2024_2_audiv027_char_rnn_quijote.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/FAU-UChile/audiv027-2024-2/blob/main/2024_2_audiv027_char_rnn_quijote.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZUD4nbAKe_8v"
      },
      "source": [
        "fuente\n",
        "https://blog.devgenius.io/python-environment-management-d920f20ea53e"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "18JMUFeUkBQi",
        "outputId": "29508c28-d781-40fd-e6c5-a0c5f77d83bc"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Cloning into 'training-charRNN'...\n",
            "remote: Enumerating objects: 105, done.\u001b[K\n",
            "remote: Counting objects: 100% (37/37), done.\u001b[K\n",
            "remote: Compressing objects: 100% (9/9), done.\u001b[K\n",
            "remote: Total 105 (delta 31), reused 28 (delta 28), pack-reused 68\u001b[K\n",
            "Receiving objects: 100% (105/105), 33.50 KiB | 11.17 MiB/s, done.\n",
            "Resolving deltas: 100% (51/51), done.\n"
          ]
        }
      ],
      "source": [
        "# clonar repositorio de ml5.js para entrenar algoritmo char-RNN\n",
        "! git clone https://github.com/ml5js/training-charRNN"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "F27hQCCikHlf",
        "outputId": "500c13ba-4d20-4365-9a4e-10fd89967361"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/content/training-charRNN\n"
          ]
        }
      ],
      "source": [
        "# cd = cambiar directorio, moverse a la carpeta instalada\n",
        "%cd training-charRNN/"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rcpUrQmpkIIr",
        "outputId": "294667c9-4be2-4026-ece7-d736b093ecd5"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/content/training-charRNN\n"
          ]
        }
      ],
      "source": [
        "# pwd es print working directory = imprimir directorio actual\n",
        "!pwd"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "P6MC89pLdvyN",
        "outputId": "86e848d0-92cc-4a8a-e2b8-7e74742dd697"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/bin/bash: line 1: pyenv: command not found\n"
          ]
        }
      ],
      "source": [
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && eval \"$(pyenv init — path)\""
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AK92XccMdXdj",
        "outputId": "299b8db8-6a9f-4ba5-fbe5-84955610992d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current\n",
            "                                 Dload  Upload   Total   Spent    Left  Speed\n",
            "100   270  100   270    0     0   1250      0 --:--:-- --:--:-- --:--:--  1255\n",
            "Cloning into '/root/.pyenv'...\n",
            "remote: Enumerating objects: 1246, done.\u001b[K\n",
            "remote: Counting objects: 100% (1246/1246), done.\u001b[K\n",
            "remote: Compressing objects: 100% (692/692), done.\u001b[K\n",
            "remote: Total 1246 (delta 728), reused 712 (delta 421), pack-reused 0\u001b[K\n",
            "Receiving objects: 100% (1246/1246), 620.33 KiB | 12.41 MiB/s, done.\n",
            "Resolving deltas: 100% (728/728), done.\n",
            "Cloning into '/root/.pyenv/plugins/pyenv-doctor'...\n",
            "remote: Enumerating objects: 11, done.\u001b[K\n",
            "remote: Counting objects: 100% (11/11), done.\u001b[K\n",
            "remote: Compressing objects: 100% (9/9), done.\u001b[K\n",
            "remote: Total 11 (delta 1), reused 5 (delta 0), pack-reused 0\u001b[K\n",
            "Receiving objects: 100% (11/11), 38.72 KiB | 9.68 MiB/s, done.\n",
            "Resolving deltas: 100% (1/1), done.\n",
            "Cloning into '/root/.pyenv/plugins/pyenv-update'...\n",
            "remote: Enumerating objects: 10, done.\u001b[K\n",
            "remote: Counting objects: 100% (10/10), done.\u001b[K\n",
            "remote: Compressing objects: 100% (6/6), done.\u001b[K\n",
            "Receiving objects: 100% (10/10), done.\n",
            "remote: Total 10 (delta 1), reused 5 (delta 0), pack-reused 0\u001b[K\n",
            "Resolving deltas: 100% (1/1), done.\n",
            "Cloning into '/root/.pyenv/plugins/pyenv-virtualenv'...\n",
            "remote: Enumerating objects: 64, done.\u001b[K\n",
            "remote: Counting objects: 100% (64/64), done.\u001b[K\n",
            "remote: Compressing objects: 100% (56/56), done.\u001b[K\n",
            "remote: Total 64 (delta 10), reused 30 (delta 1), pack-reused 0\u001b[K\n",
            "Receiving objects: 100% (64/64), 42.50 KiB | 10.62 MiB/s, done.\n",
            "Resolving deltas: 100% (10/10), done.\n",
            "\n",
            "\u001b[1mWARNING\u001b[m: seems you still have not added 'pyenv' to the load path.\n",
            "\n",
            "# Load pyenv automatically by appending\n",
            "# the following to \n",
            "# ~/.bash_profile if it exists, otherwise ~/.profile (for login shells)\n",
            "# and ~/.bashrc (for interactive shells) :\n",
            "\n",
            "export PYENV_ROOT=\"$HOME/.pyenv\"\n",
            "[[ -d $PYENV_ROOT/bin ]] && export PATH=\"$PYENV_ROOT/bin:$PATH\"\n",
            "eval \"$(pyenv init -)\"\n",
            "\n",
            "# Restart your shell for the changes to take effect.\n",
            "\n",
            "# Load pyenv-virtualenv automatically by adding\n",
            "# the following to ~/.bashrc:\n",
            "\n",
            "eval \"$(pyenv virtualenv-init -)\"\n",
            "\n"
          ]
        }
      ],
      "source": [
        "#https://github.com/pyenv/pyenv-installer\n",
        "!curl https://pyenv.run | bash"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3lLZZ9fCd2s6",
        "outputId": "ef86fc3c-2193-49b9-ef07-aafc041671bf"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "  3.3.6\n",
            "  3.6.0\n",
            "  3.6-dev\n",
            "  3.6.1\n",
            "  3.6.2\n",
            "  3.6.3\n",
            "  3.6.4\n",
            "  3.6.5\n",
            "  3.6.6\n",
            "  3.6.7\n",
            "  3.6.8\n",
            "  3.6.9\n",
            "  3.6.10\n",
            "  3.6.11\n",
            "  3.6.12\n",
            "  3.6.13\n",
            "  3.6.14\n",
            "  3.6.15\n",
            "  activepython-3.6.0\n",
            "  pypy2.7-7.3.6-src\n",
            "  pypy2.7-7.3.6\n",
            "  pypy3.6-7.0.0-src\n",
            "  pypy3.6-7.0.0\n",
            "  pypy3.6-7.1.0-src\n",
            "  pypy3.6-7.1.0\n",
            "  pypy3.6-7.1.1-src\n",
            "  pypy3.6-7.1.1\n",
            "  pypy3.6-7.2.0-src\n",
            "  pypy3.6-7.2.0\n",
            "  pypy3.6-7.3.0-src\n",
            "  pypy3.6-7.3.0\n",
            "  pypy3.6-7.3.1-src\n",
            "  pypy3.6-7.3.1\n",
            "  pypy3.6-7.3.2-src\n",
            "  pypy3.6-7.3.2\n",
            "  pypy3.6-7.3.3-src\n",
            "  pypy3.6-7.3.3\n",
            "  pypy3.7-7.3.6-src\n",
            "  pypy3.7-7.3.6\n",
            "  pypy3.8-7.3.6-src\n",
            "  pypy3.8-7.3.6\n"
          ]
        }
      ],
      "source": [
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && pyenv install -list | grep \"3\\.[6]\""
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "FlOjbW8je9bo",
        "outputId": "a4c3f744-2fa4-4c17-aadd-4b19e96a3d8a"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading Python-3.6.15.tar.xz...\n",
            "-> https://www.python.org/ftp/python/3.6.15/Python-3.6.15.tar.xz\n",
            "Installing Python-3.6.15...\n",
            "patching file configure\n",
            "patching file configure.ac\n",
            "patching file Misc/NEWS.d/next/macOS/2019-06-03-05-49-49.bpo-36231.RfmW_p.rst\n",
            "patching file setup.py\n",
            "patching file setup.py\n",
            "patching file Doc/library/ctypes.rst\n",
            "patching file Lib/test/test_unicode.py\n",
            "patching file Modules/_ctypes/_ctypes.c\n",
            "patching file Modules/_ctypes/callproc.c\n",
            "patching file Modules/_ctypes/ctypes.h\n",
            "patching file setup.py\n",
            "patching file Misc/NEWS.d/next/macOS/2020-06-24-13-51-57.bpo-41100.mcHdc5.rst\n",
            "patching file configure\n",
            "patching file configure.ac\n",
            "patching file 'Misc/NEWS.d/next/Core and Builtins/2020-06-30-04-44-29.bpo-41100.PJwA6F.rst'\n",
            "patching file Modules/_decimal/libmpdec/mpdecimal.h\n",
            "patching file setup.py\n",
            "patching file Misc/NEWS.d/next/Build/2021-10-11-16-27-38.bpo-45405.iSfdW5.rst\n",
            "patching file configure\n",
            "patching file configure.ac\n",
            "patching file 'Misc/NEWS.d/next/Core and Builtins/2019-04-16-11-52-21.bpo-27987.n2_DcQ.rst'\n",
            "patching file Objects/obmalloc.c\n",
            "patching file Include/objimpl.h\n",
            "patching file 'Misc/NEWS.d/next/Core and Builtins/2019-05-15-18-28-43.bpo-27987.FaxuLy.rst'\n",
            "Installed Python-3.6.15 to /root/.pyenv/versions/3.6.15\n",
            "* system (set by /root/.pyenv/version)\n",
            "  3.6.15\n"
          ]
        }
      ],
      "source": [
        "# install python 3.6.15\n",
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && pyenv install 3.6.15\n",
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && pyenv versions"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Cl-I1bCtfM6M",
        "outputId": "43bf1c2c-cd16-450b-8992-2aa60554eda8"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in links: /tmp/tmpdaqwmah_\n",
            "Requirement already satisfied: setuptools in /root/.pyenv/versions/3.6.15/envs/training-charRNN/lib/python3.6/site-packages (40.6.2)\n",
            "Requirement already satisfied: pip in /root/.pyenv/versions/3.6.15/envs/training-charRNN/lib/python3.6/site-packages (18.1)\n",
            "* system (set by /root/.pyenv/version)\n",
            "  3.6.15\n",
            "  3.6.15/envs/training-charRNN\n",
            "  training-charRNN --> /root/.pyenv/versions/3.6.15/envs/training-charRNN\n"
          ]
        }
      ],
      "source": [
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && mkdir -p /content/training-charRNN/ && cd /content/training-charRNN/ && pyenv virtualenv 3.6.15 training-charRNN\n",
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && pyenv versions"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rIImYKRCgshV",
        "outputId": "1ef84e94-930e-4863-c24f-2cf834d8058a"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "* system (set by /root/.pyenv/version)\n",
            "  3.6.15\n",
            "  3.6.15/envs/training-charRNN\n",
            "  training-charRNN --> /root/.pyenv/versions/3.6.15/envs/training-charRNN\n",
            "  system\n",
            "* 3.6.15 (set by /root/.pyenv/version)\n",
            "  3.6.15/envs/training-charRNN\n",
            "  training-charRNN --> /root/.pyenv/versions/3.6.15/envs/training-charRNN\n"
          ]
        }
      ],
      "source": [
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && pyenv versions\n",
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && pyenv global 3.6.15\n",
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && pyenv versions"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gOs20Mt6iRsD",
        "outputId": "19c6a7f2-a921-4a70-dd47-13ea2c9b73a0"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Python 3.10.12\n"
          ]
        }
      ],
      "source": [
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && python --version"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "iqBfmc2Mi3Dg",
        "outputId": "3ef4b8e5-a6fe-40c9-ec90-ed141d989a23"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "pyenv: version `proyecto1' not installed\n",
            "Python 3.6.15\n"
          ]
        }
      ],
      "source": [
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && pyenv local proyecto1 && pyenv which python\n",
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && eval \"$(pyenv init --path)\" && pyenv local training-charRNN/ && cd /content/training-charRNN/ && python -V"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "sHo8SvI_i-Ud",
        "outputId": "d9762b8a-0e00-406d-bce3-5484cd3c941b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Python 3.6.15\n"
          ]
        }
      ],
      "source": [
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && eval \"$(pyenv init --path)\" && pyenv local training-charRNN/ && cd /content/training-charRNN/ && python -V"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Xg8pOjMBjdKx",
        "outputId": "41813182-c91d-44f0-e05a-637a25592bb2"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "pyenv: version `proyecto1' not installed\n"
          ]
        }
      ],
      "source": [
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && eval \"$(pyenv init --path)\" && pyenv local proyecto1 && cd /content/proyecto1 && wget https://bootstrap.pypa.io/pip/3.6/get-pip.py && python3.6 get-pip.py"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "hNhoYG7ujpE6"
      },
      "outputs": [],
      "source": [
        "# prefijo\n",
        "#!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && eval \"$(pyenv init --path)\" &&"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Ag-7uOEyjsBs",
        "outputId": "e27559f9-144b-4998-f9bf-cd759c157113"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "pip 18.1 from /root/.pyenv/versions/training-charRNN/lib/python3.6/site-packages/pip (python 3.6)\n"
          ]
        }
      ],
      "source": [
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && eval \"$(pyenv init --path)\" && python3 -m pip --version"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9poFEy91mIyO",
        "outputId": "1bb6ad53-b6da-49a5-ee01-888532c321b8"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Collecting protobuf==3.19.6\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/32/27/1141a8232723dcb10a595cc0ce4321dcbbd5215300bf4acfc142343205bf/protobuf-3.19.6-py2.py3-none-any.whl (162kB)\n",
            "\u001b[K    100% |████████████████████████████████| 163kB 8.3MB/s \n",
            "\u001b[?25hInstalling collected packages: protobuf\n",
            "Successfully installed protobuf-3.19.6\n",
            "Collecting tensorflow==1.14.0\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/de/f0/96fb2e0412ae9692dbf400e5b04432885f677ad6241c088ccc5fe7724d69/tensorflow-1.14.0-cp36-cp36m-manylinux1_x86_64.whl (109.2MB)\n",
            "\u001b[K    100% |████████████████████████████████| 109.2MB 166kB/s \n",
            "\u001b[?25hCollecting keras-applications>=1.0.6 (from tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/71/e3/19762fdfc62877ae9102edf6342d71b28fbfd9dea3d2f96a882ce099b03f/Keras_Applications-1.0.8-py3-none-any.whl (50kB)\n",
            "\u001b[K    100% |████████████████████████████████| 51kB 13.7MB/s \n",
            "\u001b[?25hCollecting absl-py>=0.7.0 (from tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/dd/87/de5c32fa1b1c6c3305d576e299801d8655c175ca9557019906247b994331/absl_py-1.4.0-py3-none-any.whl (126kB)\n",
            "\u001b[K    100% |████████████████████████████████| 133kB 28.9MB/s \n",
