
            "I0711 03:03:19.259920 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334991\n",
            "I0711 03:03:19.261318 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35985\n",
            "I0711 03:03:22.270740 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332135\n",
            "I0711 03:03:22.272150 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31415\n",
            "I0711 03:03:25.237037 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337123\n",
            "I0711 03:03:25.238592 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39397\n",
            "I0711 03:03:28.254694 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.92558885 (29.796 sec)\n",
            "I0711 03:03:28.258417 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330978\n",
            "I0711 03:03:28.262791 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29565\n",
            "I0711 03:03:31.199418 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34002\n",
            "I0711 03:03:31.200990 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44031\n",
            "I0711 03:03:34.203101 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33292\n",
            "I0711 03:03:34.204585 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32671\n",
            "I0711 03:03:37.184310 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335438\n",
            "I0711 03:03:37.185690 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36701\n",
            "I0711 03:03:40.166382 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335337\n",
            "I0711 03:03:40.168172 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36539\n",
            "I0711 03:03:43.124743 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338028\n",
            "I0711 03:03:43.126613 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40845\n",
            "I0711 03:03:46.083760 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337946\n",
            "I0711 03:03:46.085381 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40714\n",
            "I0711 03:03:49.088448 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332809\n",
            "I0711 03:03:49.089840 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32494\n",
            "I0711 03:03:52.029571 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340011\n",
            "I0711 03:03:52.031247 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44017\n",
            "I0711 03:03:55.013999 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335067\n",
            "I0711 03:03:55.015706 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36108\n",
            "I0711 03:03:58.068819 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.085197926 (29.814 sec)\n",
            "I0711 03:03:58.076990 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.326486\n",
            "I0711 03:03:58.079328 140338823006080 tpu_estimator.py:2160] examples/sec: 5.22378\n",
            "I0711 03:04:01.052091 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336119\n",
            "I0711 03:04:01.053425 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3779\n",
            "I0711 03:04:04.037503 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334963\n",
            "I0711 03:04:04.039313 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35941\n",
            "I0711 03:04:06.983492 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339441\n",
            "I0711 03:04:06.984808 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43105\n",
            "I0711 03:04:09.932604 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339085\n",
            "I0711 03:04:09.934353 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42537\n",
            "I0711 03:04:12.910953 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335761\n",
            "I0711 03:04:12.912392 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37218\n",
            "I0711 03:04:15.868042 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338167\n",
            "I0711 03:04:15.869671 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41068\n",
            "I0711 03:04:18.862704 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333924\n",
            "I0711 03:04:18.864930 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34278\n",
            "I0711 03:04:21.840389 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335832\n",
            "I0711 03:04:21.841806 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37332\n",
            "I0711 03:04:24.771459 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341203\n",
            "I0711 03:04:24.773305 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45924\n",
            "I0711 03:04:27.716689 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2765456 (29.648 sec)\n",
            "I0711 03:04:27.725584 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338487\n",
            "I0711 03:04:27.726808 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4158\n",
            "I0711 03:04:30.680157 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338453\n",
            "I0711 03:04:30.681779 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41524\n",
            "I0711 03:04:33.646519 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337114\n",
            "I0711 03:04:33.647882 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39382\n",
            "I0711 03:04:36.628132 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335393\n",
            "I0711 03:04:36.629489 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36629\n",
            "I0711 03:04:39.599063 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336591\n",
            "I0711 03:04:39.600481 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38545\n",
            "I0711 03:04:42.559573 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337793\n",
            "I0711 03:04:42.561559 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40469\n",
            "I0711 03:04:45.482369 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342132\n",
            "I0711 03:04:45.484945 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47411\n",
            "I0711 03:04:48.504428 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330898\n",
            "I0711 03:04:48.505787 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29436\n",
            "I0711 03:04:51.454041 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339019\n",
            "I0711 03:04:51.455507 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42431\n",
            "I0711 03:04:54.424417 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33666\n",
            "I0711 03:04:54.425901 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38655\n",
            "I0711 03:04:57.465593 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.5433197 (29.749 sec)\n",
            "I0711 03:04:57.468522 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328504\n",
            "I0711 03:04:57.470328 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25606\n",
            "I0711 03:05:00.425665 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338167\n",
            "I0711 03:05:00.427101 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41067\n",
            "I0711 03:05:03.416784 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33432\n",
            "I0711 03:05:03.418370 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34913\n",
            "I0711 03:05:06.369922 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338626\n",
            "I0711 03:05:06.371291 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41802\n",
            "I0711 03:05:09.351465 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335394\n",
            "I0711 03:05:09.353294 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36631\n",
            "I0711 03:05:12.319293 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336948\n",
            "I0711 03:05:12.321047 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39117\n",
            "I0711 03:05:15.304222 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33502\n",
            "I0711 03:05:15.305912 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36032\n",
            "I0711 03:05:18.271307 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337037\n",
            "I0711 03:05:18.273420 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39258\n",
            "I0711 03:05:21.254889 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335157\n",
            "I0711 03:05:21.256391 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36251\n",
            "I0711 03:05:24.201271 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339406\n",
            "I0711 03:05:24.202714 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4305\n",
            "I0711 03:05:27.169547 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.8156347 (29.704 sec)\n",
            "I0711 03:05:27.172342 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336578\n",
            "I0711 03:05:27.173868 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38525\n",
            "I0711 03:05:30.189399 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331456\n",
            "I0711 03:05:30.190984 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30329\n",
            "I0711 03:05:33.153974 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337308\n",
            "I0711 03:05:33.155410 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39693\n",
            "I0711 03:05:36.172553 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33128\n",
            "I0711 03:05:36.174463 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30048\n",
            "I0711 03:05:39.170747 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333534\n",
            "I0711 03:05:39.172532 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33655\n",
            "I0711 03:05:42.159689 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334565\n",
            "I0711 03:05:42.161201 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35303\n",
            "I0711 03:05:45.137791 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335786\n",
            "I0711 03:05:45.139252 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37258\n",
            "I0711 03:05:48.114575 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335935\n",
            "I0711 03:05:48.115895 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37497\n",