            "\u001b[?25hCollecting google-pasta>=0.1.6 (from tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/a3/de/c648ef6835192e6e2cc03f40b19eeda4382c49b5bafb43d88b931c4c74ac/google_pasta-0.2.0-py3-none-any.whl (57kB)\n",
            "\u001b[K    100% |████████████████████████████████| 61kB 17.1MB/s \n",
            "\u001b[?25hCollecting six>=1.10.0 (from tensorflow==1.14.0)\n",
            "  Downloading https://files.pythonhosted.org/packages/d9/5a/e7c31adbe875f2abbb91bd84cf2dc52d792b5a01506781dbcf25c91daf11/six-1.16.0-py2.py3-none-any.whl\n",
            "Collecting grpcio>=1.8.6 (from tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/81/9a/6b33e8d15850356772f0ee6489bc8346a7aa90f0c86733283e139740865e/grpcio-1.48.2.tar.gz (22.0MB)\n",
            "\u001b[K    100% |████████████████████████████████| 22.0MB 1.9MB/s \n",
            "\u001b[?25hCollecting numpy<2.0,>=1.14.5 (from tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/45/b2/6c7545bb7a38754d63048c7696804a0d947328125d81bf12beaa692c3ae3/numpy-1.19.5-cp36-cp36m-manylinux1_x86_64.whl (13.4MB)\n",
            "\u001b[K    100% |████████████████████████████████| 13.4MB 868kB/s \n",
            "\u001b[?25hCollecting keras-preprocessing>=1.0.5 (from tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/79/4c/7c3275a01e12ef9368a892926ab932b33bb13d55794881e3573482b378a7/Keras_Preprocessing-1.1.2-py2.py3-none-any.whl (42kB)\n",
            "\u001b[K    100% |████████████████████████████████| 51kB 6.4MB/s \n",
            "\u001b[?25hCollecting gast>=0.2.0 (from tensorflow==1.14.0)\n",
            "  Downloading https://files.pythonhosted.org/packages/fa/39/5aae571e5a5f4de9c3445dae08a530498e5c53b0e74410eeeb0991c79047/gast-0.5.4-py3-none-any.whl\n",
            "Collecting astor>=0.6.0 (from tensorflow==1.14.0)\n",
            "  Downloading https://files.pythonhosted.org/packages/c3/88/97eef84f48fa04fbd6750e62dcceafba6c63c81b7ac1420856c8dcc0a3f9/astor-0.8.1-py2.py3-none-any.whl\n",
            "Collecting tensorboard<1.15.0,>=1.14.0 (from tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/91/2d/2ed263449a078cd9c8a9ba50ebd50123adf1f8cfbea1492f9084169b89d9/tensorboard-1.14.0-py3-none-any.whl (3.1MB)\n",
            "\u001b[K    100% |████████████████████████████████| 3.2MB 3.4MB/s \n",
            "\u001b[?25hCollecting tensorflow-estimator<1.15.0rc0,>=1.14.0rc0 (from tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/3c/d5/21860a5b11caf0678fbc8319341b0ae21a07156911132e0e71bffed0510d/tensorflow_estimator-1.14.0-py2.py3-none-any.whl (488kB)\n",
            "\u001b[K    100% |████████████████████████████████| 491kB 12.4MB/s \n",
            "\u001b[?25hCollecting wrapt>=1.11.1 (from tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/66/50/1c5ccb23dd63f8f3d312dc2d5e0e64c681faf7c2388fa1ab2553713cef11/wrapt-1.16.0-cp36-cp36m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_17_x86_64.manylinux2014_x86_64.whl (77kB)\n",
            "\u001b[K    100% |████████████████████████████████| 81kB 12.2MB/s \n",
            "\u001b[?25hCollecting wheel>=0.26 (from tensorflow==1.14.0)\n",
            "  Downloading https://files.pythonhosted.org/packages/27/d6/003e593296a85fd6ed616ed962795b2f87709c3eee2bca4f6d0fe55c6d00/wheel-0.37.1-py2.py3-none-any.whl\n",
            "Collecting termcolor>=1.1.0 (from tensorflow==1.14.0)\n",
            "  Downloading https://files.pythonhosted.org/packages/8a/48/a76be51647d0eb9f10e2a4511bf3ffb8cc1e6b14e9e4fab46173aa79f981/termcolor-1.1.0.tar.gz\n",
            "Requirement already satisfied: protobuf>=3.6.1 in /root/.pyenv/versions/3.6.15/envs/training-charRNN/lib/python3.6/site-packages (from tensorflow==1.14.0) (3.19.6)\n",
            "Collecting h5py (from keras-applications>=1.0.6->tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/70/7a/e53e500335afb6b1aade11227cdf107fca54106a1dca5c9d13242a043f3b/h5py-3.1.0-cp36-cp36m-manylinux1_x86_64.whl (4.0MB)\n",
            "\u001b[K    100% |████████████████████████████████| 4.0MB 1.2MB/s \n",
            "\u001b[?25hCollecting setuptools>=41.0.0 (from tensorboard<1.15.0,>=1.14.0->tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/b0/3a/88b210db68e56854d0bcf4b38e165e03be377e13907746f825790f3df5bf/setuptools-59.6.0-py3-none-any.whl (952kB)\n",
            "\u001b[K    100% |████████████████████████████████| 962kB 12.5MB/s \n",
            "\u001b[?25hCollecting markdown>=2.6.8 (from tensorboard<1.15.0,>=1.14.0->tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/f3/df/ca72f352e15b6f8ce32b74af029f1189abffb906f7c137501ffe69c98a65/Markdown-3.3.7-py3-none-any.whl (97kB)\n",
            "\u001b[K    100% |████████████████████████████████| 102kB 16.0MB/s \n",
            "\u001b[?25hCollecting werkzeug>=0.11.15 (from tensorboard<1.15.0,>=1.14.0->tensorflow==1.14.0)\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/f4/f3/22afbdb20cc4654b10c98043414a14057cd27fdba9d4ae61cea596000ba2/Werkzeug-2.0.3-py3-none-any.whl (289kB)\n",
            "\u001b[K    100% |████████████████████████████████| 296kB 24.1MB/s \n",
            "\u001b[?25hCollecting cached-property; python_version < \"3.8\" (from h5py->keras-applications>=1.0.6->tensorflow==1.14.0)\n",
            "  Downloading https://files.pythonhosted.org/packages/48/19/f2090f7dad41e225c7f2326e4cfe6fff49e57dedb5b53636c9551f86b069/cached_property-1.5.2-py2.py3-none-any.whl\n",
            "Collecting importlib-metadata>=4.4; python_version < \"3.10\" (from markdown>=2.6.8->tensorboard<1.15.0,>=1.14.0->tensorflow==1.14.0)\n",
            "  Downloading https://files.pythonhosted.org/packages/a0/a1/b153a0a4caf7a7e3f15c2cd56c7702e2cf3d89b1b359d1f1c5e59d68f4ce/importlib_metadata-4.8.3-py3-none-any.whl\n",
            "Collecting dataclasses; python_version < \"3.7\" (from werkzeug>=0.11.15->tensorboard<1.15.0,>=1.14.0->tensorflow==1.14.0)\n",
            "  Downloading https://files.pythonhosted.org/packages/fe/ca/75fac5856ab5cfa51bbbcefa250182e50441074fdc3f803f6e76451fab43/dataclasses-0.8-py3-none-any.whl\n",
            "Collecting zipp>=0.5 (from importlib-metadata>=4.4; python_version < \"3.10\"->markdown>=2.6.8->tensorboard<1.15.0,>=1.14.0->tensorflow==1.14.0)\n",
            "  Downloading https://files.pythonhosted.org/packages/bd/df/d4a4974a3e3957fd1c1fa3082366d7fff6e428ddb55f074bf64876f8e8ad/zipp-3.6.0-py3-none-any.whl\n",
            "Collecting typing-extensions>=3.6.4; python_version < \"3.8\" (from importlib-metadata>=4.4; python_version < \"3.10\"->markdown>=2.6.8->tensorboard<1.15.0,>=1.14.0->tensorflow==1.14.0)\n",
            "  Downloading https://files.pythonhosted.org/packages/45/6b/44f7f8f1e110027cf88956b59f2fad776cca7e1704396d043f89effd3a0e/typing_extensions-4.1.1-py3-none-any.whl\n",
            "Installing collected packages: cached-property, numpy, h5py, keras-applications, absl-py, six, google-pasta, grpcio, keras-preprocessing, gast, astor, setuptools, zipp, typing-extensions, importlib-metadata, markdown, dataclasses, werkzeug, wheel, tensorboard, tensorflow-estimator, wrapt, termcolor, tensorflow\n",
            "\u001b[33m  The scripts f2py, f2py3 and f2py3.6 are installed in '/root/.pyenv/versions/training-charRNN/bin' which is not on PATH.\n",
            "  Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location.\u001b[0m\n",
            "  Running setup.py install for grpcio ... \u001b[?25ldone\n",
            "\u001b[?25h  Found existing installation: setuptools 40.6.2\n",
            "    Uninstalling setuptools-40.6.2:\n",
            "      Successfully uninstalled setuptools-40.6.2\n",
            "\u001b[33m  The script markdown_py is installed in '/root/.pyenv/versions/training-charRNN/bin' which is not on PATH.\n",
            "  Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location.\u001b[0m\n",
            "\u001b[33m  The script wheel is installed in '/root/.pyenv/versions/training-charRNN/bin' which is not on PATH.\n",
            "  Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location.\u001b[0m\n",
            "\u001b[33m  The script tensorboard is installed in '/root/.pyenv/versions/training-charRNN/bin' which is not on PATH.\n",
            "  Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location.\u001b[0m\n",
            "  Running setup.py install for termcolor ... \u001b[?25ldone\n",
            "\u001b[33m  The scripts freeze_graph, saved_model_cli, tensorboard, tf_upgrade_v2, tflite_convert, toco and toco_from_protos are installed in '/root/.pyenv/versions/training-charRNN/bin' which is not on PATH.\n",
            "  Consider adding this directory to PATH or, if you prefer to suppress this warning, use --no-warn-script-location.\u001b[0m\n",
            "\u001b[?25hSuccessfully installed absl-py-1.4.0 astor-0.8.1 cached-property-1.5.2 dataclasses-0.8 gast-0.5.4 google-pasta-0.2.0 grpcio-1.48.2 h5py-3.1.0 importlib-metadata-4.8.3 keras-applications-1.0.8 keras-preprocessing-1.1.2 markdown-3.3.7 numpy-1.19.5 setuptools-59.6.0 six-1.16.0 tensorboard-1.14.0 tensorflow-1.14.0 tensorflow-estimator-1.14.0 termcolor-1.1.0 typing-extensions-4.1.1 werkzeug-2.0.3 wheel-0.37.1 wrapt-1.16.0 zipp-3.6.0\n"
          ]
        }
      ],
      "source": [
        "# usar python3.6 y su pip para instalar tensorflow v1.15.4\n",
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && eval \"$(pyenv init --path)\" && pip install protobuf==3.19.6\n",
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && eval \"$(pyenv init --path)\" && python3 -m pip install tensorflow==1.14.0"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "cRoW9mOYlf2W",
        "outputId": "4537d9d4-b4d1-40f0-fe4a-4f8d88be7703"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "--2024-04-12 20:51:26--  https://gutenberg.org/files/345/345-0.txt\n",
            "Resolving gutenberg.org (gutenberg.org)... 152.19.134.47, 2610:28:3090:3000:0:bad:cafe:47\n",
            "Connecting to gutenberg.org (gutenberg.org)|152.19.134.47|:443... connected.\n",
            "HTTP request sent, awaiting response... 200 OK\n",
            "Length: 890196 (869K) [text/plain]\n",
            "Saving to: ‘dracula.txt’\n",
            "\n",
            "dracula.txt         100%[===================>] 869.33K  --.-KB/s    in 0.07s   \n",
            "\n",
            "2024-04-12 20:51:27 (11.6 MB/s) - ‘dracula.txt’ saved [890196/890196]\n",
            "\n"
          ]
        }
      ],
      "source": [
        "# descargar archivos desde Project Gutenberg\n",
        "# esta es la lista de los más populares https://gutenberg.org/ebooks/search/?sort_order=downloads\n",
        "# para este ejemplo bajamos el quijote\n",
        "\n",
        "# ! wget -O quijote.txt https://gutenberg.org/files/2000/2000-0.txt\n",
        "# ! wget -O sherlock.txt https://gutenberg.org/files/1661/1661-0.txt\n",
        "# ! wget -O frankenstein.txt https://gutenberg.org/files/84/84-0.txt\n",
        "# ! wget -O prideAndPrejudice.txt https://gutenberg.org/files/1342/1342-0.txt\n",
        "# ! wget -O aliceInWonderland.txt https://gutenberg.org/files/11/11-0.txt\n",
        "! wget -O dracula.txt https://gutenberg.org/files/345/345-0.txt\n",
        "# ! wget -O mobyDick.txt https://gutenberg.org/files/2701/2701-0.txt\n",
        "# ! wget -O greatGatsby.txt https://gutenberg.org/ebooks/64317.txt.utf-8\n",
        "# ! wget -O metamorphosis.txt https://gutenberg.org/files/5200/5200-0.txt\n",
        "# ! wget -O iliad.txt https://gutenberg.org/files/6130/6130-0.txt\n",
        "# ! wget -O wutheringHeights.txt https://gutenberg.org/files/768/768-0.txt\n",
        "# ! wget -O mysteriousAffair.txt https://gutenberg.org/ebooks/863"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ZvuoU3j9li4k",
        "outputId": "14d5ae28-62dc-478a-94e0-2f1faccfbd34"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[1;30;43mSe truncaron las últimas líneas 5000 del resultado de transmisión.\u001b[0m\n",
            "12609/17600 (epoch 35), train_loss = 1.308, time/batch = 0.150\n",
            "12610/17600 (epoch 35), train_loss = 1.373, time/batch = 0.174\n",
            "12611/17600 (epoch 35), train_loss = 1.278, time/batch = 0.155\n",
            "12612/17600 (epoch 35), train_loss = 1.272, time/batch = 0.182\n",
            "12613/17600 (epoch 35), train_loss = 1.346, time/batch = 0.171\n",
            "12614/17600 (epoch 35), train_loss = 1.297, time/batch = 0.165\n",
            "12615/17600 (epoch 35), train_loss = 1.357, time/batch = 0.173\n",
            "12616/17600 (epoch 35), train_loss = 1.300, time/batch = 0.154\n",
            "12617/17600 (epoch 35), train_loss = 1.332, time/batch = 0.164\n",
            "12618/17600 (epoch 35), train_loss = 1.310, time/batch = 0.160\n",
            "12619/17600 (epoch 35), train_loss = 1.329, time/batch = 0.187\n",
            "12620/17600 (epoch 35), train_loss = 1.360, time/batch = 0.169\n",
            "12621/17600 (epoch 35), train_loss = 1.313, time/batch = 0.168\n",
            "12622/17600 (epoch 35), train_loss = 1.384, time/batch = 0.155\n",
            "12623/17600 (epoch 35), train_loss = 1.379, time/batch = 0.160\n",
            "12624/17600 (epoch 35), train_loss = 1.389, time/batch = 0.149\n",
            "12625/17600 (epoch 35), train_loss = 1.301, time/batch = 0.164\n",
            "12626/17600 (epoch 35), train_loss = 1.341, time/batch = 0.178\n",
            "12627/17600 (epoch 35), train_loss = 1.343, time/batch = 0.167\n",
            "12628/17600 (epoch 35), train_loss = 1.345, time/batch = 0.159\n",
            "12629/17600 (epoch 35), train_loss = 1.345, time/batch = 0.161\n",
            "12630/17600 (epoch 35), train_loss = 1.383, time/batch = 0.165\n",
            "12631/17600 (epoch 35), train_loss = 1.329, time/batch = 0.157\n",
            "12632/17600 (epoch 35), train_loss = 1.360, time/batch = 0.182\n",
            "12633/17600 (epoch 35), train_loss = 1.299, time/batch = 0.161\n",
            "12634/17600 (epoch 35), train_loss = 1.315, time/batch = 0.149\n",
            "12635/17600 (epoch 35), train_loss = 1.280, time/batch = 0.159\n",
            "12636/17600 (epoch 35), train_loss = 1.380, time/batch = 0.153\n",
            "12637/17600 (epoch 35), train_loss = 1.366, time/batch = 0.158\n",
            "12638/17600 (epoch 35), train_loss = 1.365, time/batch = 0.178\n",
            "12639/17600 (epoch 35), train_loss = 1.345, time/batch = 0.153\n",
            "12640/17600 (epoch 35), train_loss = 1.388, time/batch = 0.160\n",
            "12641/17600 (epoch 35), train_loss = 1.355, time/batch = 0.153\n",
            "12642/17600 (epoch 35), train_loss = 1.410, time/batch = 0.158\n",
            "12643/17600 (epoch 35), train_loss = 1.376, time/batch = 0.166\n",
            "12644/17600 (epoch 35), train_loss = 1.357, time/batch = 0.153\n",
            "12645/17600 (epoch 35), train_loss = 1.312, time/batch = 0.181\n",
            "12646/17600 (epoch 35), train_loss = 1.385, time/batch = 0.157\n",
            "12647/17600 (epoch 35), train_loss = 1.368, time/batch = 0.159\n",
            "12648/17600 (epoch 35), train_loss = 1.403, time/batch = 0.165\n",
            "12649/17600 (epoch 35), train_loss = 1.371, time/batch = 0.176\n",
            "12650/17600 (epoch 35), train_loss = 1.295, time/batch = 0.182\n",
            "12651/17600 (epoch 35), train_loss = 1.323, time/batch = 0.186\n",
            "12652/17600 (epoch 35), train_loss = 1.374, time/batch = 0.166\n",
            "12653/17600 (epoch 35), train_loss = 1.379, time/batch = 0.150\n",
            "12654/17600 (epoch 35), train_loss = 1.336, time/batch = 0.176\n",
            "12655/17600 (epoch 35), train_loss = 1.330, time/batch = 0.155\n",
            "12656/17600 (epoch 35), train_loss = 1.340, time/batch = 0.175\n",
            "12657/17600 (epoch 35), train_loss = 1.371, time/batch = 0.175\n",
            "12658/17600 (epoch 35), train_loss = 1.322, time/batch = 0.169\n",
            "12659/17600 (epoch 35), train_loss = 1.355, time/batch = 0.164\n",
            "12660/17600 (epoch 35), train_loss = 1.318, time/batch = 0.199\n",
            "12661/17600 (epoch 35), train_loss = 1.361, time/batch = 0.245\n",
            "12662/17600 (epoch 35), train_loss = 1.373, time/batch = 0.265\n",
            "12663/17600 (epoch 35), train_loss = 1.336, time/batch = 0.280\n",
            "12664/17600 (epoch 35), train_loss = 1.316, time/batch = 0.291\n",
            "12665/17600 (epoch 35), train_loss = 1.332, time/batch = 0.274\n",
            "12666/17600 (epoch 35), train_loss = 1.353, time/batch = 0.284\n",
            "12667/17600 (epoch 35), train_loss = 1.327, time/batch = 0.251\n",
            "12668/17600 (epoch 35), train_loss = 1.351, time/batch = 0.282\n",
            "12669/17600 (epoch 35), train_loss = 1.335, time/batch = 0.280\n",
            "12670/17600 (epoch 35), train_loss = 1.328, time/batch = 0.307\n",
            "12671/17600 (epoch 35), train_loss = 1.320, time/batch = 0.198\n",
            "12672/17600 (epoch 36), train_loss = 1.145, time/batch = 0.166\n",
            "12673/17600 (epoch 36), train_loss = 1.346, time/batch = 0.151\n",
            "12674/17600 (epoch 36), train_loss = 1.346, time/batch = 0.175\n",
            "12675/17600 (epoch 36), train_loss = 1.288, time/batch = 0.148\n",
            "12676/17600 (epoch 36), train_loss = 1.380, time/batch = 0.158\n",
            "12677/17600 (epoch 36), train_loss = 1.357, time/batch = 0.159\n",
            "12678/17600 (epoch 36), train_loss = 1.307, time/batch = 0.161\n",
            "12679/17600 (epoch 36), train_loss = 1.295, time/batch = 0.145\n",
            "12680/17600 (epoch 36), train_loss = 1.334, time/batch = 0.174\n",
            "12681/17600 (epoch 36), train_loss = 1.389, time/batch = 0.157\n",
            "12682/17600 (epoch 36), train_loss = 1.317, time/batch = 0.172\n",
            "12683/17600 (epoch 36), train_loss = 1.335, time/batch = 0.156\n",
            "12684/17600 (epoch 36), train_loss = 1.313, time/batch = 0.168\n",
            "12685/17600 (epoch 36), train_loss = 1.297, time/batch = 0.158\n",
            "12686/17600 (epoch 36), train_loss = 1.277, time/batch = 0.181\n",
            "12687/17600 (epoch 36), train_loss = 1.341, time/batch = 0.156\n",
            "12688/17600 (epoch 36), train_loss = 1.352, time/batch = 0.162\n",
            "12689/17600 (epoch 36), train_loss = 1.308, time/batch = 0.155\n",
            "12690/17600 (epoch 36), train_loss = 1.322, time/batch = 0.169\n",
            "12691/17600 (epoch 36), train_loss = 1.365, time/batch = 0.158\n",
            "12692/17600 (epoch 36), train_loss = 1.288, time/batch = 0.188\n",
            "12693/17600 (epoch 36), train_loss = 1.341, time/batch = 0.160\n",
            "12694/17600 (epoch 36), train_loss = 1.307, time/batch = 0.155\n",
            "12695/17600 (epoch 36), train_loss = 1.284, time/batch = 0.160\n",
            "12696/17600 (epoch 36), train_loss = 1.369, time/batch = 0.184\n",
            "12697/17600 (epoch 36), train_loss = 1.296, time/batch = 0.164\n",
            "12698/17600 (epoch 36), train_loss = 1.346, time/batch = 0.163\n",
            "12699/17600 (epoch 36), train_loss = 1.327, time/batch = 0.161\n",
            "12700/17600 (epoch 36), train_loss = 1.305, time/batch = 0.167\n",
            "12701/17600 (epoch 36), train_loss = 1.237, time/batch = 0.157\n",
            "12702/17600 (epoch 36), train_loss = 1.347, time/batch = 0.166\n",
            "12703/17600 (epoch 36), train_loss = 1.327, time/batch = 0.155\n",
            "12704/17600 (epoch 36), train_loss = 1.304, time/batch = 0.169\n",
            "12705/17600 (epoch 36), train_loss = 1.310, time/batch = 0.172\n",
            "12706/17600 (epoch 36), train_loss = 1.311, time/batch = 0.170\n",
            "12707/17600 (epoch 36), train_loss = 1.324, time/batch = 0.154\n",
            "12708/17600 (epoch 36), train_loss = 1.348, time/batch = 0.162\n",
            "12709/17600 (epoch 36), train_loss = 1.347, time/batch = 0.164\n",
            "12710/17600 (epoch 36), train_loss = 1.329, time/batch = 0.165\n",
            "12711/17600 (epoch 36), train_loss = 1.305, time/batch = 0.170\n",
            "12712/17600 (epoch 36), train_loss = 1.355, time/batch = 0.161\n",
            "12713/17600 (epoch 36), train_loss = 1.287, time/batch = 0.150\n",
            "12714/17600 (epoch 36), train_loss = 1.342, time/batch = 0.158\n",
            "12715/17600 (epoch 36), train_loss = 1.262, time/batch = 0.160\n",
            "12716/17600 (epoch 36), train_loss = 1.333, time/batch = 0.166\n",
            "12717/17600 (epoch 36), train_loss = 1.303, time/batch = 0.156\n",
            "12718/17600 (epoch 36), train_loss = 1.260, time/batch = 0.182\n",
            "12719/17600 (epoch 36), train_loss = 1.319, time/batch = 0.160\n",
            "12720/17600 (epoch 36), train_loss = 1.269, time/batch = 0.164\n",
            "12721/17600 (epoch 36), train_loss = 1.289, time/batch = 0.156\n",
            "12722/17600 (epoch 36), train_loss = 1.308, time/batch = 0.153\n",
            "12723/17600 (epoch 36), train_loss = 1.301, time/batch = 0.157\n",
            "12724/17600 (epoch 36), train_loss = 1.268, time/batch = 0.178\n",
            "12725/17600 (epoch 36), train_loss = 1.298, time/batch = 0.160\n",
            "12726/17600 (epoch 36), train_loss = 1.344, time/batch = 0.161\n",
            "12727/17600 (epoch 36), train_loss = 1.322, time/batch = 0.155\n",
            "12728/17600 (epoch 36), train_loss = 1.325, time/batch = 0.162\n",
            "12729/17600 (epoch 36), train_loss = 1.317, time/batch = 0.154\n",
            "12730/17600 (epoch 36), train_loss = 1.297, time/batch = 0.217\n",
            "12731/17600 (epoch 36), train_loss = 1.322, time/batch = 0.274\n",
            "12732/17600 (epoch 36), train_loss = 1.317, time/batch = 0.334\n",
            "12733/17600 (epoch 36), train_loss = 1.260, time/batch = 0.282\n",
            "12734/17600 (epoch 36), train_loss = 1.263, time/batch = 0.254\n",
            "12735/17600 (epoch 36), train_loss = 1.283, time/batch = 0.262\n",
            "12736/17600 (epoch 36), train_loss = 1.312, time/batch = 0.264\n",
            "12737/17600 (epoch 36), train_loss = 1.280, time/batch = 0.268\n",
            "12738/17600 (epoch 36), train_loss = 1.318, time/batch = 0.284\n",
            "12739/17600 (epoch 36), train_loss = 1.273, time/batch = 0.316\n",
            "12740/17600 (epoch 36), train_loss = 1.322, time/batch = 0.270\n",
            "12741/17600 (epoch 36), train_loss = 1.253, time/batch = 0.155\n",
            "12742/17600 (epoch 36), train_loss = 1.301, time/batch = 0.182\n",
            "12743/17600 (epoch 36), train_loss = 1.344, time/batch = 0.159\n",
            "12744/17600 (epoch 36), train_loss = 1.305, time/batch = 0.157\n",
            "12745/17600 (epoch 36), train_loss = 1.378, time/batch = 0.156\n",
            "12746/17600 (epoch 36), train_loss = 1.307, time/batch = 0.164\n",
            "12747/17600 (epoch 36), train_loss = 1.310, time/batch = 0.161\n",
            "12748/17600 (epoch 36), train_loss = 1.264, time/batch = 0.182\n",
            "12749/17600 (epoch 36), train_loss = 1.318, time/batch = 0.151\n",
            "12750/17600 (epoch 36), train_loss = 1.327, time/batch = 0.175\n",
            "12751/17600 (epoch 36), train_loss = 1.400, time/batch = 0.152\n",
            "12752/17600 (epoch 36), train_loss = 1.370, time/batch = 0.161\n",
            "12753/17600 (epoch 36), train_loss = 1.364, time/batch = 0.154\n",
            "12754/17600 (epoch 36), train_loss = 1.345, time/batch = 0.162\n",
            "12755/17600 (epoch 36), train_loss = 1.315, time/batch = 0.171\n",
            "12756/17600 (epoch 36), train_loss = 1.292, time/batch = 0.170\n",
            "12757/17600 (epoch 36), train_loss = 1.327, time/batch = 0.161\n",
            "12758/17600 (epoch 36), train_loss = 1.319, time/batch = 0.164\n",
            "12759/17600 (epoch 36), train_loss = 1.259, time/batch = 0.149\n",
            "12760/17600 (epoch 36), train_loss = 1.297, time/batch = 0.184\n",
            "12761/17600 (epoch 36), train_loss = 1.239, time/batch = 0.165\n",
            "12762/17600 (epoch 36), train_loss = 1.305, time/batch = 0.164\n",
            "12763/17600 (epoch 36), train_loss = 1.289, time/batch = 0.147\n",
            "12764/17600 (epoch 36), train_loss = 1.268, time/batch = 0.163\n",
            "12765/17600 (epoch 36), train_loss = 1.322, time/batch = 0.152\n",
            "12766/17600 (epoch 36), train_loss = 1.305, time/batch = 0.179\n",
            "12767/17600 (epoch 36), train_loss = 1.270, time/batch = 0.175\n",
            "12768/17600 (epoch 36), train_loss = 1.246, time/batch = 0.160\n",
            "12769/17600 (epoch 36), train_loss = 1.295, time/batch = 0.156\n",
            "12770/17600 (epoch 36), train_loss = 1.328, time/batch = 0.169\n",
            "12771/17600 (epoch 36), train_loss = 1.287, time/batch = 0.162\n",
            "12772/17600 (epoch 36), train_loss = 1.260, time/batch = 0.167\n",
            "12773/17600 (epoch 36), train_loss = 1.345, time/batch = 0.174\n",
            "12774/17600 (epoch 36), train_loss = 1.286, time/batch = 0.159\n",
            "12775/17600 (epoch 36), train_loss = 1.396, time/batch = 0.158\n",
            "12776/17600 (epoch 36), train_loss = 1.321, time/batch = 0.165\n",
            "12777/17600 (epoch 36), train_loss = 1.294, time/batch = 0.145\n",
            "12778/17600 (epoch 36), train_loss = 1.336, time/batch = 0.173\n",
            "12779/17600 (epoch 36), train_loss = 1.321, time/batch = 0.158\n",
            "12780/17600 (epoch 36), train_loss = 1.326, time/batch = 0.178\n",
            "12781/17600 (epoch 36), train_loss = 1.311, time/batch = 0.169\n",
            "12782/17600 (epoch 36), train_loss = 1.267, time/batch = 0.172\n",
            "12783/17600 (epoch 36), train_loss = 1.290, time/batch = 0.158\n",
            "12784/17600 (epoch 36), train_loss = 1.319, time/batch = 0.165\n",
            "12785/17600 (epoch 36), train_loss = 1.212, time/batch = 0.156\n",
            "12786/17600 (epoch 36), train_loss = 1.288, time/batch = 0.185\n",
            "12787/17600 (epoch 36), train_loss = 1.320, time/batch = 0.158\n",
            "12788/17600 (epoch 36), train_loss = 1.289, time/batch = 0.166\n",
            "12789/17600 (epoch 36), train_loss = 1.313, time/batch = 0.162\n",
            "12790/17600 (epoch 36), train_loss = 1.257, time/batch = 0.162\n",
            "12791/17600 (epoch 36), train_loss = 1.359, time/batch = 0.158\n",
            "12792/17600 (epoch 36), train_loss = 1.265, time/batch = 0.187\n",
            "12793/17600 (epoch 36), train_loss = 1.327, time/batch = 0.157\n",
            "12794/17600 (epoch 36), train_loss = 1.351, time/batch = 0.154\n",
            "12795/17600 (epoch 36), train_loss = 1.339, time/batch = 0.170\n",
            "12796/17600 (epoch 36), train_loss = 1.254, time/batch = 0.157\n",
            "12797/17600 (epoch 36), train_loss = 1.319, time/batch = 0.164\n",
            "12798/17600 (epoch 36), train_loss = 1.317, time/batch = 0.178\n",
            "12799/17600 (epoch 36), train_loss = 1.304, time/batch = 0.165\n",
            "12800/17600 (epoch 36), train_loss = 1.286, time/batch = 0.153\n",
            "12801/17600 (epoch 36), train_loss = 1.320, time/batch = 0.226\n",
            "12802/17600 (epoch 36), train_loss = 1.333, time/batch = 0.281\n",
            "12803/17600 (epoch 36), train_loss = 1.336, time/batch = 0.303\n",
            "12804/17600 (epoch 36), train_loss = 1.305, time/batch = 0.279\n",
            "12805/17600 (epoch 36), train_loss = 1.279, time/batch = 0.275\n",
            "12806/17600 (epoch 36), train_loss = 1.288, time/batch = 0.267\n",
            "12807/17600 (epoch 36), train_loss = 1.260, time/batch = 0.277\n",
            "12808/17600 (epoch 36), train_loss = 1.332, time/batch = 0.250\n",
            "12809/17600 (epoch 36), train_loss = 1.353, time/batch = 0.259\n",
            "12810/17600 (epoch 36), train_loss = 1.299, time/batch = 0.305\n",
            "12811/17600 (epoch 36), train_loss = 1.349, time/batch = 0.282\n",
            "12812/17600 (epoch 36), train_loss = 1.380, time/batch = 0.170\n",
            "12813/17600 (epoch 36), train_loss = 1.287, time/batch = 0.172\n",
            "12814/17600 (epoch 36), train_loss = 1.315, time/batch = 0.159\n",
            "12815/17600 (epoch 36), train_loss = 1.280, time/batch = 0.172\n",
            "12816/17600 (epoch 36), train_loss = 1.335, time/batch = 0.177\n",
            "12817/17600 (epoch 36), train_loss = 1.317, time/batch = 0.169\n",
            "12818/17600 (epoch 36), train_loss = 1.294, time/batch = 0.152\n",
            "12819/17600 (epoch 36), train_loss = 1.301, time/batch = 0.165\n",
            "12820/17600 (epoch 36), train_loss = 1.284, time/batch = 0.154\n",
            "12821/17600 (epoch 36), train_loss = 1.346, time/batch = 0.164\n",
            "12822/17600 (epoch 36), train_loss = 1.236, time/batch = 0.164\n",
            "12823/17600 (epoch 36), train_loss = 1.307, time/batch = 0.176\n",
            "12824/17600 (epoch 36), train_loss = 1.333, time/batch = 0.146\n",
            "12825/17600 (epoch 36), train_loss = 1.300, time/batch = 0.162\n",
            "12826/17600 (epoch 36), train_loss = 1.318, time/batch = 0.162\n",
            "12827/17600 (epoch 36), train_loss = 1.253, time/batch = 0.163\n",
            "12828/17600 (epoch 36), train_loss = 1.252, time/batch = 0.149\n",
            "12829/17600 (epoch 36), train_loss = 1.269, time/batch = 0.178\n",
            "12830/17600 (epoch 36), train_loss = 1.330, time/batch = 0.153\n",
            "12831/17600 (epoch 36), train_loss = 1.374, time/batch = 0.170\n",
            "12832/17600 (epoch 36), train_loss = 1.329, time/batch = 0.155\n",
            "12833/17600 (epoch 36), train_loss = 1.330, time/batch = 0.163\n",
            "12834/17600 (epoch 36), train_loss = 1.260, time/batch = 0.155\n",
            "12835/17600 (epoch 36), train_loss = 1.290, time/batch = 0.189\n",
            "12836/17600 (epoch 36), train_loss = 1.311, time/batch = 0.161\n",
            "12837/17600 (epoch 36), train_loss = 1.298, time/batch = 0.157\n",
            "12838/17600 (epoch 36), train_loss = 1.311, time/batch = 0.154\n",
            "12839/17600 (epoch 36), train_loss = 1.338, time/batch = 0.170\n",
            "12840/17600 (epoch 36), train_loss = 1.327, time/batch = 0.153\n",
            "12841/17600 (epoch 36), train_loss = 1.307, time/batch = 0.172\n",
            "12842/17600 (epoch 36), train_loss = 1.244, time/batch = 0.149\n",