            "I0711 03:05:51.098759 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335095\n",
            "I0711 03:05:51.100137 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36152\n",
            "I0711 03:05:54.061054 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337575\n",
            "I0711 03:05:54.062483 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4012\n",
            "I0711 03:05:57.017879 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.40494633 (29.848 sec)\n",
            "I0711 03:05:57.025109 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337392\n",
            "I0711 03:05:57.027780 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39828\n",
            "I0711 03:05:59.990248 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337247\n",
            "I0711 03:05:59.991811 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39596\n",
            "I0711 03:06:02.952303 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337598\n",
            "I0711 03:06:02.953945 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40157\n",
            "I0711 03:06:05.945061 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334135\n",
            "I0711 03:06:05.946949 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34616\n",
            "I0711 03:06:08.946441 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333185\n",
            "I0711 03:06:08.950889 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33096\n",
            "I0711 03:06:11.929402 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335246\n",
            "I0711 03:06:11.931130 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36393\n",
            "I0711 03:06:14.947849 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331286\n",
            "I0711 03:06:14.949281 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30058\n",
            "I0711 03:06:17.924409 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335952\n",
            "I0711 03:06:17.926108 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37523\n",
            "I0711 03:06:20.928439 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332896\n",
            "I0711 03:06:20.930559 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32633\n",
            "I0711 03:06:23.947017 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331279\n",
            "I0711 03:06:23.948499 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30046\n",
            "I0711 03:06:26.909363 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.25560188 (29.891 sec)\n",
            "I0711 03:06:26.920361 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336334\n",
            "I0711 03:06:26.921725 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38135\n",
            "I0711 03:06:29.901293 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335454\n",
            "I0711 03:06:29.902602 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36726\n",
            "I0711 03:06:32.904536 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33297\n",
            "I0711 03:06:32.905933 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32752\n",
            "I0711 03:06:35.882177 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335838\n",
            "I0711 03:06:35.883842 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37341\n",
            "I0711 03:06:38.855938 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336268\n",
            "I0711 03:06:38.857280 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38029\n",
            "I0711 03:06:41.802214 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33943\n",
            "I0711 03:06:41.804014 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43089\n",
            "I0711 03:06:44.745283 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339769\n",
            "I0711 03:06:44.746883 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4363\n",
            "I0711 03:06:47.692577 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339303\n",
            "I0711 03:06:47.694190 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42886\n",
            "I0711 03:06:50.678421 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334902\n",
            "I0711 03:06:50.679974 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35843\n",
            "I0711 03:06:53.630578 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338739\n",
            "I0711 03:06:53.632090 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41982\n",
            "I0711 03:06:56.581057 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.105863094 (29.672 sec)\n",
            "I0711 03:06:56.589623 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337946\n",
            "I0711 03:06:56.591645 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40714\n",
            "I0711 03:06:59.532972 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339744\n",
            "I0711 03:06:59.534396 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43591\n",
            "I0711 03:07:02.499707 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337071\n",
            "I0711 03:07:02.501386 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39314\n",
            "I0711 03:07:05.481134 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335421\n",
            "I0711 03:07:05.482882 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36674\n",
            "I0711 03:07:08.412779 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341096\n",
            "I0711 03:07:08.414427 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45753\n",
            "I0711 03:07:11.371788 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337957\n",
            "I0711 03:07:11.373283 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40732\n",
            "I0711 03:07:14.290883 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.342565\n",
            "I0711 03:07:14.292763 140338823006080 tpu_estimator.py:2160] examples/sec: 5.48104\n",
            "I0711 03:07:17.252944 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337609\n",
            "I0711 03:07:17.254334 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40174\n",
            "I0711 03:07:20.192672 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340162\n",
            "I0711 03:07:20.193994 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44259\n",
            "I0711 03:07:23.150811 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338053\n",
            "I0711 03:07:23.152300 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40885\n",
            "I0711 03:07:26.074153 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.300098 (29.493 sec)\n",
            "I0711 03:07:26.083693 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340957\n",
            "I0711 03:07:26.085144 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45531\n",
            "I0711 03:07:29.040356 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338218\n",
            "I0711 03:07:29.041778 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41149\n",
            "I0711 03:07:32.105504 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.326255\n",
            "I0711 03:07:32.106796 140338823006080 tpu_estimator.py:2160] examples/sec: 5.22008\n",
            "I0711 03:07:35.059777 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338489\n",
            "I0711 03:07:35.062141 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41582\n",
            "I0711 03:07:38.095665 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329397\n",
            "I0711 03:07:38.097422 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27036\n",
            "I0711 03:07:41.037554 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339919\n",
            "I0711 03:07:41.038985 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43871\n",
            "I0711 03:07:44.004577 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33704\n",
            "I0711 03:07:44.006124 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39264\n",
            "I0711 03:07:46.977015 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336415\n",
            "I0711 03:07:46.978648 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38264\n",
            "I0711 03:07:49.969133 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334219\n",
            "I0711 03:07:49.970511 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3475\n",
            "I0711 03:07:52.923143 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33852\n",
            "I0711 03:07:52.924593 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41633\n",
            "I0711 03:07:55.914363 140338823006080 basic_session_run_hooks.py:260] batch_loss = 3.5757396 (29.840 sec)\n",
            "I0711 03:07:55.922676 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333437\n",
            "I0711 03:07:55.924315 140338823006080 tpu_estimator.py:2160] examples/sec: 5.335\n",
            "I0711 03:07:58.882422 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337829\n",
            "I0711 03:07:58.884056 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40526\n",
            "I0711 03:08:01.808895 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341689\n",
            "I0711 03:08:01.810549 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46703\n",
            "I0711 03:08:04.733735 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341903\n",
            "I0711 03:08:04.735313 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47045\n",
            "I0711 03:08:07.664266 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341234\n",
            "I0711 03:08:07.665619 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45974\n",
            "I0711 03:08:10.575959 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34345\n",
            "I0711 03:08:10.577342 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4952\n",
            "I0711 03:08:13.553636 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335825\n",
            "I0711 03:08:13.555760 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37319\n",
            "I0711 03:08:16.495944 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339866\n",
            "I0711 03:08:16.497835 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43785\n",
            "I0711 03:08:19.446166 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338959\n",
            "I0711 03:08:19.447561 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42334\n",
            "I0711 03:08:22.417950 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3365\n",
            "I0711 03:08:22.420064 140338823006080 tpu_estimator.py:2160] examples/sec: 5.384\n",
            "I0711 03:08:25.400720 140338823006080 basic_session_run_hooks.py:260] batch_loss = 4.0976634 (29.486 sec)\n",
            "I0711 03:08:25.408880 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334497\n",
            "I0711 03:08:25.411005 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35196\n",
            "I0711 03:08:28.373412 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337172\n",
            "I0711 03:08:28.374885 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39475\n",
            "I0711 03:08:31.383574 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332204\n",
            "I0711 03:08:31.384959 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31527\n",
            "I0711 03:08:34.350613 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337031\n",
            "I0711 03:08:34.352221 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39249\n",
            "I0711 03:08:37.307781 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338163\n",
            "I0711 03:08:37.309140 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41062\n",
            "I0711 03:08:40.287983 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335548\n",
            "I0711 03:08:40.289955 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36877\n",
            "I0711 03:08:43.265603 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335838\n",
            "I0711 03:08:43.267286 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3734\n",
            "I0711 03:08:46.247230 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335396\n",
            "I0711 03:08:46.249197 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36633\n",
            "I0711 03:08:49.232612 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334959\n",
            "I0711 03:08:49.234216 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35934\n",
            "I0711 03:08:52.157572 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341882\n",
            "I0711 03:08:52.159242 140338823006080 tpu_estimator.py:2160] examples/sec: 5.47011\n",
            "I0711 03:08:55.107640 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.07228494 (29.707 sec)\n",
            "I0711 03:08:55.116825 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337923\n",
            "I0711 03:08:55.118421 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40677\n",
            "I0711 03:08:58.116169 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333408\n",
            "I0711 03:08:58.117534 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33452\n",
            "I0711 03:09:01.056285 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340132\n",
            "I0711 03:09:01.058089 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44211\n",
            "I0711 03:09:03.969099 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343299\n",
            "I0711 03:09:03.970596 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49279\n",
            "I0711 03:09:06.938423 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336785\n",
            "I0711 03:09:06.939813 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38856\n",
            "I0711 03:09:09.950525 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331986\n",
            "I0711 03:09:09.952028 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31178\n",
            "I0711 03:09:12.934554 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335119\n",
            "I0711 03:09:12.936355 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3619\n",
            "I0711 03:09:15.878749 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339653\n",
            "I0711 03:09:15.880139 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43444\n",
            "I0711 03:09:18.815944 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340472\n",
            "I0711 03:09:18.817675 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44756\n",
            "I0711 03:09:21.832621 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331477\n",
            "I0711 03:09:21.834199 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30363\n",
            "I0711 03:09:24.804851 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.6062155 (29.697 sec)\n",
            "I0711 03:09:24.813003 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335533\n",
            "I0711 03:09:24.820596 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36853\n",
            "I0711 03:09:27.818954 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332668\n",
            "I0711 03:09:27.820570 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32269\n",
            "I0711 03:09:30.778160 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337931\n",
            "I0711 03:09:30.779518 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40689\n",
            "I0711 03:09:33.803738 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330515\n",
            "I0711 03:09:33.805292 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28824\n",
            "I0711 03:09:36.808481 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332805\n",
            "I0711 03:09:36.809842 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32488\n",
            "I0711 03:09:39.792431 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335128\n",
            "I0711 03:09:39.794062 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36205\n",
            "I0711 03:09:42.742651 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338961\n",
            "I0711 03:09:42.744128 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42338\n",
            "I0711 03:09:45.690270 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339271\n",
            "I0711 03:09:45.692093 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42833\n",
            "I0711 03:09:48.734140 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328516\n",
            "I0711 03:09:48.736776 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25626\n",
            "I0711 03:09:51.706207 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336463\n",
            "I0711 03:09:51.707751 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38341\n",
            "I0711 03:09:54.673846 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.7680148 (29.869 sec)\n",
            "I0711 03:09:54.682098 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336041\n",
            "I0711 03:09:54.683928 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37666\n",
            "I0711 03:09:57.646440 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337334\n",
            "I0711 03:09:57.647846 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39735\n",
            "I0711 03:10:00.643598 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333648\n",
            "I0711 03:10:00.645203 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33836\n",
            "I0711 03:10:03.615415 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336495\n",
            "I0711 03:10:03.616913 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38392\n",
            "I0711 03:10:06.605257 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33447\n",
            "I0711 03:10:06.606772 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35152\n",
            "I0711 03:10:09.593943 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334605\n",
            "I0711 03:10:09.595308 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35368\n",
            "I0711 03:10:12.571342 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335853\n",
            "I0711 03:10:12.572901 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37365\n",
            "I0711 03:10:15.582733 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332076\n",
            "I0711 03:10:15.584011 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31321\n",
            "I0711 03:10:18.612908 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330009\n",
            "I0711 03:10:18.614307 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28014\n",
            "I0711 03:10:21.558518 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339487\n",
            "I0711 03:10:21.560122 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4318\n",
            "I0711 03:10:24.607132 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.069965124 (29.933 sec)\n",
            "I0711 03:10:24.614155 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.32728\n",
            "I0711 03:10:24.617161 140338823006080 tpu_estimator.py:2160] examples/sec: 5.23648\n",
            "I0711 03:10:27.582770 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336844\n",
            "I0711 03:10:27.584506 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38951\n",
            "I0711 03:10:30.545828 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337489\n",
            "I0711 03:10:30.547390 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39983\n",
            "I0711 03:10:33.556651 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332146\n",
            "I0711 03:10:33.557971 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31434\n",
            "I0711 03:10:36.566632 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332213\n",
            "I0711 03:10:36.567949 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3154\n",
            "I0711 03:10:39.612172 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.328358\n",
            "I0711 03:10:39.613970 140338823006080 tpu_estimator.py:2160] examples/sec: 5.25373\n",
            "I0711 03:10:42.592885 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33548\n",
            "I0711 03:10:42.594316 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36768\n",
            "I0711 03:10:45.577634 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335038\n",
            "I0711 03:10:45.579100 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36061\n",
            "I0711 03:10:48.612176 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329551\n",
            "I0711 03:10:48.613842 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27282\n",
            "I0711 03:10:51.628587 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331506\n",