            "12843/17600 (epoch 36), train_loss = 1.256, time/batch = 0.169\n",
            "12844/17600 (epoch 36), train_loss = 1.344, time/batch = 0.156\n",
            "12845/17600 (epoch 36), train_loss = 1.310, time/batch = 0.163\n",
            "12846/17600 (epoch 36), train_loss = 1.301, time/batch = 0.164\n",
            "12847/17600 (epoch 36), train_loss = 1.305, time/batch = 0.165\n",
            "12848/17600 (epoch 36), train_loss = 1.221, time/batch = 0.171\n",
            "12849/17600 (epoch 36), train_loss = 1.285, time/batch = 0.166\n",
            "12850/17600 (epoch 36), train_loss = 1.284, time/batch = 0.159\n",
            "12851/17600 (epoch 36), train_loss = 1.261, time/batch = 0.159\n",
            "12852/17600 (epoch 36), train_loss = 1.260, time/batch = 0.151\n",
            "12853/17600 (epoch 36), train_loss = 1.313, time/batch = 0.156\n",
            "12854/17600 (epoch 36), train_loss = 1.369, time/batch = 0.165\n",
            "12855/17600 (epoch 36), train_loss = 1.286, time/batch = 0.163\n",
            "12856/17600 (epoch 36), train_loss = 1.286, time/batch = 0.157\n",
            "12857/17600 (epoch 36), train_loss = 1.250, time/batch = 0.164\n",
            "12858/17600 (epoch 36), train_loss = 1.327, time/batch = 0.150\n",
            "12859/17600 (epoch 36), train_loss = 1.335, time/batch = 0.174\n",
            "12860/17600 (epoch 36), train_loss = 1.253, time/batch = 0.174\n",
            "12861/17600 (epoch 36), train_loss = 1.352, time/batch = 0.162\n",
            "12862/17600 (epoch 36), train_loss = 1.312, time/batch = 0.152\n",
            "12863/17600 (epoch 36), train_loss = 1.295, time/batch = 0.166\n",
            "12864/17600 (epoch 36), train_loss = 1.305, time/batch = 0.164\n",
            "12865/17600 (epoch 36), train_loss = 1.322, time/batch = 0.172\n",
            "12866/17600 (epoch 36), train_loss = 1.323, time/batch = 0.152\n",
            "12867/17600 (epoch 36), train_loss = 1.333, time/batch = 0.175\n",
            "12868/17600 (epoch 36), train_loss = 1.291, time/batch = 0.153\n",
            "12869/17600 (epoch 36), train_loss = 1.340, time/batch = 0.179\n",
            "12870/17600 (epoch 36), train_loss = 1.288, time/batch = 0.148\n",
            "12871/17600 (epoch 36), train_loss = 1.322, time/batch = 0.165\n",
            "12872/17600 (epoch 36), train_loss = 1.326, time/batch = 0.168\n",
            "12873/17600 (epoch 36), train_loss = 1.292, time/batch = 0.284\n",
            "12874/17600 (epoch 36), train_loss = 1.284, time/batch = 0.289\n",
            "12875/17600 (epoch 36), train_loss = 1.310, time/batch = 0.288\n",
            "12876/17600 (epoch 36), train_loss = 1.346, time/batch = 0.292\n",
            "12877/17600 (epoch 36), train_loss = 1.392, time/batch = 0.280\n",
            "12878/17600 (epoch 36), train_loss = 1.306, time/batch = 0.277\n",
            "12879/17600 (epoch 36), train_loss = 1.307, time/batch = 0.296\n",
            "12880/17600 (epoch 36), train_loss = 1.307, time/batch = 0.280\n",
            "12881/17600 (epoch 36), train_loss = 1.287, time/batch = 0.251\n",
            "12882/17600 (epoch 36), train_loss = 1.331, time/batch = 0.273\n",
            "12883/17600 (epoch 36), train_loss = 1.322, time/batch = 0.205\n",
            "12884/17600 (epoch 36), train_loss = 1.290, time/batch = 0.151\n",
            "12885/17600 (epoch 36), train_loss = 1.365, time/batch = 0.167\n",
            "12886/17600 (epoch 36), train_loss = 1.408, time/batch = 0.155\n",
            "12887/17600 (epoch 36), train_loss = 1.296, time/batch = 0.145\n",
            "12888/17600 (epoch 36), train_loss = 1.320, time/batch = 0.160\n",
            "12889/17600 (epoch 36), train_loss = 1.272, time/batch = 0.148\n",
            "12890/17600 (epoch 36), train_loss = 1.310, time/batch = 0.167\n",
            "12891/17600 (epoch 36), train_loss = 1.291, time/batch = 0.166\n",
            "12892/17600 (epoch 36), train_loss = 1.281, time/batch = 0.168\n",
            "12893/17600 (epoch 36), train_loss = 1.309, time/batch = 0.182\n",
            "12894/17600 (epoch 36), train_loss = 1.375, time/batch = 0.172\n",
            "12895/17600 (epoch 36), train_loss = 1.347, time/batch = 0.153\n",
            "12896/17600 (epoch 36), train_loss = 1.264, time/batch = 0.170\n",
            "12897/17600 (epoch 36), train_loss = 1.315, time/batch = 0.168\n",
            "12898/17600 (epoch 36), train_loss = 1.319, time/batch = 0.168\n",
            "12899/17600 (epoch 36), train_loss = 1.301, time/batch = 0.154\n",
            "12900/17600 (epoch 36), train_loss = 1.329, time/batch = 0.171\n",
            "12901/17600 (epoch 36), train_loss = 1.294, time/batch = 0.148\n",
            "12902/17600 (epoch 36), train_loss = 1.284, time/batch = 0.158\n",
            "12903/17600 (epoch 36), train_loss = 1.350, time/batch = 0.171\n",
            "12904/17600 (epoch 36), train_loss = 1.344, time/batch = 0.179\n",
            "12905/17600 (epoch 36), train_loss = 1.360, time/batch = 0.152\n",
            "12906/17600 (epoch 36), train_loss = 1.273, time/batch = 0.170\n",
            "12907/17600 (epoch 36), train_loss = 1.305, time/batch = 0.150\n",
            "12908/17600 (epoch 36), train_loss = 1.293, time/batch = 0.168\n",
            "12909/17600 (epoch 36), train_loss = 1.364, time/batch = 0.151\n",
            "12910/17600 (epoch 36), train_loss = 1.359, time/batch = 0.177\n",
            "12911/17600 (epoch 36), train_loss = 1.255, time/batch = 0.150\n",
            "12912/17600 (epoch 36), train_loss = 1.346, time/batch = 0.154\n",
            "12913/17600 (epoch 36), train_loss = 1.292, time/batch = 0.163\n",
            "12914/17600 (epoch 36), train_loss = 1.327, time/batch = 0.156\n",
            "12915/17600 (epoch 36), train_loss = 1.320, time/batch = 0.155\n",
            "12916/17600 (epoch 36), train_loss = 1.290, time/batch = 0.167\n",
            "12917/17600 (epoch 36), train_loss = 1.329, time/batch = 0.168\n",
            "12918/17600 (epoch 36), train_loss = 1.322, time/batch = 0.161\n",
            "12919/17600 (epoch 36), train_loss = 1.295, time/batch = 0.158\n",
            "12920/17600 (epoch 36), train_loss = 1.272, time/batch = 0.154\n",
            "12921/17600 (epoch 36), train_loss = 1.250, time/batch = 0.153\n",
            "12922/17600 (epoch 36), train_loss = 1.386, time/batch = 0.167\n",
            "12923/17600 (epoch 36), train_loss = 1.324, time/batch = 0.179\n",
            "12924/17600 (epoch 36), train_loss = 1.319, time/batch = 0.173\n",
            "12925/17600 (epoch 36), train_loss = 1.325, time/batch = 0.156\n",
            "12926/17600 (epoch 36), train_loss = 1.341, time/batch = 0.163\n",
            "12927/17600 (epoch 36), train_loss = 1.351, time/batch = 0.162\n",
            "12928/17600 (epoch 36), train_loss = 1.397, time/batch = 0.176\n",
            "12929/17600 (epoch 36), train_loss = 1.365, time/batch = 0.179\n",
            "12930/17600 (epoch 36), train_loss = 1.317, time/batch = 0.152\n",
            "12931/17600 (epoch 36), train_loss = 1.318, time/batch = 0.169\n",
            "12932/17600 (epoch 36), train_loss = 1.297, time/batch = 0.157\n",
            "12933/17600 (epoch 36), train_loss = 1.302, time/batch = 0.172\n",
            "12934/17600 (epoch 36), train_loss = 1.285, time/batch = 0.158\n",
            "12935/17600 (epoch 36), train_loss = 1.242, time/batch = 0.176\n",
            "12936/17600 (epoch 36), train_loss = 1.292, time/batch = 0.147\n",
            "12937/17600 (epoch 36), train_loss = 1.300, time/batch = 0.160\n",
            "12938/17600 (epoch 36), train_loss = 1.253, time/batch = 0.145\n",
            "12939/17600 (epoch 36), train_loss = 1.257, time/batch = 0.165\n",
            "12940/17600 (epoch 36), train_loss = 1.290, time/batch = 0.160\n",
            "12941/17600 (epoch 36), train_loss = 1.324, time/batch = 0.167\n",
            "12942/17600 (epoch 36), train_loss = 1.249, time/batch = 0.181\n",
            "12943/17600 (epoch 36), train_loss = 1.317, time/batch = 0.165\n",
            "12944/17600 (epoch 36), train_loss = 1.283, time/batch = 0.209\n",
            "12945/17600 (epoch 36), train_loss = 1.305, time/batch = 0.280\n",
            "12946/17600 (epoch 36), train_loss = 1.291, time/batch = 0.255\n",
            "12947/17600 (epoch 36), train_loss = 1.346, time/batch = 0.278\n",
            "12948/17600 (epoch 36), train_loss = 1.310, time/batch = 0.243\n",
            "12949/17600 (epoch 36), train_loss = 1.337, time/batch = 0.269\n",
            "12950/17600 (epoch 36), train_loss = 1.305, time/batch = 0.285\n",
            "12951/17600 (epoch 36), train_loss = 1.309, time/batch = 0.305\n",
            "12952/17600 (epoch 36), train_loss = 1.333, time/batch = 0.282\n",
            "12953/17600 (epoch 36), train_loss = 1.379, time/batch = 0.251\n",
            "12954/17600 (epoch 36), train_loss = 1.349, time/batch = 0.291\n",
            "12955/17600 (epoch 36), train_loss = 1.348, time/batch = 0.174\n",
            "12956/17600 (epoch 36), train_loss = 1.369, time/batch = 0.154\n",
            "12957/17600 (epoch 36), train_loss = 1.379, time/batch = 0.159\n",
            "12958/17600 (epoch 36), train_loss = 1.322, time/batch = 0.160\n",
            "12959/17600 (epoch 36), train_loss = 1.369, time/batch = 0.166\n",
            "12960/17600 (epoch 36), train_loss = 1.307, time/batch = 0.185\n",
            "12961/17600 (epoch 36), train_loss = 1.306, time/batch = 0.155\n",
            "12962/17600 (epoch 36), train_loss = 1.371, time/batch = 0.172\n",
            "12963/17600 (epoch 36), train_loss = 1.276, time/batch = 0.158\n",
            "12964/17600 (epoch 36), train_loss = 1.270, time/batch = 0.165\n",
            "12965/17600 (epoch 36), train_loss = 1.344, time/batch = 0.160\n",
            "12966/17600 (epoch 36), train_loss = 1.295, time/batch = 0.158\n",
            "12967/17600 (epoch 36), train_loss = 1.354, time/batch = 0.170\n",
            "12968/17600 (epoch 36), train_loss = 1.299, time/batch = 0.163\n",
            "12969/17600 (epoch 36), train_loss = 1.329, time/batch = 0.160\n",
            "12970/17600 (epoch 36), train_loss = 1.308, time/batch = 0.169\n",
            "12971/17600 (epoch 36), train_loss = 1.326, time/batch = 0.164\n",
            "12972/17600 (epoch 36), train_loss = 1.357, time/batch = 0.168\n",
            "12973/17600 (epoch 36), train_loss = 1.311, time/batch = 0.167\n",
            "12974/17600 (epoch 36), train_loss = 1.382, time/batch = 0.153\n",
            "12975/17600 (epoch 36), train_loss = 1.376, time/batch = 0.147\n",
            "12976/17600 (epoch 36), train_loss = 1.386, time/batch = 0.159\n",
            "12977/17600 (epoch 36), train_loss = 1.298, time/batch = 0.163\n",
            "12978/17600 (epoch 36), train_loss = 1.338, time/batch = 0.179\n",
            "12979/17600 (epoch 36), train_loss = 1.340, time/batch = 0.184\n",
            "12980/17600 (epoch 36), train_loss = 1.343, time/batch = 0.168\n",
            "12981/17600 (epoch 36), train_loss = 1.343, time/batch = 0.169\n",
            "12982/17600 (epoch 36), train_loss = 1.380, time/batch = 0.158\n",
            "12983/17600 (epoch 36), train_loss = 1.327, time/batch = 0.180\n",
            "12984/17600 (epoch 36), train_loss = 1.358, time/batch = 0.147\n",
            "12985/17600 (epoch 36), train_loss = 1.297, time/batch = 0.189\n",
            "12986/17600 (epoch 36), train_loss = 1.312, time/batch = 0.160\n",
            "12987/17600 (epoch 36), train_loss = 1.278, time/batch = 0.180\n",
            "12988/17600 (epoch 36), train_loss = 1.377, time/batch = 0.149\n",
            "12989/17600 (epoch 36), train_loss = 1.364, time/batch = 0.162\n",
            "12990/17600 (epoch 36), train_loss = 1.363, time/batch = 0.161\n",
            "12991/17600 (epoch 36), train_loss = 1.343, time/batch = 0.169\n",
            "12992/17600 (epoch 36), train_loss = 1.385, time/batch = 0.170\n",
            "12993/17600 (epoch 36), train_loss = 1.352, time/batch = 0.176\n",
            "12994/17600 (epoch 36), train_loss = 1.407, time/batch = 0.157\n",
            "12995/17600 (epoch 36), train_loss = 1.374, time/batch = 0.166\n",
            "12996/17600 (epoch 36), train_loss = 1.355, time/batch = 0.155\n",
            "12997/17600 (epoch 36), train_loss = 1.309, time/batch = 0.161\n",
            "12998/17600 (epoch 36), train_loss = 1.383, time/batch = 0.164\n",
            "12999/17600 (epoch 36), train_loss = 1.366, time/batch = 0.169\n",
            "13000/17600 (epoch 36), train_loss = 1.401, time/batch = 0.149\n",
            "Model saved to checkpoints/dracula/dracula!\n",
            "13001/17600 (epoch 36), train_loss = 1.368, time/batch = 0.160\n",
            "13002/17600 (epoch 36), train_loss = 1.293, time/batch = 0.161\n",
            "13003/17600 (epoch 36), train_loss = 1.320, time/batch = 0.159\n",
            "13004/17600 (epoch 36), train_loss = 1.371, time/batch = 0.178\n",
            "13005/17600 (epoch 36), train_loss = 1.377, time/batch = 0.172\n",
            "13006/17600 (epoch 36), train_loss = 1.334, time/batch = 0.167\n",
            "13007/17600 (epoch 36), train_loss = 1.327, time/batch = 0.157\n",
            "13008/17600 (epoch 36), train_loss = 1.337, time/batch = 0.169\n",
            "13009/17600 (epoch 36), train_loss = 1.369, time/batch = 0.148\n",
            "13010/17600 (epoch 36), train_loss = 1.320, time/batch = 0.160\n",
            "13011/17600 (epoch 36), train_loss = 1.352, time/batch = 0.171\n",
            "13012/17600 (epoch 36), train_loss = 1.316, time/batch = 0.167\n",
            "13013/17600 (epoch 36), train_loss = 1.358, time/batch = 0.161\n",
            "13014/17600 (epoch 36), train_loss = 1.370, time/batch = 0.170\n",
            "13015/17600 (epoch 36), train_loss = 1.333, time/batch = 0.155\n",
            "13016/17600 (epoch 36), train_loss = 1.313, time/batch = 0.166\n",
            "13017/17600 (epoch 36), train_loss = 1.330, time/batch = 0.170\n",
            "13018/17600 (epoch 36), train_loss = 1.351, time/batch = 0.165\n",
            "13019/17600 (epoch 36), train_loss = 1.324, time/batch = 0.166\n",
            "13020/17600 (epoch 36), train_loss = 1.349, time/batch = 0.162\n",
            "13021/17600 (epoch 36), train_loss = 1.333, time/batch = 0.165\n",
            "13022/17600 (epoch 36), train_loss = 1.325, time/batch = 0.159\n",
            "13023/17600 (epoch 36), train_loss = 1.317, time/batch = 0.151\n",
            "13024/17600 (epoch 37), train_loss = 1.135, time/batch = 0.162\n",
            "13025/17600 (epoch 37), train_loss = 1.345, time/batch = 0.166\n",
            "13026/17600 (epoch 37), train_loss = 1.343, time/batch = 0.171\n",
            "13027/17600 (epoch 37), train_loss = 1.285, time/batch = 0.158\n",
            "13028/17600 (epoch 37), train_loss = 1.379, time/batch = 0.177\n",
            "13029/17600 (epoch 37), train_loss = 1.356, time/batch = 0.158\n",
            "13030/17600 (epoch 37), train_loss = 1.305, time/batch = 0.206\n",
            "13031/17600 (epoch 37), train_loss = 1.292, time/batch = 0.170\n",
            "13032/17600 (epoch 37), train_loss = 1.331, time/batch = 0.169\n",
            "13033/17600 (epoch 37), train_loss = 1.387, time/batch = 0.151\n",
            "13034/17600 (epoch 37), train_loss = 1.314, time/batch = 0.178\n",
            "13035/17600 (epoch 37), train_loss = 1.331, time/batch = 0.151\n",
            "13036/17600 (epoch 37), train_loss = 1.310, time/batch = 0.168\n",
            "13037/17600 (epoch 37), train_loss = 1.294, time/batch = 0.165\n",
            "13038/17600 (epoch 37), train_loss = 1.274, time/batch = 0.163\n",
            "13039/17600 (epoch 37), train_loss = 1.339, time/batch = 0.156\n",
            "13040/17600 (epoch 37), train_loss = 1.350, time/batch = 0.192\n",
            "13041/17600 (epoch 37), train_loss = 1.306, time/batch = 0.154\n",
            "13042/17600 (epoch 37), train_loss = 1.320, time/batch = 0.167\n",
            "13043/17600 (epoch 37), train_loss = 1.363, time/batch = 0.155\n",
            "13044/17600 (epoch 37), train_loss = 1.285, time/batch = 0.155\n",
            "13045/17600 (epoch 37), train_loss = 1.339, time/batch = 0.149\n",
            "13046/17600 (epoch 37), train_loss = 1.304, time/batch = 0.154\n",
            "13047/17600 (epoch 37), train_loss = 1.281, time/batch = 0.170\n",
            "13048/17600 (epoch 37), train_loss = 1.366, time/batch = 0.164\n",
            "13049/17600 (epoch 37), train_loss = 1.295, time/batch = 0.154\n",
            "13050/17600 (epoch 37), train_loss = 1.344, time/batch = 0.154\n",
            "13051/17600 (epoch 37), train_loss = 1.325, time/batch = 0.152\n",
            "13052/17600 (epoch 37), train_loss = 1.303, time/batch = 0.158\n",
            "13053/17600 (epoch 37), train_loss = 1.236, time/batch = 0.172\n",
            "13054/17600 (epoch 37), train_loss = 1.345, time/batch = 0.249\n",
            "13055/17600 (epoch 37), train_loss = 1.325, time/batch = 0.264\n",
            "13056/17600 (epoch 37), train_loss = 1.301, time/batch = 0.282\n",
            "13057/17600 (epoch 37), train_loss = 1.307, time/batch = 0.269\n",
            "13058/17600 (epoch 37), train_loss = 1.308, time/batch = 0.292\n",
            "13059/17600 (epoch 37), train_loss = 1.322, time/batch = 0.266\n",
            "13060/17600 (epoch 37), train_loss = 1.347, time/batch = 0.298\n",
            "13061/17600 (epoch 37), train_loss = 1.344, time/batch = 0.301\n",
            "13062/17600 (epoch 37), train_loss = 1.326, time/batch = 0.280\n",
            "13063/17600 (epoch 37), train_loss = 1.302, time/batch = 0.278\n",
            "13064/17600 (epoch 37), train_loss = 1.352, time/batch = 0.165\n",
            "13065/17600 (epoch 37), train_loss = 1.284, time/batch = 0.180\n",
            "13066/17600 (epoch 37), train_loss = 1.340, time/batch = 0.158\n",
            "13067/17600 (epoch 37), train_loss = 1.260, time/batch = 0.172\n",
            "13068/17600 (epoch 37), train_loss = 1.331, time/batch = 0.150\n",
            "13069/17600 (epoch 37), train_loss = 1.300, time/batch = 0.142\n",
            "13070/17600 (epoch 37), train_loss = 1.257, time/batch = 0.159\n",
            "13071/17600 (epoch 37), train_loss = 1.317, time/batch = 0.145\n",
            "13072/17600 (epoch 37), train_loss = 1.267, time/batch = 0.179\n",
            "13073/17600 (epoch 37), train_loss = 1.286, time/batch = 0.149\n",
            "13074/17600 (epoch 37), train_loss = 1.306, time/batch = 0.168\n",
            "13075/17600 (epoch 37), train_loss = 1.298, time/batch = 0.172\n",
            "13076/17600 (epoch 37), train_loss = 1.266, time/batch = 0.164\n",
            "13077/17600 (epoch 37), train_loss = 1.296, time/batch = 0.152\n",
            "13078/17600 (epoch 37), train_loss = 1.341, time/batch = 0.188\n",
            "13079/17600 (epoch 37), train_loss = 1.320, time/batch = 0.155\n",
            "13080/17600 (epoch 37), train_loss = 1.322, time/batch = 0.162\n",
            "13081/17600 (epoch 37), train_loss = 1.314, time/batch = 0.162\n",
            "13082/17600 (epoch 37), train_loss = 1.294, time/batch = 0.154\n",
            "13083/17600 (epoch 37), train_loss = 1.319, time/batch = 0.167\n",
            "13084/17600 (epoch 37), train_loss = 1.315, time/batch = 0.184\n",
            "13085/17600 (epoch 37), train_loss = 1.257, time/batch = 0.175\n",
            "13086/17600 (epoch 37), train_loss = 1.260, time/batch = 0.156\n",
            "13087/17600 (epoch 37), train_loss = 1.280, time/batch = 0.169\n",
            "13088/17600 (epoch 37), train_loss = 1.310, time/batch = 0.151\n",
            "13089/17600 (epoch 37), train_loss = 1.277, time/batch = 0.162\n",
            "13090/17600 (epoch 37), train_loss = 1.315, time/batch = 0.156\n",
            "13091/17600 (epoch 37), train_loss = 1.271, time/batch = 0.179\n",
            "13092/17600 (epoch 37), train_loss = 1.320, time/batch = 0.152\n",
            "13093/17600 (epoch 37), train_loss = 1.251, time/batch = 0.164\n",
            "13094/17600 (epoch 37), train_loss = 1.298, time/batch = 0.151\n",
            "13095/17600 (epoch 37), train_loss = 1.342, time/batch = 0.161\n",
            "13096/17600 (epoch 37), train_loss = 1.302, time/batch = 0.147\n",
            "13097/17600 (epoch 37), train_loss = 1.375, time/batch = 0.181\n",
            "13098/17600 (epoch 37), train_loss = 1.304, time/batch = 0.161\n",
            "13099/17600 (epoch 37), train_loss = 1.307, time/batch = 0.170\n",
            "13100/17600 (epoch 37), train_loss = 1.261, time/batch = 0.151\n",
            "13101/17600 (epoch 37), train_loss = 1.316, time/batch = 0.164\n",
            "13102/17600 (epoch 37), train_loss = 1.324, time/batch = 0.156\n",
            "13103/17600 (epoch 37), train_loss = 1.398, time/batch = 0.178\n",
            "13104/17600 (epoch 37), train_loss = 1.368, time/batch = 0.149\n",
            "13105/17600 (epoch 37), train_loss = 1.362, time/batch = 0.163\n",
            "13106/17600 (epoch 37), train_loss = 1.342, time/batch = 0.155\n",
            "13107/17600 (epoch 37), train_loss = 1.313, time/batch = 0.168\n",
            "13108/17600 (epoch 37), train_loss = 1.291, time/batch = 0.160\n",
            "13109/17600 (epoch 37), train_loss = 1.325, time/batch = 0.158\n",
            "13110/17600 (epoch 37), train_loss = 1.316, time/batch = 0.185\n",
            "13111/17600 (epoch 37), train_loss = 1.256, time/batch = 0.147\n",
            "13112/17600 (epoch 37), train_loss = 1.295, time/batch = 0.170\n",
            "13113/17600 (epoch 37), train_loss = 1.236, time/batch = 0.150\n",
            "13114/17600 (epoch 37), train_loss = 1.301, time/batch = 0.186\n",
            "13115/17600 (epoch 37), train_loss = 1.287, time/batch = 0.167\n",
            "13116/17600 (epoch 37), train_loss = 1.266, time/batch = 0.176\n",
            "13117/17600 (epoch 37), train_loss = 1.320, time/batch = 0.155\n",
            "13118/17600 (epoch 37), train_loss = 1.303, time/batch = 0.166\n",
            "13119/17600 (epoch 37), train_loss = 1.267, time/batch = 0.151\n",
            "13120/17600 (epoch 37), train_loss = 1.243, time/batch = 0.158\n",
            "13121/17600 (epoch 37), train_loss = 1.292, time/batch = 0.156\n",
            "13122/17600 (epoch 37), train_loss = 1.326, time/batch = 0.177\n",
            "13123/17600 (epoch 37), train_loss = 1.284, time/batch = 0.152\n",
            "13124/17600 (epoch 37), train_loss = 1.257, time/batch = 0.170\n",
            "13125/17600 (epoch 37), train_loss = 1.343, time/batch = 0.257\n",
            "13126/17600 (epoch 37), train_loss = 1.283, time/batch = 0.307\n",
            "13127/17600 (epoch 37), train_loss = 1.393, time/batch = 0.318\n",
            "13128/17600 (epoch 37), train_loss = 1.318, time/batch = 0.280\n",
            "13129/17600 (epoch 37), train_loss = 1.292, time/batch = 0.272\n",
            "13130/17600 (epoch 37), train_loss = 1.334, time/batch = 0.296\n",
            "13131/17600 (epoch 37), train_loss = 1.319, time/batch = 0.289\n",
            "13132/17600 (epoch 37), train_loss = 1.324, time/batch = 0.266\n",
            "13133/17600 (epoch 37), train_loss = 1.308, time/batch = 0.271\n",
            "13134/17600 (epoch 37), train_loss = 1.264, time/batch = 0.305\n",
            "13135/17600 (epoch 37), train_loss = 1.288, time/batch = 0.182\n",
            "13136/17600 (epoch 37), train_loss = 1.317, time/batch = 0.154\n",
            "13137/17600 (epoch 37), train_loss = 1.210, time/batch = 0.154\n",
            "13138/17600 (epoch 37), train_loss = 1.286, time/batch = 0.175\n",
            "13139/17600 (epoch 37), train_loss = 1.318, time/batch = 0.159\n",
            "13140/17600 (epoch 37), train_loss = 1.287, time/batch = 0.187\n",
            "13141/17600 (epoch 37), train_loss = 1.311, time/batch = 0.164\n",
            "13142/17600 (epoch 37), train_loss = 1.255, time/batch = 0.174\n",
            "13143/17600 (epoch 37), train_loss = 1.357, time/batch = 0.153\n",
            "13144/17600 (epoch 37), train_loss = 1.263, time/batch = 0.168\n",
            "13145/17600 (epoch 37), train_loss = 1.324, time/batch = 0.162\n",
            "13146/17600 (epoch 37), train_loss = 1.349, time/batch = 0.178\n",
            "13147/17600 (epoch 37), train_loss = 1.336, time/batch = 0.161\n",
            "13148/17600 (epoch 37), train_loss = 1.252, time/batch = 0.170\n",
            "13149/17600 (epoch 37), train_loss = 1.317, time/batch = 0.158\n",
            "13150/17600 (epoch 37), train_loss = 1.315, time/batch = 0.164\n",
            "13151/17600 (epoch 37), train_loss = 1.301, time/batch = 0.161\n",
            "13152/17600 (epoch 37), train_loss = 1.284, time/batch = 0.175\n",
            "13153/17600 (epoch 37), train_loss = 1.317, time/batch = 0.168\n",
            "13154/17600 (epoch 37), train_loss = 1.330, time/batch = 0.156\n",
            "13155/17600 (epoch 37), train_loss = 1.334, time/batch = 0.159\n",
            "13156/17600 (epoch 37), train_loss = 1.303, time/batch = 0.158\n",
            "13157/17600 (epoch 37), train_loss = 1.276, time/batch = 0.158\n",
            "13158/17600 (epoch 37), train_loss = 1.286, time/batch = 0.170\n",
            "13159/17600 (epoch 37), train_loss = 1.258, time/batch = 0.170\n",
            "13160/17600 (epoch 37), train_loss = 1.329, time/batch = 0.166\n",
            "13161/17600 (epoch 37), train_loss = 1.351, time/batch = 0.153\n",
            "13162/17600 (epoch 37), train_loss = 1.297, time/batch = 0.171\n",
            "13163/17600 (epoch 37), train_loss = 1.347, time/batch = 0.152\n",
            "13164/17600 (epoch 37), train_loss = 1.378, time/batch = 0.159\n",
            "13165/17600 (epoch 37), train_loss = 1.286, time/batch = 0.175\n",
            "13166/17600 (epoch 37), train_loss = 1.312, time/batch = 0.171\n",
            "13167/17600 (epoch 37), train_loss = 1.278, time/batch = 0.145\n",
            "13168/17600 (epoch 37), train_loss = 1.332, time/batch = 0.160\n",
            "13169/17600 (epoch 37), train_loss = 1.315, time/batch = 0.166\n",
            "13170/17600 (epoch 37), train_loss = 1.291, time/batch = 0.166\n",
            "13171/17600 (epoch 37), train_loss = 1.299, time/batch = 0.181\n",
            "13172/17600 (epoch 37), train_loss = 1.281, time/batch = 0.173\n",
            "13173/17600 (epoch 37), train_loss = 1.344, time/batch = 0.156\n",
            "13174/17600 (epoch 37), train_loss = 1.233, time/batch = 0.178\n",
            "13175/17600 (epoch 37), train_loss = 1.305, time/batch = 0.167\n",
            "13176/17600 (epoch 37), train_loss = 1.330, time/batch = 0.162\n",
            "13177/17600 (epoch 37), train_loss = 1.297, time/batch = 0.159\n",
            "13178/17600 (epoch 37), train_loss = 1.315, time/batch = 0.174\n",
            "13179/17600 (epoch 37), train_loss = 1.251, time/batch = 0.163\n",
            "13180/17600 (epoch 37), train_loss = 1.250, time/batch = 0.169\n",
            "13181/17600 (epoch 37), train_loss = 1.267, time/batch = 0.162\n",
            "13182/17600 (epoch 37), train_loss = 1.328, time/batch = 0.155\n",
            "13183/17600 (epoch 37), train_loss = 1.372, time/batch = 0.158\n",
            "13184/17600 (epoch 37), train_loss = 1.326, time/batch = 0.183\n",
            "13185/17600 (epoch 37), train_loss = 1.327, time/batch = 0.147\n",
            "13186/17600 (epoch 37), train_loss = 1.258, time/batch = 0.162\n",
            "13187/17600 (epoch 37), train_loss = 1.288, time/batch = 0.159\n",
            "13188/17600 (epoch 37), train_loss = 1.309, time/batch = 0.167\n",
            "13189/17600 (epoch 37), train_loss = 1.296, time/batch = 0.169\n",
            "13190/17600 (epoch 37), train_loss = 1.309, time/batch = 0.166\n",
            "13191/17600 (epoch 37), train_loss = 1.335, time/batch = 0.163\n",
            "13192/17600 (epoch 37), train_loss = 1.325, time/batch = 0.165\n",
            "13193/17600 (epoch 37), train_loss = 1.305, time/batch = 0.159\n",
            "13194/17600 (epoch 37), train_loss = 1.242, time/batch = 0.166\n",
            "13195/17600 (epoch 37), train_loss = 1.253, time/batch = 0.218\n",
            "13196/17600 (epoch 37), train_loss = 1.341, time/batch = 0.314\n",
            "13197/17600 (epoch 37), train_loss = 1.308, time/batch = 0.322\n",
            "13198/17600 (epoch 37), train_loss = 1.299, time/batch = 0.291\n",
            "13199/17600 (epoch 37), train_loss = 1.302, time/batch = 0.302\n",
            "13200/17600 (epoch 37), train_loss = 1.219, time/batch = 0.273\n",
            "13201/17600 (epoch 37), train_loss = 1.282, time/batch = 0.267\n",
            "13202/17600 (epoch 37), train_loss = 1.282, time/batch = 0.274\n",
            "13203/17600 (epoch 37), train_loss = 1.259, time/batch = 0.287\n",
            "13204/17600 (epoch 37), train_loss = 1.257, time/batch = 0.301\n",
            "13205/17600 (epoch 37), train_loss = 1.310, time/batch = 0.257\n",
            "13206/17600 (epoch 37), train_loss = 1.367, time/batch = 0.155\n",
            "13207/17600 (epoch 37), train_loss = 1.283, time/batch = 0.162\n",
            "13208/17600 (epoch 37), train_loss = 1.284, time/batch = 0.178\n",
            "13209/17600 (epoch 37), train_loss = 1.247, time/batch = 0.169\n",
            "13210/17600 (epoch 37), train_loss = 1.325, time/batch = 0.161\n",
            "13211/17600 (epoch 37), train_loss = 1.332, time/batch = 0.165\n",
            "13212/17600 (epoch 37), train_loss = 1.251, time/batch = 0.146\n",
            "13213/17600 (epoch 37), train_loss = 1.349, time/batch = 0.164\n",
            "13214/17600 (epoch 37), train_loss = 1.311, time/batch = 0.174\n",
            "13215/17600 (epoch 37), train_loss = 1.293, time/batch = 0.160\n",
            "13216/17600 (epoch 37), train_loss = 1.302, time/batch = 0.173\n",
            "13217/17600 (epoch 37), train_loss = 1.320, time/batch = 0.155\n",
            "13218/17600 (epoch 37), train_loss = 1.321, time/batch = 0.167\n",
            "13219/17600 (epoch 37), train_loss = 1.331, time/batch = 0.145\n",
            "13220/17600 (epoch 37), train_loss = 1.288, time/batch = 0.178\n",
            "13221/17600 (epoch 37), train_loss = 1.338, time/batch = 0.149\n",
            "13222/17600 (epoch 37), train_loss = 1.286, time/batch = 0.172\n",
            "13223/17600 (epoch 37), train_loss = 1.319, time/batch = 0.148\n",
            "13224/17600 (epoch 37), train_loss = 1.323, time/batch = 0.169\n",