            "I0711 03:10:51.630135 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3041\n",
            "I0711 03:10:54.590850 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.105041504 (29.984 sec)\n",
            "I0711 03:10:54.599410 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336607\n",
            "I0711 03:10:54.600858 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38571\n",
            "I0711 03:10:57.590493 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334331\n",
            "I0711 03:10:57.591781 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3493\n",
            "I0711 03:11:00.557247 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33707\n",
            "I0711 03:11:00.558624 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39311\n",
            "I0711 03:11:03.504553 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33929\n",
            "I0711 03:11:03.509583 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42864\n",
            "I0711 03:11:06.491880 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334746\n",
            "I0711 03:11:06.493419 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35594\n",
            "I0711 03:11:09.499941 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332452\n",
            "I0711 03:11:09.501766 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31923\n",
            "I0711 03:11:12.450794 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338879\n",
            "I0711 03:11:12.456864 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42206\n",
            "I0711 03:11:15.430472 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335614\n",
            "I0711 03:11:15.432421 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36983\n",
            "I0711 03:11:18.454273 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3307\n",
            "I0711 03:11:18.455971 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2912\n",
            "I0711 03:11:21.455384 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33321\n",
            "I0711 03:11:21.461297 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33137\n",
            "I0711 03:11:24.455286 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.25937068 (29.864 sec)\n",
            "I0711 03:11:24.457883 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333055\n",
            "I0711 03:11:24.460563 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32888\n",
            "I0711 03:11:27.412325 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338479\n",
            "I0711 03:11:27.413948 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41566\n",
            "I0711 03:11:30.393969 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335389\n",
            "I0711 03:11:30.395457 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36622\n",
            "I0711 03:11:33.336343 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339862\n",
            "I0711 03:11:33.338124 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43779\n",
            "I0711 03:11:36.359598 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330762\n",
            "I0711 03:11:36.361170 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29219\n",
            "I0711 03:11:39.389836 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330009\n",
            "I0711 03:11:39.390989 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28015\n",
            "I0711 03:11:42.366489 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335956\n",
            "I0711 03:11:42.368113 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37529\n",
            "I0711 03:11:45.332223 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337175\n",
            "I0711 03:11:45.334051 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3948\n",
            "I0711 03:11:48.319709 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334724\n",
            "I0711 03:11:48.321316 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35559\n",
            "I0711 03:11:51.311977 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334202\n",
            "I0711 03:11:51.313878 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34723\n",
            "I0711 03:11:54.300214 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.38675714 (29.845 sec)\n",
            "I0711 03:11:54.307569 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33384\n",
            "I0711 03:11:54.311349 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34144\n",
            "I0711 03:11:57.292907 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33495\n",
            "I0711 03:11:57.294331 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3592\n",
            "I0711 03:12:00.245668 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338669\n",
            "I0711 03:12:00.247112 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4187\n",
            "I0711 03:12:03.223352 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335829\n",
            "I0711 03:12:03.224958 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37327\n",
            "I0711 03:12:06.184960 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337661\n",
            "I0711 03:12:06.187173 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40258\n",
            "I0711 03:12:09.147112 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337586\n",
            "I0711 03:12:09.148631 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40137\n",
            "I0711 03:12:12.121715 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336185\n",
            "I0711 03:12:12.123512 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37896\n",
            "I0711 03:12:15.083267 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337663\n",
            "I0711 03:12:15.084627 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40261\n",
            "I0711 03:12:18.017252 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340829\n",
            "I0711 03:12:18.018818 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45326\n",
            "I0711 03:12:21.003164 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334905\n",
            "I0711 03:12:21.004660 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35848\n",
            "I0711 03:12:23.984215 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.8984971 (29.684 sec)\n",
            "I0711 03:12:23.990471 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334773\n",
            "I0711 03:12:23.992681 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35637\n",
            "I0711 03:12:26.992202 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333115\n",
            "I0711 03:12:26.994186 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32984\n",
            "I0711 03:12:29.977051 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335026\n",
            "I0711 03:12:29.978715 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36041\n",
            "I0711 03:12:32.970015 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334118\n",
            "I0711 03:12:32.971645 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34589\n",
            "I0711 03:12:35.965390 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333849\n",
            "I0711 03:12:35.966905 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34158\n",
            "I0711 03:12:38.933070 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336966\n",
            "I0711 03:12:38.934893 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39145\n",
            "I0711 03:12:41.879747 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339363\n",
            "I0711 03:12:41.881509 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4298\n",
            "I0711 03:12:44.853313 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336303\n",
            "I0711 03:12:44.855998 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38085\n",
            "I0711 03:12:47.835001 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335373\n",
            "I0711 03:12:47.836470 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36597\n",
            "I0711 03:12:50.817967 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335243\n",
            "I0711 03:12:50.819507 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36389\n",
            "I0711 03:12:53.796536 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.7751093 (29.812 sec)\n",
            "I0711 03:12:53.799418 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335396\n",
            "I0711 03:12:53.805619 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36634\n",
            "I0711 03:12:56.834310 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329507\n",
            "I0711 03:12:56.835950 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27211\n",
            "I0711 03:12:59.808113 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336271\n",
            "I0711 03:12:59.809430 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38033\n",
            "I0711 03:13:02.822057 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331791\n",
            "I0711 03:13:02.823412 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30866\n",
            "I0711 03:13:05.813489 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334292\n",
            "I0711 03:13:05.814877 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34867\n",
            "I0711 03:13:08.770099 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338221\n",
            "I0711 03:13:08.772080 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41154\n",
            "I0711 03:13:11.715379 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339522\n",
            "I0711 03:13:11.716934 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43235\n",
            "I0711 03:13:14.701444 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334894\n",
            "I0711 03:13:14.702739 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35831\n",
            "I0711 03:13:17.714898 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331843\n",
            "I0711 03:13:17.716613 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30949\n",
            "I0711 03:13:20.715969 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333215\n",
            "I0711 03:13:20.717646 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33145\n",
            "I0711 03:13:23.727482 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.8295705 (29.931 sec)\n",