            "13225/17600 (epoch 37), train_loss = 1.289, time/batch = 0.150\n",
            "13226/17600 (epoch 37), train_loss = 1.281, time/batch = 0.169\n",
            "13227/17600 (epoch 37), train_loss = 1.308, time/batch = 0.168\n",
            "13228/17600 (epoch 37), train_loss = 1.343, time/batch = 0.165\n",
            "13229/17600 (epoch 37), train_loss = 1.390, time/batch = 0.165\n",
            "13230/17600 (epoch 37), train_loss = 1.304, time/batch = 0.162\n",
            "13231/17600 (epoch 37), train_loss = 1.304, time/batch = 0.160\n",
            "13232/17600 (epoch 37), train_loss = 1.305, time/batch = 0.147\n",
            "13233/17600 (epoch 37), train_loss = 1.285, time/batch = 0.184\n",
            "13234/17600 (epoch 37), train_loss = 1.329, time/batch = 0.160\n",
            "13235/17600 (epoch 37), train_loss = 1.320, time/batch = 0.168\n",
            "13236/17600 (epoch 37), train_loss = 1.288, time/batch = 0.161\n",
            "13237/17600 (epoch 37), train_loss = 1.363, time/batch = 0.164\n",
            "13238/17600 (epoch 37), train_loss = 1.405, time/batch = 0.145\n",
            "13239/17600 (epoch 37), train_loss = 1.293, time/batch = 0.178\n",
            "13240/17600 (epoch 37), train_loss = 1.317, time/batch = 0.153\n",
            "13241/17600 (epoch 37), train_loss = 1.270, time/batch = 0.170\n",
            "13242/17600 (epoch 37), train_loss = 1.307, time/batch = 0.166\n",
            "13243/17600 (epoch 37), train_loss = 1.288, time/batch = 0.172\n",
            "13244/17600 (epoch 37), train_loss = 1.278, time/batch = 0.154\n",
            "13245/17600 (epoch 37), train_loss = 1.305, time/batch = 0.182\n",
            "13246/17600 (epoch 37), train_loss = 1.372, time/batch = 0.155\n",
            "13247/17600 (epoch 37), train_loss = 1.345, time/batch = 0.179\n",
            "13248/17600 (epoch 37), train_loss = 1.261, time/batch = 0.155\n",
            "13249/17600 (epoch 37), train_loss = 1.312, time/batch = 0.158\n",
            "13250/17600 (epoch 37), train_loss = 1.316, time/batch = 0.157\n",
            "13251/17600 (epoch 37), train_loss = 1.297, time/batch = 0.154\n",
            "13252/17600 (epoch 37), train_loss = 1.326, time/batch = 0.165\n",
            "13253/17600 (epoch 37), train_loss = 1.292, time/batch = 0.175\n",
            "13254/17600 (epoch 37), train_loss = 1.281, time/batch = 0.159\n",
            "13255/17600 (epoch 37), train_loss = 1.347, time/batch = 0.163\n",
            "13256/17600 (epoch 37), train_loss = 1.342, time/batch = 0.147\n",
            "13257/17600 (epoch 37), train_loss = 1.357, time/batch = 0.174\n",
            "13258/17600 (epoch 37), train_loss = 1.270, time/batch = 0.164\n",
            "13259/17600 (epoch 37), train_loss = 1.302, time/batch = 0.161\n",
            "13260/17600 (epoch 37), train_loss = 1.290, time/batch = 0.157\n",
            "13261/17600 (epoch 37), train_loss = 1.361, time/batch = 0.175\n",
            "13262/17600 (epoch 37), train_loss = 1.357, time/batch = 0.155\n",
            "13263/17600 (epoch 37), train_loss = 1.252, time/batch = 0.161\n",
            "13264/17600 (epoch 37), train_loss = 1.344, time/batch = 0.174\n",
            "13265/17600 (epoch 37), train_loss = 1.290, time/batch = 0.170\n",
            "13266/17600 (epoch 37), train_loss = 1.325, time/batch = 0.193\n",
            "13267/17600 (epoch 37), train_loss = 1.318, time/batch = 0.252\n",
            "13268/17600 (epoch 37), train_loss = 1.287, time/batch = 0.262\n",
            "13269/17600 (epoch 37), train_loss = 1.327, time/batch = 0.285\n",
            "13270/17600 (epoch 37), train_loss = 1.319, time/batch = 0.268\n",
            "13271/17600 (epoch 37), train_loss = 1.292, time/batch = 0.283\n",
            "13272/17600 (epoch 37), train_loss = 1.269, time/batch = 0.265\n",
            "13273/17600 (epoch 37), train_loss = 1.248, time/batch = 0.296\n",
            "13274/17600 (epoch 37), train_loss = 1.383, time/batch = 0.277\n",
            "13275/17600 (epoch 37), train_loss = 1.321, time/batch = 0.299\n",
            "13276/17600 (epoch 37), train_loss = 1.316, time/batch = 0.289\n",
            "13277/17600 (epoch 37), train_loss = 1.322, time/batch = 0.170\n",
            "13278/17600 (epoch 37), train_loss = 1.339, time/batch = 0.160\n",
            "13279/17600 (epoch 37), train_loss = 1.348, time/batch = 0.157\n",
            "13280/17600 (epoch 37), train_loss = 1.395, time/batch = 0.157\n",
            "13281/17600 (epoch 37), train_loss = 1.363, time/batch = 0.150\n",
            "13282/17600 (epoch 37), train_loss = 1.314, time/batch = 0.167\n",
            "13283/17600 (epoch 37), train_loss = 1.316, time/batch = 0.173\n",
            "13284/17600 (epoch 37), train_loss = 1.295, time/batch = 0.163\n",
            "13285/17600 (epoch 37), train_loss = 1.300, time/batch = 0.147\n",
            "13286/17600 (epoch 37), train_loss = 1.283, time/batch = 0.162\n",
            "13287/17600 (epoch 37), train_loss = 1.240, time/batch = 0.150\n",
            "13288/17600 (epoch 37), train_loss = 1.290, time/batch = 0.172\n",
            "13289/17600 (epoch 37), train_loss = 1.297, time/batch = 0.178\n",
            "13290/17600 (epoch 37), train_loss = 1.251, time/batch = 0.172\n",
            "13291/17600 (epoch 37), train_loss = 1.254, time/batch = 0.154\n",
            "13292/17600 (epoch 37), train_loss = 1.288, time/batch = 0.160\n",
            "13293/17600 (epoch 37), train_loss = 1.322, time/batch = 0.144\n",
            "13294/17600 (epoch 37), train_loss = 1.246, time/batch = 0.163\n",
            "13295/17600 (epoch 37), train_loss = 1.314, time/batch = 0.161\n",
            "13296/17600 (epoch 37), train_loss = 1.280, time/batch = 0.166\n",
            "13297/17600 (epoch 37), train_loss = 1.303, time/batch = 0.165\n",
            "13298/17600 (epoch 37), train_loss = 1.289, time/batch = 0.150\n",
            "13299/17600 (epoch 37), train_loss = 1.344, time/batch = 0.165\n",
            "13300/17600 (epoch 37), train_loss = 1.308, time/batch = 0.162\n",
            "13301/17600 (epoch 37), train_loss = 1.335, time/batch = 0.160\n",
            "13302/17600 (epoch 37), train_loss = 1.302, time/batch = 0.172\n",
            "13303/17600 (epoch 37), train_loss = 1.307, time/batch = 0.163\n",
            "13304/17600 (epoch 37), train_loss = 1.330, time/batch = 0.156\n",
            "13305/17600 (epoch 37), train_loss = 1.376, time/batch = 0.164\n",
            "13306/17600 (epoch 37), train_loss = 1.347, time/batch = 0.159\n",
            "13307/17600 (epoch 37), train_loss = 1.346, time/batch = 0.168\n",
            "13308/17600 (epoch 37), train_loss = 1.366, time/batch = 0.170\n",
            "13309/17600 (epoch 37), train_loss = 1.377, time/batch = 0.154\n",
            "13310/17600 (epoch 37), train_loss = 1.320, time/batch = 0.152\n",
            "13311/17600 (epoch 37), train_loss = 1.367, time/batch = 0.159\n",
            "13312/17600 (epoch 37), train_loss = 1.305, time/batch = 0.148\n",
            "13313/17600 (epoch 37), train_loss = 1.303, time/batch = 0.169\n",
            "13314/17600 (epoch 37), train_loss = 1.368, time/batch = 0.158\n",
            "13315/17600 (epoch 37), train_loss = 1.273, time/batch = 0.185\n",
            "13316/17600 (epoch 37), train_loss = 1.268, time/batch = 0.150\n",
            "13317/17600 (epoch 37), train_loss = 1.341, time/batch = 0.171\n",
            "13318/17600 (epoch 37), train_loss = 1.292, time/batch = 0.163\n",
            "13319/17600 (epoch 37), train_loss = 1.351, time/batch = 0.177\n",
            "13320/17600 (epoch 37), train_loss = 1.296, time/batch = 0.154\n",
            "13321/17600 (epoch 37), train_loss = 1.326, time/batch = 0.183\n",
            "13322/17600 (epoch 37), train_loss = 1.307, time/batch = 0.148\n",
            "13323/17600 (epoch 37), train_loss = 1.323, time/batch = 0.168\n",
            "13324/17600 (epoch 37), train_loss = 1.354, time/batch = 0.157\n",
            "13325/17600 (epoch 37), train_loss = 1.309, time/batch = 0.164\n",
            "13326/17600 (epoch 37), train_loss = 1.379, time/batch = 0.157\n",
            "13327/17600 (epoch 37), train_loss = 1.374, time/batch = 0.190\n",
            "13328/17600 (epoch 37), train_loss = 1.383, time/batch = 0.158\n",
            "13329/17600 (epoch 37), train_loss = 1.296, time/batch = 0.165\n",
            "13330/17600 (epoch 37), train_loss = 1.335, time/batch = 0.165\n",
            "13331/17600 (epoch 37), train_loss = 1.338, time/batch = 0.167\n",
            "13332/17600 (epoch 37), train_loss = 1.340, time/batch = 0.155\n",
            "13333/17600 (epoch 37), train_loss = 1.340, time/batch = 0.182\n",
            "13334/17600 (epoch 37), train_loss = 1.377, time/batch = 0.156\n",
            "13335/17600 (epoch 37), train_loss = 1.325, time/batch = 0.174\n",
            "13336/17600 (epoch 37), train_loss = 1.356, time/batch = 0.163\n",
            "13337/17600 (epoch 37), train_loss = 1.294, time/batch = 0.186\n",
            "13338/17600 (epoch 37), train_loss = 1.310, time/batch = 0.257\n",
            "13339/17600 (epoch 37), train_loss = 1.276, time/batch = 0.286\n",
            "13340/17600 (epoch 37), train_loss = 1.375, time/batch = 0.290\n",
            "13341/17600 (epoch 37), train_loss = 1.361, time/batch = 0.285\n",
            "13342/17600 (epoch 37), train_loss = 1.361, time/batch = 0.258\n",
            "13343/17600 (epoch 37), train_loss = 1.341, time/batch = 0.296\n",
            "13344/17600 (epoch 37), train_loss = 1.383, time/batch = 0.267\n",
            "13345/17600 (epoch 37), train_loss = 1.349, time/batch = 0.257\n",
            "13346/17600 (epoch 37), train_loss = 1.404, time/batch = 0.266\n",
            "13347/17600 (epoch 37), train_loss = 1.372, time/batch = 0.294\n",
            "13348/17600 (epoch 37), train_loss = 1.354, time/batch = 0.174\n",
            "13349/17600 (epoch 37), train_loss = 1.307, time/batch = 0.172\n",
            "13350/17600 (epoch 37), train_loss = 1.381, time/batch = 0.150\n",
            "13351/17600 (epoch 37), train_loss = 1.363, time/batch = 0.164\n",
            "13352/17600 (epoch 37), train_loss = 1.399, time/batch = 0.161\n",
            "13353/17600 (epoch 37), train_loss = 1.366, time/batch = 0.156\n",
            "13354/17600 (epoch 37), train_loss = 1.291, time/batch = 0.144\n",
            "13355/17600 (epoch 37), train_loss = 1.317, time/batch = 0.171\n",
            "13356/17600 (epoch 37), train_loss = 1.368, time/batch = 0.160\n",
            "13357/17600 (epoch 37), train_loss = 1.375, time/batch = 0.161\n",
            "13358/17600 (epoch 37), train_loss = 1.331, time/batch = 0.185\n",
            "13359/17600 (epoch 37), train_loss = 1.324, time/batch = 0.166\n",
            "13360/17600 (epoch 37), train_loss = 1.334, time/batch = 0.166\n",
            "13361/17600 (epoch 37), train_loss = 1.368, time/batch = 0.177\n",
            "13362/17600 (epoch 37), train_loss = 1.318, time/batch = 0.152\n",
            "13363/17600 (epoch 37), train_loss = 1.349, time/batch = 0.170\n",
            "13364/17600 (epoch 37), train_loss = 1.314, time/batch = 0.168\n",
            "13365/17600 (epoch 37), train_loss = 1.356, time/batch = 0.165\n",
            "13366/17600 (epoch 37), train_loss = 1.367, time/batch = 0.155\n",
            "13367/17600 (epoch 37), train_loss = 1.331, time/batch = 0.161\n",
            "13368/17600 (epoch 37), train_loss = 1.310, time/batch = 0.152\n",
            "13369/17600 (epoch 37), train_loss = 1.327, time/batch = 0.157\n",
            "13370/17600 (epoch 37), train_loss = 1.349, time/batch = 0.170\n",
            "13371/17600 (epoch 37), train_loss = 1.322, time/batch = 0.172\n",
            "13372/17600 (epoch 37), train_loss = 1.346, time/batch = 0.146\n",
            "13373/17600 (epoch 37), train_loss = 1.330, time/batch = 0.162\n",
            "13374/17600 (epoch 37), train_loss = 1.322, time/batch = 0.150\n",
            "13375/17600 (epoch 37), train_loss = 1.315, time/batch = 0.149\n",
            "13376/17600 (epoch 38), train_loss = 1.134, time/batch = 0.172\n",
            "13377/17600 (epoch 38), train_loss = 1.348, time/batch = 0.155\n",
            "13378/17600 (epoch 38), train_loss = 1.339, time/batch = 0.161\n",
            "13379/17600 (epoch 38), train_loss = 1.282, time/batch = 0.158\n",
            "13380/17600 (epoch 38), train_loss = 1.377, time/batch = 0.166\n",
            "13381/17600 (epoch 38), train_loss = 1.353, time/batch = 0.148\n",
            "13382/17600 (epoch 38), train_loss = 1.304, time/batch = 0.185\n",
            "13383/17600 (epoch 38), train_loss = 1.289, time/batch = 0.159\n",
            "13384/17600 (epoch 38), train_loss = 1.329, time/batch = 0.177\n",
            "13385/17600 (epoch 38), train_loss = 1.385, time/batch = 0.161\n",
            "13386/17600 (epoch 38), train_loss = 1.312, time/batch = 0.164\n",
            "13387/17600 (epoch 38), train_loss = 1.328, time/batch = 0.153\n",
            "13388/17600 (epoch 38), train_loss = 1.308, time/batch = 0.186\n",
            "13389/17600 (epoch 38), train_loss = 1.292, time/batch = 0.160\n",
            "13390/17600 (epoch 38), train_loss = 1.272, time/batch = 0.162\n",
            "13391/17600 (epoch 38), train_loss = 1.337, time/batch = 0.161\n",
            "13392/17600 (epoch 38), train_loss = 1.348, time/batch = 0.160\n",
            "13393/17600 (epoch 38), train_loss = 1.303, time/batch = 0.152\n",
            "13394/17600 (epoch 38), train_loss = 1.318, time/batch = 0.170\n",
            "13395/17600 (epoch 38), train_loss = 1.360, time/batch = 0.167\n",
            "13396/17600 (epoch 38), train_loss = 1.283, time/batch = 0.158\n",
            "13397/17600 (epoch 38), train_loss = 1.337, time/batch = 0.157\n",
            "13398/17600 (epoch 38), train_loss = 1.303, time/batch = 0.160\n",
            "13399/17600 (epoch 38), train_loss = 1.278, time/batch = 0.162\n",
            "13400/17600 (epoch 38), train_loss = 1.363, time/batch = 0.164\n",
            "13401/17600 (epoch 38), train_loss = 1.292, time/batch = 0.180\n",
            "13402/17600 (epoch 38), train_loss = 1.341, time/batch = 0.164\n",
            "13403/17600 (epoch 38), train_loss = 1.322, time/batch = 0.158\n",
            "13404/17600 (epoch 38), train_loss = 1.301, time/batch = 0.166\n",
            "13405/17600 (epoch 38), train_loss = 1.233, time/batch = 0.152\n",
            "13406/17600 (epoch 38), train_loss = 1.342, time/batch = 0.163\n",
            "13407/17600 (epoch 38), train_loss = 1.323, time/batch = 0.179\n",
            "13408/17600 (epoch 38), train_loss = 1.299, time/batch = 0.302\n",
            "13409/17600 (epoch 38), train_loss = 1.304, time/batch = 0.254\n",
            "13410/17600 (epoch 38), train_loss = 1.306, time/batch = 0.275\n",
            "13411/17600 (epoch 38), train_loss = 1.320, time/batch = 0.289\n",
            "13412/17600 (epoch 38), train_loss = 1.345, time/batch = 0.267\n",
            "13413/17600 (epoch 38), train_loss = 1.342, time/batch = 0.284\n",
            "13414/17600 (epoch 38), train_loss = 1.323, time/batch = 0.296\n",
            "13415/17600 (epoch 38), train_loss = 1.300, time/batch = 0.296\n",
            "13416/17600 (epoch 38), train_loss = 1.349, time/batch = 0.271\n",