            "I0711 03:13:23.730441 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331733\n",
            "I0711 03:13:23.735968 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30773\n",
            "I0711 03:13:26.745562 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33166\n",
            "I0711 03:13:26.746910 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30657\n",
            "I0711 03:13:29.778583 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329706\n",
            "I0711 03:13:29.780057 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2753\n",
            "I0711 03:13:32.768347 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334482\n",
            "I0711 03:13:32.769945 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35172\n",
            "I0711 03:13:35.726981 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33798\n",
            "I0711 03:13:35.728671 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40767\n",
            "I0711 03:13:38.717610 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334383\n",
            "I0711 03:13:38.719260 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35013\n",
            "I0711 03:13:41.727318 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332259\n",
            "I0711 03:13:41.728847 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31614\n",
            "I0711 03:13:44.692754 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337215\n",
            "I0711 03:13:44.694022 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39544\n",
            "I0711 03:13:47.679605 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334803\n",
            "I0711 03:13:47.681100 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35685\n",
            "I0711 03:13:50.703550 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330693\n",
            "I0711 03:13:50.705478 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29108\n",
            "I0711 03:13:53.655741 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.566069 (29.928 sec)\n",
            "I0711 03:13:53.658242 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338449\n",
            "I0711 03:13:53.664017 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41518\n",
            "I0711 03:13:56.590446 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341041\n",
            "I0711 03:13:56.592334 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45666\n",
            "I0711 03:13:59.580577 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334431\n",
            "I0711 03:13:59.582326 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35089\n",
            "I0711 03:14:02.510995 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.341244\n",
            "I0711 03:14:02.512686 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4599\n",
            "I0711 03:14:05.504445 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334065\n",
            "I0711 03:14:05.506161 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34503\n",
            "I0711 03:14:08.511984 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332496\n",
            "I0711 03:14:08.513802 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31994\n",
            "I0711 03:14:11.527971 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331567\n",
            "I0711 03:14:11.529318 140338823006080 tpu_estimator.py:2160] examples/sec: 5.30508\n",
            "I0711 03:14:14.478657 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338908\n",
            "I0711 03:14:14.479968 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42253\n",
            "I0711 03:14:17.507298 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330182\n",
            "I0711 03:14:17.508694 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28292\n",
            "I0711 03:14:20.453514 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339416\n",
            "I0711 03:14:20.455394 140338823006080 tpu_estimator.py:2160] examples/sec: 5.43065\n",
            "I0711 03:14:23.425933 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.4938734 (29.770 sec)\n",
            "I0711 03:14:23.428414 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336142\n",
            "I0711 03:14:23.430574 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37827\n",
            "I0711 03:14:26.396872 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336879\n",
            "I0711 03:14:26.398334 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39007\n",
            "I0711 03:14:29.366663 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336718\n",
            "I0711 03:14:29.368736 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38749\n",
            "I0711 03:14:32.313846 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339313\n",
            "I0711 03:14:32.315285 140338823006080 tpu_estimator.py:2160] examples/sec: 5.429\n",
            "I0711 03:14:35.272917 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337943\n",
            "I0711 03:14:35.274330 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40709\n",
            "I0711 03:14:38.246522 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336295\n",
            "I0711 03:14:38.247984 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38071\n",
            "I0711 03:14:41.218516 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336467\n",
            "I0711 03:14:41.220120 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38348\n",
            "I0711 03:14:44.204424 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334908\n",
            "I0711 03:14:44.205828 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35853\n",
            "I0711 03:14:47.199045 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333933\n",
            "I0711 03:14:47.200608 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34293\n",
            "I0711 03:14:50.151572 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338693\n",
            "I0711 03:14:50.153193 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41909\n",
            "I0711 03:14:53.084125 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.037646294 (29.658 sec)\n",
            "I0711 03:14:53.091042 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340198\n",
            "I0711 03:14:53.093843 140338823006080 tpu_estimator.py:2160] examples/sec: 5.44318\n",
            "I0711 03:14:56.027667 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340537\n",
            "I0711 03:14:56.029948 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4486\n",
            "I0711 03:14:59.006552 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335684\n",
            "I0711 03:14:59.008262 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37094\n",
            "I0711 03:15:01.983981 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335861\n",
            "I0711 03:15:01.985555 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37377\n",
            "I0711 03:15:04.962417 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335751\n",
            "I0711 03:15:04.964107 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37202\n",
            "I0711 03:15:07.928125 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337183\n",
            "I0711 03:15:07.929871 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39493\n",
            "I0711 03:15:10.894894 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337066\n",
            "I0711 03:15:10.896420 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39306\n",
            "I0711 03:15:13.891871 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333676\n",
            "I0711 03:15:13.893773 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33881\n",
            "I0711 03:15:16.853097 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337698\n",
            "I0711 03:15:16.854513 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40317\n",
            "I0711 03:15:19.821476 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336889\n",
            "I0711 03:15:19.823369 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39023\n",
            "I0711 03:15:22.796721 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.8985661 (29.713 sec)\n",
            "I0711 03:15:22.806447 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335036\n",
            "I0711 03:15:22.808280 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36057\n",
            "I0711 03:15:25.739223 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340942\n",
            "I0711 03:15:25.741290 140338823006080 tpu_estimator.py:2160] examples/sec: 5.45507\n",
            "I0711 03:15:28.722166 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335241\n",
            "I0711 03:15:28.723683 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36386\n",
            "I0711 03:15:31.701583 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335629\n",
            "I0711 03:15:31.703193 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37006\n",
            "I0711 03:15:34.707101 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332726\n",
            "I0711 03:15:34.708670 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32361\n",
            "I0711 03:15:37.703014 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333788\n",
            "I0711 03:15:37.704466 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34061\n",
            "I0711 03:15:40.658189 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338397\n",
            "I0711 03:15:40.659617 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41436\n",
            "I0711 03:15:43.611028 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338647\n",
            "I0711 03:15:43.612560 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41835\n",
            "I0711 03:15:46.636367 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330542\n",
            "I0711 03:15:46.637749 140338823006080 tpu_estimator.py:2160] examples/sec: 5.28867\n",
            "I0711 03:15:49.593293 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338194\n",
            "I0711 03:15:49.594693 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41111\n",
            "I0711 03:15:52.574178 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.3691678 (29.777 sec)\n",
            "I0711 03:15:52.583389 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334455\n",
            "I0711 03:15:52.585022 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35128\n",