            "13417/17600 (epoch 38), train_loss = 1.282, time/batch = 0.295\n",
            "13418/17600 (epoch 38), train_loss = 1.338, time/batch = 0.162\n",
            "13419/17600 (epoch 38), train_loss = 1.258, time/batch = 0.177\n",
            "13420/17600 (epoch 38), train_loss = 1.328, time/batch = 0.154\n",
            "13421/17600 (epoch 38), train_loss = 1.298, time/batch = 0.164\n",
            "13422/17600 (epoch 38), train_loss = 1.255, time/batch = 0.153\n",
            "13423/17600 (epoch 38), train_loss = 1.314, time/batch = 0.166\n",
            "13424/17600 (epoch 38), train_loss = 1.264, time/batch = 0.145\n",
            "13425/17600 (epoch 38), train_loss = 1.283, time/batch = 0.181\n",
            "13426/17600 (epoch 38), train_loss = 1.304, time/batch = 0.153\n",
            "13427/17600 (epoch 38), train_loss = 1.296, time/batch = 0.162\n",
            "13428/17600 (epoch 38), train_loss = 1.264, time/batch = 0.160\n",
            "13429/17600 (epoch 38), train_loss = 1.293, time/batch = 0.175\n",
            "13430/17600 (epoch 38), train_loss = 1.339, time/batch = 0.157\n",
            "13431/17600 (epoch 38), train_loss = 1.317, time/batch = 0.180\n",
            "13432/17600 (epoch 38), train_loss = 1.319, time/batch = 0.169\n",
            "13433/17600 (epoch 38), train_loss = 1.312, time/batch = 0.162\n",
            "13434/17600 (epoch 38), train_loss = 1.292, time/batch = 0.145\n",
            "13435/17600 (epoch 38), train_loss = 1.317, time/batch = 0.163\n",
            "13436/17600 (epoch 38), train_loss = 1.313, time/batch = 0.154\n",
            "13437/17600 (epoch 38), train_loss = 1.255, time/batch = 0.167\n",
            "13438/17600 (epoch 38), train_loss = 1.258, time/batch = 0.178\n",
            "13439/17600 (epoch 38), train_loss = 1.277, time/batch = 0.169\n",
            "13440/17600 (epoch 38), train_loss = 1.308, time/batch = 0.174\n",
            "13441/17600 (epoch 38), train_loss = 1.275, time/batch = 0.194\n",
            "13442/17600 (epoch 38), train_loss = 1.312, time/batch = 0.162\n",
            "13443/17600 (epoch 38), train_loss = 1.269, time/batch = 0.180\n",
            "13444/17600 (epoch 38), train_loss = 1.317, time/batch = 0.158\n",
            "13445/17600 (epoch 38), train_loss = 1.248, time/batch = 0.173\n",
            "13446/17600 (epoch 38), train_loss = 1.296, time/batch = 0.165\n",
            "13447/17600 (epoch 38), train_loss = 1.339, time/batch = 0.163\n",
            "13448/17600 (epoch 38), train_loss = 1.300, time/batch = 0.164\n",
            "13449/17600 (epoch 38), train_loss = 1.372, time/batch = 0.164\n",
            "13450/17600 (epoch 38), train_loss = 1.301, time/batch = 0.187\n",
            "13451/17600 (epoch 38), train_loss = 1.305, time/batch = 0.150\n",
            "13452/17600 (epoch 38), train_loss = 1.259, time/batch = 0.152\n",
            "13453/17600 (epoch 38), train_loss = 1.314, time/batch = 0.158\n",
            "13454/17600 (epoch 38), train_loss = 1.322, time/batch = 0.156\n",
            "13455/17600 (epoch 38), train_loss = 1.396, time/batch = 0.152\n",
            "13456/17600 (epoch 38), train_loss = 1.365, time/batch = 0.181\n",
            "13457/17600 (epoch 38), train_loss = 1.359, time/batch = 0.155\n",
            "13458/17600 (epoch 38), train_loss = 1.340, time/batch = 0.171\n",
            "13459/17600 (epoch 38), train_loss = 1.311, time/batch = 0.150\n",
            "13460/17600 (epoch 38), train_loss = 1.288, time/batch = 0.169\n",
            "13461/17600 (epoch 38), train_loss = 1.323, time/batch = 0.167\n",
            "13462/17600 (epoch 38), train_loss = 1.314, time/batch = 0.192\n",
            "13463/17600 (epoch 38), train_loss = 1.254, time/batch = 0.159\n",
            "13464/17600 (epoch 38), train_loss = 1.293, time/batch = 0.158\n",
            "13465/17600 (epoch 38), train_loss = 1.234, time/batch = 0.163\n",
            "13466/17600 (epoch 38), train_loss = 1.299, time/batch = 0.171\n",
            "13467/17600 (epoch 38), train_loss = 1.285, time/batch = 0.155\n",
            "13468/17600 (epoch 38), train_loss = 1.263, time/batch = 0.185\n",
            "13469/17600 (epoch 38), train_loss = 1.317, time/batch = 0.162\n",
            "13470/17600 (epoch 38), train_loss = 1.301, time/batch = 0.165\n",
            "13471/17600 (epoch 38), train_loss = 1.265, time/batch = 0.148\n",
            "13472/17600 (epoch 38), train_loss = 1.240, time/batch = 0.174\n",
            "13473/17600 (epoch 38), train_loss = 1.290, time/batch = 0.163\n",
            "13474/17600 (epoch 38), train_loss = 1.323, time/batch = 0.171\n",
            "13475/17600 (epoch 38), train_loss = 1.282, time/batch = 0.163\n",
            "13476/17600 (epoch 38), train_loss = 1.255, time/batch = 0.176\n",
            "13477/17600 (epoch 38), train_loss = 1.340, time/batch = 0.159\n",
            "13478/17600 (epoch 38), train_loss = 1.281, time/batch = 0.214\n",
            "13479/17600 (epoch 38), train_loss = 1.390, time/batch = 0.277\n",
            "13480/17600 (epoch 38), train_loss = 1.316, time/batch = 0.265\n",
            "13481/17600 (epoch 38), train_loss = 1.290, time/batch = 0.288\n",
            "13482/17600 (epoch 38), train_loss = 1.331, time/batch = 0.260\n",
            "13483/17600 (epoch 38), train_loss = 1.317, time/batch = 0.271\n",
            "13484/17600 (epoch 38), train_loss = 1.321, time/batch = 0.280\n",
            "13485/17600 (epoch 38), train_loss = 1.306, time/batch = 0.296\n",
            "13486/17600 (epoch 38), train_loss = 1.262, time/batch = 0.311\n",
            "13487/17600 (epoch 38), train_loss = 1.285, time/batch = 0.291\n",
            "13488/17600 (epoch 38), train_loss = 1.315, time/batch = 0.256\n",
            "13489/17600 (epoch 38), train_loss = 1.209, time/batch = 0.160\n",
            "13490/17600 (epoch 38), train_loss = 1.284, time/batch = 0.152\n",
            "13491/17600 (epoch 38), train_loss = 1.315, time/batch = 0.168\n",
            "13492/17600 (epoch 38), train_loss = 1.284, time/batch = 0.156\n",
            "13493/17600 (epoch 38), train_loss = 1.309, time/batch = 0.198\n",
            "13494/17600 (epoch 38), train_loss = 1.252, time/batch = 0.152\n",
            "13495/17600 (epoch 38), train_loss = 1.355, time/batch = 0.160\n",
            "13496/17600 (epoch 38), train_loss = 1.261, time/batch = 0.145\n",
            "13497/17600 (epoch 38), train_loss = 1.322, time/batch = 0.179\n",
            "13498/17600 (epoch 38), train_loss = 1.347, time/batch = 0.151\n",
            "13499/17600 (epoch 38), train_loss = 1.334, time/batch = 0.179\n",
            "13500/17600 (epoch 38), train_loss = 1.249, time/batch = 0.148\n",
            "13501/17600 (epoch 38), train_loss = 1.315, time/batch = 0.167\n",
            "13502/17600 (epoch 38), train_loss = 1.313, time/batch = 0.157\n",
            "13503/17600 (epoch 38), train_loss = 1.299, time/batch = 0.167\n",
            "13504/17600 (epoch 38), train_loss = 1.281, time/batch = 0.165\n",
            "13505/17600 (epoch 38), train_loss = 1.315, time/batch = 0.173\n",
            "13506/17600 (epoch 38), train_loss = 1.328, time/batch = 0.160\n",
            "13507/17600 (epoch 38), train_loss = 1.332, time/batch = 0.161\n",
            "13508/17600 (epoch 38), train_loss = 1.301, time/batch = 0.161\n",
            "13509/17600 (epoch 38), train_loss = 1.274, time/batch = 0.183\n",
            "13510/17600 (epoch 38), train_loss = 1.284, time/batch = 0.158\n",
            "13511/17600 (epoch 38), train_loss = 1.256, time/batch = 0.167\n",
            "13512/17600 (epoch 38), train_loss = 1.326, time/batch = 0.155\n",
            "13513/17600 (epoch 38), train_loss = 1.349, time/batch = 0.160\n",
            "13514/17600 (epoch 38), train_loss = 1.295, time/batch = 0.150\n",
            "13515/17600 (epoch 38), train_loss = 1.345, time/batch = 0.178\n",
            "13516/17600 (epoch 38), train_loss = 1.375, time/batch = 0.162\n",
            "13517/17600 (epoch 38), train_loss = 1.284, time/batch = 0.165\n",
            "13518/17600 (epoch 38), train_loss = 1.310, time/batch = 0.180\n",
            "13519/17600 (epoch 38), train_loss = 1.275, time/batch = 0.162\n",
            "13520/17600 (epoch 38), train_loss = 1.330, time/batch = 0.157\n",
            "13521/17600 (epoch 38), train_loss = 1.312, time/batch = 0.164\n",
            "13522/17600 (epoch 38), train_loss = 1.288, time/batch = 0.149\n",
            "13523/17600 (epoch 38), train_loss = 1.296, time/batch = 0.157\n",
            "13524/17600 (epoch 38), train_loss = 1.279, time/batch = 0.184\n",
            "13525/17600 (epoch 38), train_loss = 1.342, time/batch = 0.165\n",
            "13526/17600 (epoch 38), train_loss = 1.230, time/batch = 0.144\n",
            "13527/17600 (epoch 38), train_loss = 1.302, time/batch = 0.169\n",
            "13528/17600 (epoch 38), train_loss = 1.328, time/batch = 0.150\n",
            "13529/17600 (epoch 38), train_loss = 1.295, time/batch = 0.170\n",
            "13530/17600 (epoch 38), train_loss = 1.313, time/batch = 0.163\n",
            "13531/17600 (epoch 38), train_loss = 1.249, time/batch = 0.156\n",
            "13532/17600 (epoch 38), train_loss = 1.247, time/batch = 0.165\n",
            "13533/17600 (epoch 38), train_loss = 1.265, time/batch = 0.161\n",
            "13534/17600 (epoch 38), train_loss = 1.326, time/batch = 0.169\n",
            "13535/17600 (epoch 38), train_loss = 1.370, time/batch = 0.165\n",
            "13536/17600 (epoch 38), train_loss = 1.324, time/batch = 0.189\n",
            "13537/17600 (epoch 38), train_loss = 1.325, time/batch = 0.167\n",
            "13538/17600 (epoch 38), train_loss = 1.255, time/batch = 0.167\n",
            "13539/17600 (epoch 38), train_loss = 1.285, time/batch = 0.150\n",
            "13540/17600 (epoch 38), train_loss = 1.306, time/batch = 0.170\n",
            "13541/17600 (epoch 38), train_loss = 1.293, time/batch = 0.165\n",
            "13542/17600 (epoch 38), train_loss = 1.306, time/batch = 0.173\n",
            "13543/17600 (epoch 38), train_loss = 1.333, time/batch = 0.173\n",
            "13544/17600 (epoch 38), train_loss = 1.323, time/batch = 0.163\n",
            "13545/17600 (epoch 38), train_loss = 1.303, time/batch = 0.158\n",
            "13546/17600 (epoch 38), train_loss = 1.239, time/batch = 0.171\n",
            "13547/17600 (epoch 38), train_loss = 1.250, time/batch = 0.176\n",
            "13548/17600 (epoch 38), train_loss = 1.339, time/batch = 0.166\n",
            "13549/17600 (epoch 38), train_loss = 1.306, time/batch = 0.251\n",
            "13550/17600 (epoch 38), train_loss = 1.297, time/batch = 0.274\n",
            "13551/17600 (epoch 38), train_loss = 1.300, time/batch = 0.263\n",
            "13552/17600 (epoch 38), train_loss = 1.216, time/batch = 0.288\n",
            "13553/17600 (epoch 38), train_loss = 1.279, time/batch = 0.282\n",
            "13554/17600 (epoch 38), train_loss = 1.279, time/batch = 0.281\n",
            "13555/17600 (epoch 38), train_loss = 1.256, time/batch = 0.281\n",
            "13556/17600 (epoch 38), train_loss = 1.255, time/batch = 0.278\n",
            "13557/17600 (epoch 38), train_loss = 1.308, time/batch = 0.261\n",
            "13558/17600 (epoch 38), train_loss = 1.365, time/batch = 0.275\n",
            "13559/17600 (epoch 38), train_loss = 1.281, time/batch = 0.239\n",
            "13560/17600 (epoch 38), train_loss = 1.282, time/batch = 0.186\n",
            "13561/17600 (epoch 38), train_loss = 1.245, time/batch = 0.173\n",
            "13562/17600 (epoch 38), train_loss = 1.322, time/batch = 0.165\n",
            "13563/17600 (epoch 38), train_loss = 1.330, time/batch = 0.155\n",
            "13564/17600 (epoch 38), train_loss = 1.249, time/batch = 0.172\n",
            "13565/17600 (epoch 38), train_loss = 1.346, time/batch = 0.160\n",
            "13566/17600 (epoch 38), train_loss = 1.309, time/batch = 0.170\n",
            "13567/17600 (epoch 38), train_loss = 1.291, time/batch = 0.182\n",
            "13568/17600 (epoch 38), train_loss = 1.299, time/batch = 0.172\n",
            "13569/17600 (epoch 38), train_loss = 1.317, time/batch = 0.156\n",
            "13570/17600 (epoch 38), train_loss = 1.319, time/batch = 0.158\n",
            "13571/17600 (epoch 38), train_loss = 1.328, time/batch = 0.161\n",
            "13572/17600 (epoch 38), train_loss = 1.285, time/batch = 0.162\n",
            "13573/17600 (epoch 38), train_loss = 1.336, time/batch = 0.179\n",
            "13574/17600 (epoch 38), train_loss = 1.283, time/batch = 0.167\n",
            "13575/17600 (epoch 38), train_loss = 1.317, time/batch = 0.157\n",
            "13576/17600 (epoch 38), train_loss = 1.321, time/batch = 0.175\n",
            "13577/17600 (epoch 38), train_loss = 1.287, time/batch = 0.150\n",
            "13578/17600 (epoch 38), train_loss = 1.278, time/batch = 0.175\n",
            "13579/17600 (epoch 38), train_loss = 1.305, time/batch = 0.170\n",
            "13580/17600 (epoch 38), train_loss = 1.340, time/batch = 0.166\n",
            "13581/17600 (epoch 38), train_loss = 1.387, time/batch = 0.152\n",
            "13582/17600 (epoch 38), train_loss = 1.302, time/batch = 0.170\n",
            "13583/17600 (epoch 38), train_loss = 1.302, time/batch = 0.158\n",
            "13584/17600 (epoch 38), train_loss = 1.303, time/batch = 0.151\n",
            "13585/17600 (epoch 38), train_loss = 1.283, time/batch = 0.155\n",
            "13586/17600 (epoch 38), train_loss = 1.327, time/batch = 0.171\n",
            "13587/17600 (epoch 38), train_loss = 1.318, time/batch = 0.158\n",
            "13588/17600 (epoch 38), train_loss = 1.286, time/batch = 0.167\n",
            "13589/17600 (epoch 38), train_loss = 1.361, time/batch = 0.151\n",
            "13590/17600 (epoch 38), train_loss = 1.402, time/batch = 0.162\n",
            "13591/17600 (epoch 38), train_loss = 1.291, time/batch = 0.170\n",
            "13592/17600 (epoch 38), train_loss = 1.314, time/batch = 0.167\n",
            "13593/17600 (epoch 38), train_loss = 1.267, time/batch = 0.163\n",
            "13594/17600 (epoch 38), train_loss = 1.304, time/batch = 0.157\n",
            "13595/17600 (epoch 38), train_loss = 1.286, time/batch = 0.172\n",
            "13596/17600 (epoch 38), train_loss = 1.276, time/batch = 0.157\n",
            "13597/17600 (epoch 38), train_loss = 1.303, time/batch = 0.170\n",
            "13598/17600 (epoch 38), train_loss = 1.369, time/batch = 0.168\n",
            "13599/17600 (epoch 38), train_loss = 1.342, time/batch = 0.166\n",
            "13600/17600 (epoch 38), train_loss = 1.259, time/batch = 0.158\n",
            "13601/17600 (epoch 38), train_loss = 1.309, time/batch = 0.164\n",
            "13602/17600 (epoch 38), train_loss = 1.314, time/batch = 0.162\n",
            "13603/17600 (epoch 38), train_loss = 1.294, time/batch = 0.161\n",
            "13604/17600 (epoch 38), train_loss = 1.324, time/batch = 0.171\n",
            "13605/17600 (epoch 38), train_loss = 1.289, time/batch = 0.173\n",
            "13606/17600 (epoch 38), train_loss = 1.278, time/batch = 0.156\n",
            "13607/17600 (epoch 38), train_loss = 1.344, time/batch = 0.179\n",