            "I0711 03:15:55.567165 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335127\n",
            "I0711 03:15:55.568566 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36203\n",
            "I0711 03:15:58.551771 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335048\n",
            "I0711 03:15:58.553465 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36077\n",
            "I0711 03:16:01.521395 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336742\n",
            "I0711 03:16:01.523237 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38787\n",
            "I0711 03:16:04.512181 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334364\n",
            "I0711 03:16:04.513694 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34982\n",
            "I0711 03:16:07.548973 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3293\n",
            "I0711 03:16:07.551852 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2688\n",
            "I0711 03:16:10.534837 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334901\n",
            "I0711 03:16:10.536360 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35841\n",
            "I0711 03:16:13.519651 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335032\n",
            "I0711 03:16:13.521114 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36051\n",
            "I0711 03:16:16.525336 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332706\n",
            "I0711 03:16:16.527097 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3233\n",
            "I0711 03:16:19.513914 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334603\n",
            "I0711 03:16:19.515343 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35365\n",
            "I0711 03:16:22.511405 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.18419743 (29.937 sec)\n",
            "I0711 03:16:22.514546 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333265\n",
            "I0711 03:16:22.517222 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33225\n",
            "I0711 03:16:25.455658 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.340006\n",
            "I0711 03:16:25.457187 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4401\n",
            "I0711 03:16:28.443714 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334664\n",
            "I0711 03:16:28.445294 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35462\n",
            "I0711 03:16:31.437496 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334021\n",
            "I0711 03:16:31.440222 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34433\n",
            "I0711 03:16:34.423115 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33495\n",
            "I0711 03:16:34.424832 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3592\n",
            "I0711 03:16:37.569576 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.317816\n",
            "I0711 03:16:37.570999 140338823006080 tpu_estimator.py:2160] examples/sec: 5.08505\n",
            "I0711 03:16:40.530550 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337726\n",
            "I0711 03:16:40.532994 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40362\n",
            "I0711 03:16:43.535729 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332756\n",
            "I0711 03:16:43.537255 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3241\n",
            "I0711 03:16:46.527063 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334301\n",
            "I0711 03:16:46.528710 140338823006080 tpu_estimator.py:2160] examples/sec: 5.34882\n",
            "I0711 03:16:49.559116 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329807\n",
            "I0711 03:16:49.560468 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27691\n",
            "I0711 03:16:52.583895 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.2517664 (30.072 sec)\n",
            "I0711 03:16:52.594200 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.329484\n",
            "I0711 03:16:52.596129 140338823006080 tpu_estimator.py:2160] examples/sec: 5.27175\n",
            "I0711 03:16:55.580147 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.3349\n",
            "I0711 03:16:55.581568 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35841\n",
            "I0711 03:16:58.565056 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335018\n",
            "I0711 03:16:58.566663 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36029\n",
            "I0711 03:17:01.554093 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334552\n",
            "I0711 03:17:01.560458 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35284\n",
            "I0711 03:17:04.507586 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33859\n",
            "I0711 03:17:04.509134 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41743\n",
            "I0711 03:17:07.576772 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.325814\n",
            "I0711 03:17:07.578128 140338823006080 tpu_estimator.py:2160] examples/sec: 5.21302\n",
            "I0711 03:17:10.548772 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336484\n",
            "I0711 03:17:10.550097 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38374\n",
            "I0711 03:17:13.529682 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335461\n",
            "I0711 03:17:13.531102 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36737\n",
            "I0711 03:17:16.539923 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332196\n",
            "I0711 03:17:16.541506 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31513\n",
            "I0711 03:17:19.519257 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335648\n",
            "I0711 03:17:19.520794 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37036\n",
            "I0711 03:17:22.485334 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.314096 (29.901 sec)\n",
            "I0711 03:17:22.493769 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336202\n",
            "I0711 03:17:22.497435 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37924\n",
            "I0711 03:17:25.467484 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33627\n",
            "I0711 03:17:25.469207 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38032\n",
            "I0711 03:17:28.457287 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334469\n",
            "I0711 03:17:28.458581 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3515\n",
            "I0711 03:17:31.431140 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336262\n",
            "I0711 03:17:31.432455 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38019\n",
            "I0711 03:17:34.416257 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334996\n",
            "I0711 03:17:34.417681 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35993\n",
            "I0711 03:17:37.381510 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337234\n",
            "I0711 03:17:37.383111 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39574\n",
            "I0711 03:17:40.386290 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332809\n",
            "I0711 03:17:40.387599 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32494\n",
            "I0711 03:17:43.365208 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335694\n",
            "I0711 03:17:43.366940 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3711\n",
            "I0711 03:17:46.337704 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336411\n",
            "I0711 03:17:46.339036 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38257\n",
            "I0711 03:17:49.301528 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337415\n",
            "I0711 03:17:49.303104 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39864\n",
            "I0711 03:17:52.265613 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.04476595 (29.780 sec)\n",
            "I0711 03:17:52.270299 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336831\n",
            "I0711 03:17:52.272173 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38929\n",
            "I0711 03:17:55.259921 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334492\n",
            "I0711 03:17:55.261389 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35187\n",
            "I0711 03:17:58.319694 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.326817\n",
            "I0711 03:17:58.321242 140338823006080 tpu_estimator.py:2160] examples/sec: 5.22907\n",
            "I0711 03:18:01.329030 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332305\n",
            "I0711 03:18:01.330318 140338823006080 tpu_estimator.py:2160] examples/sec: 5.31687\n",
            "I0711 03:18:04.306442 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335857\n",
            "I0711 03:18:04.308188 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37372\n",
            "I0711 03:18:07.304812 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333519\n",
            "I0711 03:18:07.306185 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3363\n",
            "I0711 03:18:10.287904 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335219\n",
            "I0711 03:18:10.289321 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3635\n",
            "I0711 03:18:13.316048 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330238\n",
            "I0711 03:18:13.317718 140338823006080 tpu_estimator.py:2160] examples/sec: 5.2838\n",
            "I0711 03:18:16.366822 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.327791\n",
            "I0711 03:18:16.368130 140338823006080 tpu_estimator.py:2160] examples/sec: 5.24465\n",
            "I0711 03:18:19.323253 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33824\n",
            "I0711 03:18:19.324664 140338823006080 tpu_estimator.py:2160] examples/sec: 5.41184\n",
            "I0711 03:18:22.335105 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.381309 (30.069 sec)\n",
            "I0711 03:18:22.345824 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330841\n",
            "I0711 03:18:22.347440 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29345\n",
            "I0711 03:18:25.329586 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.33515\n",
            "I0711 03:18:25.331043 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3624\n",
            "I0711 03:18:28.290328 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.337753\n",
            "I0711 03:18:28.291643 140338823006080 tpu_estimator.py:2160] examples/sec: 5.40404\n",
            "I0711 03:18:31.265349 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336147\n",
            "I0711 03:18:31.266917 140338823006080 tpu_estimator.py:2160] examples/sec: 5.37835\n",
            "I0711 03:18:34.284887 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331161\n",
            "I0711 03:18:34.286844 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29857\n",
            "I0711 03:18:37.270002 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335\n",
            "I0711 03:18:37.271519 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35999\n",
            "I0711 03:18:40.292736 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.330824\n",
            "I0711 03:18:40.294570 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29318\n",
            "I0711 03:18:43.206830 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.343166\n",
            "I0711 03:18:43.208694 140338823006080 tpu_estimator.py:2160] examples/sec: 5.49066\n",
            "I0711 03:18:46.192547 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.334923\n",
            "I0711 03:18:46.194552 140338823006080 tpu_estimator.py:2160] examples/sec: 5.35877\n",
            "I0711 03:18:49.198320 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.332697\n",
            "I0711 03:18:49.199689 140338823006080 tpu_estimator.py:2160] examples/sec: 5.32316\n",
            "I0711 03:18:52.208191 140338823006080 basic_session_run_hooks.py:260] batch_loss = 1.0613494 (29.873 sec)\n",
            "I0711 03:18:52.217470 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.331221\n",
            "I0711 03:18:52.218971 140338823006080 tpu_estimator.py:2160] examples/sec: 5.29953\n",
            "I0711 03:18:55.201159 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335154\n",
            "I0711 03:18:55.202646 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36247\n",
            "I0711 03:18:58.159672 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338007\n",
            "I0711 03:18:58.161326 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4081\n",
            "I0711 03:19:01.108038 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.339165\n",
            "I0711 03:19:01.109846 140338823006080 tpu_estimator.py:2160] examples/sec: 5.42664\n",
            "I0711 03:19:04.106284 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.333535\n",
            "I0711 03:19:04.107910 140338823006080 tpu_estimator.py:2160] examples/sec: 5.33656\n",
            "I0711 03:19:07.074292 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336924\n",
            "I0711 03:19:07.076268 140338823006080 tpu_estimator.py:2160] examples/sec: 5.39078\n",
            "I0711 03:19:10.000149 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.34178\n",
            "I0711 03:19:10.001550 140338823006080 tpu_estimator.py:2160] examples/sec: 5.46848\n",
            "I0711 03:19:12.957495 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.338138\n",
            "I0711 03:19:12.958983 140338823006080 tpu_estimator.py:2160] examples/sec: 5.4102\n",
            "I0711 03:19:15.927696 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336678\n",
            "I0711 03:19:15.929298 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38685\n",
            "I0711 03:19:18.910686 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335237\n",
            "I0711 03:19:18.912168 140338823006080 tpu_estimator.py:2160] examples/sec: 5.3638\n",
            "I0711 03:19:21.876692 140338823006080 basic_session_run_hooks.py:260] batch_loss = 0.62055564 (29.668 sec)\n",
            "I0711 03:19:21.879449 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.336841\n",
            "I0711 03:19:21.887832 140338823006080 tpu_estimator.py:2160] examples/sec: 5.38945\n",
            "I0711 03:19:24.860127 140338823006080 tpu_estimator.py:2159] global_step/sec: 0.335493\n",
            "I0711 03:19:24.861660 140338823006080 tpu_estimator.py:2160] examples/sec: 5.36788\n",
            "I0711 03:19:24.875245 140338823006080 basic_session_run_hooks.py:606] Saving checkpoints for 2632 into /content/drive/My Drive/ner_checkpoints/model.ckpt.\n",
            "I0711 03:19:55.600639 140338823006080 estimator.py:1145] Calling model_fn.\n",
            "I0711 03:19:55.602169 140338823006080 tpu_estimator.py:2965] Running eval on CPU\n",
            "I0711 03:19:55.605373 140338823006080 <ipython-input-9-2f5db7366c31>:282] *** Features ***\n",
            "I0711 03:19:55.609315 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_ids, shape = (?, 128)\n",
            "I0711 03:19:55.611574 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_mask, shape = (?, 128)\n",
            "I0711 03:19:55.613608 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = label_ids, shape = (?, 128)\n",
            "I0711 03:19:55.615765 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = segment_ids, shape = (?, 128)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "shape of input_ids (?, 128)\n",
            "shape of label_ids (?, 128)\n",
            "seq_length 128\n",
            "lengths Tensor(\"Sum:0\", shape=(?,), dtype=int32)\n",
            "####################\n",
            "shape of output_layer: (?, 128, 1024)\n",
            "embedding_size:1024\n",
            "seq_length:128\n",
            "shape of logit (?, 128, 11)\n",
            "shape of loss ()\n",
            "shape of per_example_loss (?,)\n",
            "num labels:11\n",
            "####################\n",
            "shape of pred_ids (?, 128)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "I0711 03:20:03.212475 140338823006080 estimator.py:1147] Done calling model_fn.\n",
            "I0711 03:20:03.243102 140338823006080 evaluation.py:255] Starting evaluation at 2019-07-11T03:20:03Z\n",
            "I0711 03:20:04.471555 140338823006080 monitored_session.py:240] Graph was finalized.\n",
            "I0711 03:20:04.530614 140338823006080 saver.py:1280] Restoring parameters from /content/drive/My Drive/ner_checkpoints/model.ckpt-2632\n",
            "I0711 03:20:14.245001 140338823006080 session_manager.py:500] Running local_init_op.\n",
            "I0711 03:20:14.355391 140338823006080 session_manager.py:502] Done running local_init_op.\n",
            "I0711 03:20:21.430558 140338823006080 evaluation.py:167] Evaluation [10/100]\n",
            "I0711 03:20:26.674819 140338823006080 evaluation.py:167] Evaluation [20/100]\n",
            "I0711 03:20:31.855310 140338823006080 evaluation.py:167] Evaluation [30/100]\n",
            "I0711 03:20:37.046827 140338823006080 evaluation.py:167] Evaluation [40/100]\n",
            "I0711 03:20:42.471990 140338823006080 evaluation.py:167] Evaluation [50/100]\n",
            "I0711 03:20:47.338465 140338823006080 evaluation.py:167] Evaluation [60/100]\n",
            "I0711 03:20:52.206989 140338823006080 evaluation.py:167] Evaluation [70/100]\n",
            "I0711 03:20:57.509884 140338823006080 evaluation.py:167] Evaluation [80/100]\n",
            "I0711 03:21:02.818662 140338823006080 evaluation.py:167] Evaluation [90/100]\n",
            "I0711 03:21:08.496021 140338823006080 evaluation.py:167] Evaluation [100/100]\n",
            "I0711 03:21:08.722022 140338823006080 evaluation.py:275] Finished evaluation at 2019-07-11-03:21:08\n",
            "I0711 03:21:08.723363 140338823006080 estimator.py:2039] Saving dict for global step 2632: eval_accuracy = 0.98800147, eval_f = 0.9499271, eval_loss = 1.1111144, eval_precision = 0.9531707, eval_recall = 0.9467054, global_step = 2632, loss = 1.1111144\n",
            "I0711 03:21:08.731120 140338823006080 estimator.py:2099] Saving 'checkpoint_path' summary for global step 2632: /content/drive/My Drive/ner_checkpoints/model.ckpt-2632\n",
            "I0711 03:21:08.735283 140338823006080 error_handling.py:96] evaluation_loop marked as finished\n",
            "I0711 03:21:09.147538 140338823006080 estimator.py:368] Loss for final step: 1.349946.\n",
            "I0711 03:21:09.149876 140338823006080 error_handling.py:96] training_loop marked as finished\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "***** Eval results *****\n",
            "  eval_accuracy = 0.98800147\n",
            "  eval_f = 0.9499271\n",
            "  eval_loss = 1.1111144\n",
            "  eval_precision = 0.9531707\n",
            "  eval_recall = 0.9467054\n",
            "  global_step = 2632\n",
            "  loss = 1.1111144\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "I0711 03:21:10.428016 140338823006080 <ipython-input-9-2f5db7366c31>:113] Writing example 0 of 3453\n",
            "I0711 03:21:10.434016 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 03:21:10.437298 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: test-0\n",
            "I0711 03:21:10.441936 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: S ##OC ##CE ##R - J ##AP ##AN GE ##T L ##UC ##K ##Y W ##IN , CH ##IN ##A IN S ##UR ##PR ##IS ##E DE ##F ##EA ##T .\n",
            "I0711 03:21:10.444185 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 156 9244 10954 2069 118 147 12240 14962 25075 1942 149 21986 2428 3663 160 11607 117 24890 11607 1592 15969 156 19556 22861 6258 2036 18581 2271 12420 1942 119 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.446401 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.448364 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.450721 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 1 10 10 10 1 6 10 10 1 10 1 10 10 10 1 10 1 2 10 10 1 1 10 10 10 10 1 10 10 10 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.453655 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 03:21:10.455363 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: test-1\n",
            "I0711 03:21:10.457158 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: Na ##di ##m La ##d ##ki\n",
            "I0711 03:21:10.458606 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 11896 3309 1306 2001 1181 2293 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.460340 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.462485 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.464349 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 2 10 10 3 10 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.467710 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 03:21:10.469881 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: test-2\n",
            "I0711 03:21:10.472161 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: AL - AI ##N , United Arab Emirates 1996 - 12 - 06\n",
            "I0711 03:21:10.474159 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 18589 118 19016 2249 117 1244 4699 14832 1820 118 1367 118 5037 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.476191 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.478046 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.479893 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 6 10 10 10 1 6 7 7 1 10 10 10 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.483685 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 03:21:10.485580 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: test-3\n",
            "I0711 03:21:10.487943 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: Japan began the defence of their Asian Cup title with a lucky 2 - 1 win against Syria in a Group C championship match on Friday .\n",
            "I0711 03:21:10.489865 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 1999 1310 1103 6465 1104 1147 3141 1635 1641 1114 170 6918 123 118 122 1782 1222 7303 1107 170 1990 140 2899 1801 1113 5286 119 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.491715 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.493640 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.495529 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 6 1 1 1 1 1 8 9 1 1 1 1 1 10 10 1 1 6 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.499477 140338823006080 <ipython-input-9-2f5db7366c31>:69] *** Example ***\n",
            "I0711 03:21:10.501673 140338823006080 <ipython-input-9-2f5db7366c31>:70] guid: test-4\n",
            "I0711 03:21:10.504130 140338823006080 <ipython-input-9-2f5db7366c31>:72] tokens: But China saw their luck desert them in the second match of the group , crashing to a surprise 2 - 0 defeat to newcomer ##s Uzbekistan .\n",
            "I0711 03:21:10.505671 140338823006080 <ipython-input-9-2f5db7366c31>:73] input_ids: 101 1252 1975 1486 1147 6920 6941 1172 1107 1103 1248 1801 1104 1103 1372 117 13423 1106 170 3774 123 118 121 3326 1106 25551 1116 18245 119 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.507782 140338823006080 <ipython-input-9-2f5db7366c31>:74] input_mask: 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.510178 140338823006080 <ipython-input-9-2f5db7366c31>:75] segment_ids: 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:10.511646 140338823006080 <ipython-input-9-2f5db7366c31>:76] label_ids: 0 1 6 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 10 10 1 1 1 10 6 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 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",
            "I0711 03:21:12.923662 140338823006080 estimator.py:1145] Calling model_fn.\n",
            "I0711 03:21:12.925157 140338823006080 tpu_estimator.py:2965] Running eval on CPU\n",
            "I0711 03:21:12.926908 140338823006080 <ipython-input-9-2f5db7366c31>:282] *** Features ***\n",
            "I0711 03:21:12.933950 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_ids, shape = (?, 128)\n",
            "I0711 03:21:12.936290 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_mask, shape = (?, 128)\n",
            "I0711 03:21:12.939271 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = label_ids, shape = (?, 128)\n",
            "I0711 03:21:12.941752 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = segment_ids, shape = (?, 128)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "***** Running prediction*****\n",
            "  Num examples = 3453\n",
            "  Batch size = 8\n",
            "shape of input_ids (?, 128)\n",
            "shape of label_ids (?, 128)\n",
            "seq_length 128\n",
            "lengths Tensor(\"Sum:0\", shape=(?,), dtype=int32)\n",
            "####################\n",
            "shape of output_layer: (?, 128, 1024)\n",
            "embedding_size:1024\n",
            "seq_length:128\n",
            "shape of logit (?, 128, 11)\n",
            "shape of loss ()\n",
            "shape of per_example_loss (?,)\n",
            "num labels:11\n",
            "####################\n",
            "shape of pred_ids (?, 128)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "I0711 03:21:19.560024 140338823006080 estimator.py:1147] Done calling model_fn.\n",
            "I0711 03:21:19.590540 140338823006080 evaluation.py:255] Starting evaluation at 2019-07-11T03:21:19Z\n",
            "I0711 03:21:21.562604 140338823006080 monitored_session.py:240] Graph was finalized.\n",
            "I0711 03:21:21.578215 140338823006080 saver.py:1280] Restoring parameters from /content/drive/My Drive/ner_checkpoints/model.ckpt-2632\n",
            "I0711 03:21:24.888319 140338823006080 session_manager.py:500] Running local_init_op.\n",
            "I0711 03:21:24.994400 140338823006080 session_manager.py:502] Done running local_init_op.\n",
            "I0711 03:25:12.208883 140338823006080 evaluation.py:275] Finished evaluation at 2019-07-11-03:25:12\n",
            "I0711 03:25:12.210873 140338823006080 estimator.py:2039] Saving dict for global step 2632: eval_accuracy = 0.9813488, eval_f = 0.907698, eval_loss = 1.800308, eval_precision = 0.901839, eval_recall = 0.9136336, global_step = 2632, loss = 1.8011435\n",
            "I0711 03:25:12.215716 140338823006080 estimator.py:2099] Saving 'checkpoint_path' summary for global step 2632: /content/drive/My Drive/ner_checkpoints/model.ckpt-2632\n",
            "I0711 03:25:12.222175 140338823006080 error_handling.py:96] evaluation_loop marked as finished\n",
            "I0711 03:25:12.232628 140338823006080 <ipython-input-10-16741e56f63c>:181] ***** Predict results *****\n",
            "I0711 03:25:12.234656 140338823006080 <ipython-input-10-16741e56f63c>:183]   eval_accuracy = 0.9813488\n",
            "I0711 03:25:12.237040 140338823006080 <ipython-input-10-16741e56f63c>:183]   eval_f = 0.907698\n",
            "I0711 03:25:12.238562 140338823006080 <ipython-input-10-16741e56f63c>:183]   eval_loss = 1.800308\n",
            "I0711 03:25:12.240889 140338823006080 <ipython-input-10-16741e56f63c>:183]   eval_precision = 0.901839\n",
            "I0711 03:25:12.242280 140338823006080 <ipython-input-10-16741e56f63c>:183]   eval_recall = 0.9136336\n",
            "I0711 03:25:12.243930 140338823006080 <ipython-input-10-16741e56f63c>:183]   global_step = 2632\n",
            "I0711 03:25:12.245979 140338823006080 <ipython-input-10-16741e56f63c>:183]   loss = 1.8011435\n",
            "I0711 03:25:12.329592 140338823006080 estimator.py:1145] Calling model_fn.\n",
            "I0711 03:25:12.330657 140338823006080 tpu_estimator.py:2965] Running infer on CPU\n",
            "I0711 03:25:12.332222 140338823006080 <ipython-input-9-2f5db7366c31>:282] *** Features ***\n",
            "I0711 03:25:12.334113 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_ids, shape = (?, 128)\n",
            "I0711 03:25:12.336304 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = input_mask, shape = (?, 128)\n",
            "I0711 03:25:12.338193 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = label_ids, shape = (?, 128)\n",
            "I0711 03:25:12.340410 140338823006080 <ipython-input-9-2f5db7366c31>:284]   name = segment_ids, shape = (?, 128)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "********************\n",
            "type of result:<class 'generator'>, type of predict_examples:<class 'list'>\n",
            "********************\n",
            "shape of input_ids (?, 128)\n",
            "shape of label_ids (?, 128)\n",
            "seq_length 128\n",
            "lengths Tensor(\"Sum:0\", shape=(?,), dtype=int32)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "I0711 03:25:18.422541 140338823006080 estimator.py:1147] Done calling model_fn.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "####################\n",
            "shape of output_layer: (?, 128, 1024)\n",
            "embedding_size:1024\n",
            "seq_length:128\n",
            "shape of logit (?, 128, 11)\n",
            "shape of loss ()\n",
            "shape of per_example_loss (?,)\n",
            "num labels:11\n",
            "####################\n",
            "shape of pred_ids (?, 128)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "I0711 03:25:19.623170 140338823006080 monitored_session.py:240] Graph was finalized.\n",
            "I0711 03:25:19.632440 140338823006080 saver.py:1280] Restoring parameters from /content/drive/My Drive/ner_checkpoints/model.ckpt-2632\n",
            "I0711 03:25:22.924409 140338823006080 session_manager.py:500] Running local_init_op.\n",
            "I0711 03:25:23.012550 140338823006080 session_manager.py:502] Done running local_init_op.\n",
            "I0711 03:28:59.760147 140338823006080 error_handling.py:96] prediction_loop marked as finished\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "error",
          "ename": "SystemExit",
          "evalue": "ignored",
          "traceback": [
            "An exception has occurred, use %tb to see the full traceback.\n",
            "\u001b[0;31mSystemExit\u001b[0m\n"
          ]
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/IPython/core/interactiveshell.py:2890: UserWarning: To exit: use 'exit', 'quit', or Ctrl-D.\n",
            "  warn(\"To exit: use 'exit', 'quit', or Ctrl-D.\", stacklevel=1)\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Sx6ZrFKxnc9K",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!ps ax | grep python\n",
        "\n",
        "print('x1231)\n",
        "print('x567')\n",
        "print('x890')\n",
        "      \n",
        "print('x0123')\n",
        "print('x0567')\n",
        "print('x0890')\n",
        "      \n",
        "print('x9123')\n",
        "print('x9567')\n",
        "print('x9890')\n",
        "\n",
        "print('x2123')\n",
        "print('x2567')\n",
        "print('x2890')\n",
        "\n",
        "print('x7123)\n",
        "print('x7567')\n",
        "print('x7890')\n",
        "\n",
        "print('x3123')\n",
        "print('x3756')\n",
        "print('x3789')\n",
        "\n",
        "print('x4123')\n",
        "print('x4756')\n",
        "print('x4789')\n",
        "\n",
        "print('x5123')\n",
        "print('x57567')\n",
        "print('x57890')\n",
        "      \n",
        "\n"
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}