            "13608/17600 (epoch 38), train_loss = 1.340, time/batch = 0.165\n",
            "13609/17600 (epoch 38), train_loss = 1.355, time/batch = 0.169\n",
            "13610/17600 (epoch 38), train_loss = 1.268, time/batch = 0.151\n",
            "13611/17600 (epoch 38), train_loss = 1.300, time/batch = 0.172\n",
            "13612/17600 (epoch 38), train_loss = 1.288, time/batch = 0.164\n",
            "13613/17600 (epoch 38), train_loss = 1.359, time/batch = 0.158\n",
            "13614/17600 (epoch 38), train_loss = 1.355, time/batch = 0.153\n",
            "13615/17600 (epoch 38), train_loss = 1.250, time/batch = 0.151\n",
            "13616/17600 (epoch 38), train_loss = 1.342, time/batch = 0.160\n",
            "13617/17600 (epoch 38), train_loss = 1.288, time/batch = 0.168\n",
            "13618/17600 (epoch 38), train_loss = 1.323, time/batch = 0.176\n",
            "13619/17600 (epoch 38), train_loss = 1.316, time/batch = 0.151\n",
            "13620/17600 (epoch 38), train_loss = 1.285, time/batch = 0.236\n",
            "13621/17600 (epoch 38), train_loss = 1.325, time/batch = 0.281\n",
            "13622/17600 (epoch 38), train_loss = 1.317, time/batch = 0.290\n",
            "13623/17600 (epoch 38), train_loss = 1.290, time/batch = 0.285\n",
            "13624/17600 (epoch 38), train_loss = 1.267, time/batch = 0.266\n",
            "13625/17600 (epoch 38), train_loss = 1.246, time/batch = 0.273\n",
            "13626/17600 (epoch 38), train_loss = 1.381, time/batch = 0.273\n",
            "13627/17600 (epoch 38), train_loss = 1.319, time/batch = 0.276\n",
            "13628/17600 (epoch 38), train_loss = 1.313, time/batch = 0.306\n",
            "13629/17600 (epoch 38), train_loss = 1.320, time/batch = 0.307\n",
            "13630/17600 (epoch 38), train_loss = 1.337, time/batch = 0.253\n",
            "13631/17600 (epoch 38), train_loss = 1.346, time/batch = 0.151\n",
            "13632/17600 (epoch 38), train_loss = 1.393, time/batch = 0.172\n",
            "13633/17600 (epoch 38), train_loss = 1.361, time/batch = 0.162\n",
            "13634/17600 (epoch 38), train_loss = 1.312, time/batch = 0.172\n",
            "13635/17600 (epoch 38), train_loss = 1.314, time/batch = 0.167\n",
            "13636/17600 (epoch 38), train_loss = 1.292, time/batch = 0.165\n",
            "13637/17600 (epoch 38), train_loss = 1.298, time/batch = 0.155\n",
            "13638/17600 (epoch 38), train_loss = 1.280, time/batch = 0.165\n",
            "13639/17600 (epoch 38), train_loss = 1.238, time/batch = 0.144\n",
            "13640/17600 (epoch 38), train_loss = 1.287, time/batch = 0.169\n",
            "13641/17600 (epoch 38), train_loss = 1.295, time/batch = 0.181\n",
            "13642/17600 (epoch 38), train_loss = 1.248, time/batch = 0.168\n",
            "13643/17600 (epoch 38), train_loss = 1.252, time/batch = 0.150\n",
            "13644/17600 (epoch 38), train_loss = 1.285, time/batch = 0.162\n",
            "13645/17600 (epoch 38), train_loss = 1.319, time/batch = 0.147\n",
            "13646/17600 (epoch 38), train_loss = 1.244, time/batch = 0.170\n",
            "13647/17600 (epoch 38), train_loss = 1.312, time/batch = 0.164\n",
            "13648/17600 (epoch 38), train_loss = 1.278, time/batch = 0.176\n",
            "13649/17600 (epoch 38), train_loss = 1.301, time/batch = 0.153\n",
            "13650/17600 (epoch 38), train_loss = 1.286, time/batch = 0.169\n",
            "13651/17600 (epoch 38), train_loss = 1.342, time/batch = 0.164\n",
            "13652/17600 (epoch 38), train_loss = 1.305, time/batch = 0.167\n",
            "13653/17600 (epoch 38), train_loss = 1.332, time/batch = 0.163\n",
            "13654/17600 (epoch 38), train_loss = 1.300, time/batch = 0.194\n",
            "13655/17600 (epoch 38), train_loss = 1.304, time/batch = 0.155\n",
            "13656/17600 (epoch 38), train_loss = 1.328, time/batch = 0.170\n",
            "13657/17600 (epoch 38), train_loss = 1.374, time/batch = 0.160\n",
            "13658/17600 (epoch 38), train_loss = 1.345, time/batch = 0.171\n",
            "13659/17600 (epoch 38), train_loss = 1.344, time/batch = 0.168\n",
            "13660/17600 (epoch 38), train_loss = 1.363, time/batch = 0.196\n",
            "13661/17600 (epoch 38), train_loss = 1.374, time/batch = 0.164\n",
            "13662/17600 (epoch 38), train_loss = 1.318, time/batch = 0.160\n",
            "13663/17600 (epoch 38), train_loss = 1.365, time/batch = 0.152\n",
            "13664/17600 (epoch 38), train_loss = 1.302, time/batch = 0.156\n",
            "13665/17600 (epoch 38), train_loss = 1.301, time/batch = 0.154\n",
            "13666/17600 (epoch 38), train_loss = 1.366, time/batch = 0.191\n",
            "13667/17600 (epoch 38), train_loss = 1.271, time/batch = 0.148\n",
            "13668/17600 (epoch 38), train_loss = 1.265, time/batch = 0.172\n",
            "13669/17600 (epoch 38), train_loss = 1.339, time/batch = 0.158\n",
            "13670/17600 (epoch 38), train_loss = 1.289, time/batch = 0.152\n",
            "13671/17600 (epoch 38), train_loss = 1.349, time/batch = 0.155\n",
            "13672/17600 (epoch 38), train_loss = 1.294, time/batch = 0.186\n",
            "13673/17600 (epoch 38), train_loss = 1.323, time/batch = 0.163\n",
            "13674/17600 (epoch 38), train_loss = 1.305, time/batch = 0.169\n",
            "13675/17600 (epoch 38), train_loss = 1.321, time/batch = 0.153\n",
            "13676/17600 (epoch 38), train_loss = 1.352, time/batch = 0.168\n",
            "13677/17600 (epoch 38), train_loss = 1.307, time/batch = 0.155\n",
            "13678/17600 (epoch 38), train_loss = 1.377, time/batch = 0.175\n",
            "13679/17600 (epoch 38), train_loss = 1.371, time/batch = 0.155\n",
            "13680/17600 (epoch 38), train_loss = 1.381, time/batch = 0.175\n",
            "13681/17600 (epoch 38), train_loss = 1.294, time/batch = 0.149\n",
            "13682/17600 (epoch 38), train_loss = 1.332, time/batch = 0.180\n",
            "13683/17600 (epoch 38), train_loss = 1.335, time/batch = 0.164\n",
            "13684/17600 (epoch 38), train_loss = 1.337, time/batch = 0.156\n",
            "13685/17600 (epoch 38), train_loss = 1.338, time/batch = 0.172\n",
            "13686/17600 (epoch 38), train_loss = 1.374, time/batch = 0.164\n",
            "13687/17600 (epoch 38), train_loss = 1.323, time/batch = 0.168\n",
            "13688/17600 (epoch 38), train_loss = 1.354, time/batch = 0.163\n",
            "13689/17600 (epoch 38), train_loss = 1.292, time/batch = 0.170\n",
            "13690/17600 (epoch 38), train_loss = 1.307, time/batch = 0.210\n",
            "13691/17600 (epoch 38), train_loss = 1.274, time/batch = 0.303\n",
            "13692/17600 (epoch 38), train_loss = 1.372, time/batch = 0.265\n",
            "13693/17600 (epoch 38), train_loss = 1.359, time/batch = 0.292\n",
            "13694/17600 (epoch 38), train_loss = 1.359, time/batch = 0.272\n",
            "13695/17600 (epoch 38), train_loss = 1.339, time/batch = 0.273\n",
            "13696/17600 (epoch 38), train_loss = 1.380, time/batch = 0.289\n",
            "13697/17600 (epoch 38), train_loss = 1.346, time/batch = 0.292\n",
            "13698/17600 (epoch 38), train_loss = 1.401, time/batch = 0.285\n",
            "13699/17600 (epoch 38), train_loss = 1.370, time/batch = 0.285\n",
            "13700/17600 (epoch 38), train_loss = 1.352, time/batch = 0.280\n",
            "13701/17600 (epoch 38), train_loss = 1.305, time/batch = 0.175\n",
            "13702/17600 (epoch 38), train_loss = 1.380, time/batch = 0.165\n",
            "13703/17600 (epoch 38), train_loss = 1.361, time/batch = 0.171\n",
            "13704/17600 (epoch 38), train_loss = 1.397, time/batch = 0.159\n",
            "13705/17600 (epoch 38), train_loss = 1.364, time/batch = 0.149\n",
            "13706/17600 (epoch 38), train_loss = 1.288, time/batch = 0.172\n",
            "13707/17600 (epoch 38), train_loss = 1.315, time/batch = 0.164\n",
            "13708/17600 (epoch 38), train_loss = 1.365, time/batch = 0.169\n",
            "13709/17600 (epoch 38), train_loss = 1.373, time/batch = 0.166\n",
            "13710/17600 (epoch 38), train_loss = 1.329, time/batch = 0.155\n",
            "13711/17600 (epoch 38), train_loss = 1.322, time/batch = 0.148\n",
            "13712/17600 (epoch 38), train_loss = 1.332, time/batch = 0.159\n",
            "13713/17600 (epoch 38), train_loss = 1.366, time/batch = 0.156\n",
            "13714/17600 (epoch 38), train_loss = 1.315, time/batch = 0.155\n",
            "13715/17600 (epoch 38), train_loss = 1.346, time/batch = 0.164\n",
            "13716/17600 (epoch 38), train_loss = 1.311, time/batch = 0.171\n",
            "13717/17600 (epoch 38), train_loss = 1.354, time/batch = 0.159\n",
            "13718/17600 (epoch 38), train_loss = 1.364, time/batch = 0.179\n",
            "13719/17600 (epoch 38), train_loss = 1.328, time/batch = 0.165\n",
            "13720/17600 (epoch 38), train_loss = 1.307, time/batch = 0.162\n",
            "13721/17600 (epoch 38), train_loss = 1.325, time/batch = 0.151\n",
            "13722/17600 (epoch 38), train_loss = 1.347, time/batch = 0.181\n",
            "13723/17600 (epoch 38), train_loss = 1.320, time/batch = 0.157\n",
            "13724/17600 (epoch 38), train_loss = 1.344, time/batch = 0.165\n",
            "13725/17600 (epoch 38), train_loss = 1.328, time/batch = 0.161\n",
            "13726/17600 (epoch 38), train_loss = 1.320, time/batch = 0.161\n",
            "13727/17600 (epoch 38), train_loss = 1.312, time/batch = 0.165\n",
            "13728/17600 (epoch 39), train_loss = 1.129, time/batch = 0.155\n",
            "13729/17600 (epoch 39), train_loss = 1.344, time/batch = 0.166\n",
            "13730/17600 (epoch 39), train_loss = 1.338, time/batch = 0.173\n",
            "13731/17600 (epoch 39), train_loss = 1.281, time/batch = 0.158\n",
            "13732/17600 (epoch 39), train_loss = 1.375, time/batch = 0.178\n",
            "13733/17600 (epoch 39), train_loss = 1.352, time/batch = 0.157\n",
            "13734/17600 (epoch 39), train_loss = 1.299, time/batch = 0.163\n",
            "13735/17600 (epoch 39), train_loss = 1.287, time/batch = 0.150\n",
            "13736/17600 (epoch 39), train_loss = 1.327, time/batch = 0.161\n",
            "13737/17600 (epoch 39), train_loss = 1.381, time/batch = 0.150\n",
            "13738/17600 (epoch 39), train_loss = 1.308, time/batch = 0.160\n",
            "13739/17600 (epoch 39), train_loss = 1.324, time/batch = 0.183\n",
            "13740/17600 (epoch 39), train_loss = 1.305, time/batch = 0.165\n",
            "13741/17600 (epoch 39), train_loss = 1.289, time/batch = 0.164\n",
            "13742/17600 (epoch 39), train_loss = 1.270, time/batch = 0.183\n",
            "13743/17600 (epoch 39), train_loss = 1.335, time/batch = 0.163\n",
            "13744/17600 (epoch 39), train_loss = 1.346, time/batch = 0.171\n",
            "13745/17600 (epoch 39), train_loss = 1.301, time/batch = 0.175\n",
            "13746/17600 (epoch 39), train_loss = 1.315, time/batch = 0.170\n",
            "13747/17600 (epoch 39), train_loss = 1.358, time/batch = 0.165\n",
            "13748/17600 (epoch 39), train_loss = 1.281, time/batch = 0.168\n",
            "13749/17600 (epoch 39), train_loss = 1.335, time/batch = 0.161\n",
            "13750/17600 (epoch 39), train_loss = 1.301, time/batch = 0.177\n",
            "13751/17600 (epoch 39), train_loss = 1.275, time/batch = 0.164\n",
            "13752/17600 (epoch 39), train_loss = 1.360, time/batch = 0.161\n",
            "13753/17600 (epoch 39), train_loss = 1.290, time/batch = 0.166\n",
            "13754/17600 (epoch 39), train_loss = 1.339, time/batch = 0.169\n",
            "13755/17600 (epoch 39), train_loss = 1.321, time/batch = 0.156\n",
            "13756/17600 (epoch 39), train_loss = 1.299, time/batch = 0.167\n",
            "13757/17600 (epoch 39), train_loss = 1.232, time/batch = 0.167\n",
            "13758/17600 (epoch 39), train_loss = 1.340, time/batch = 0.166\n",
            "13759/17600 (epoch 39), train_loss = 1.320, time/batch = 0.151\n",
            "13760/17600 (epoch 39), train_loss = 1.296, time/batch = 0.242\n",
            "13761/17600 (epoch 39), train_loss = 1.301, time/batch = 0.266\n",
            "13762/17600 (epoch 39), train_loss = 1.303, time/batch = 0.287\n",
            "13763/17600 (epoch 39), train_loss = 1.318, time/batch = 0.292\n",
            "13764/17600 (epoch 39), train_loss = 1.343, time/batch = 0.266\n",
            "13765/17600 (epoch 39), train_loss = 1.340, time/batch = 0.280\n",
            "13766/17600 (epoch 39), train_loss = 1.321, time/batch = 0.286\n",
            "13767/17600 (epoch 39), train_loss = 1.298, time/batch = 0.309\n",
            "13768/17600 (epoch 39), train_loss = 1.346, time/batch = 0.300\n",
            "13769/17600 (epoch 39), train_loss = 1.280, time/batch = 0.274\n",
            "13770/17600 (epoch 39), train_loss = 1.336, time/batch = 0.242\n",
            "13771/17600 (epoch 39), train_loss = 1.256, time/batch = 0.168\n",
            "13772/17600 (epoch 39), train_loss = 1.325, time/batch = 0.153\n",
            "13773/17600 (epoch 39), train_loss = 1.296, time/batch = 0.157\n",
            "13774/17600 (epoch 39), train_loss = 1.252, time/batch = 0.146\n",
            "13775/17600 (epoch 39), train_loss = 1.312, time/batch = 0.182\n",
            "13776/17600 (epoch 39), train_loss = 1.262, time/batch = 0.151\n",
            "13777/17600 (epoch 39), train_loss = 1.281, time/batch = 0.169\n",
            "13778/17600 (epoch 39), train_loss = 1.303, time/batch = 0.164\n",
            "13779/17600 (epoch 39), train_loss = 1.294, time/batch = 0.159\n",
            "13780/17600 (epoch 39), train_loss = 1.262, time/batch = 0.146\n",
            "13781/17600 (epoch 39), train_loss = 1.291, time/batch = 0.182\n",
            "13782/17600 (epoch 39), train_loss = 1.337, time/batch = 0.157\n",
            "13783/17600 (epoch 39), train_loss = 1.315, time/batch = 0.179\n",
            "13784/17600 (epoch 39), train_loss = 1.317, time/batch = 0.152\n",
            "13785/17600 (epoch 39), train_loss = 1.309, time/batch = 0.175\n",
            "13786/17600 (epoch 39), train_loss = 1.289, time/batch = 0.159\n",
            "13787/17600 (epoch 39), train_loss = 1.315, time/batch = 0.158\n",
            "13788/17600 (epoch 39), train_loss = 1.311, time/batch = 0.169\n",
            "13789/17600 (epoch 39), train_loss = 1.253, time/batch = 0.165\n",
            "13790/17600 (epoch 39), train_loss = 1.255, time/batch = 0.155\n",
            "13791/17600 (epoch 39), train_loss = 1.275, time/batch = 0.166\n",
            "13792/17600 (epoch 39), train_loss = 1.306, time/batch = 0.148\n",
            "13793/17600 (epoch 39), train_loss = 1.272, time/batch = 0.170\n",
            "13794/17600 (epoch 39), train_loss = 1.309, time/batch = 0.177\n",
            "13795/17600 (epoch 39), train_loss = 1.267, time/batch = 0.179\n",
            "13796/17600 (epoch 39), train_loss = 1.315, time/batch = 0.177\n",
            "13797/17600 (epoch 39), train_loss = 1.246, time/batch = 0.155\n",
            "13798/17600 (epoch 39), train_loss = 1.293, time/batch = 0.165\n",
            "13799/17600 (epoch 39), train_loss = 1.337, time/batch = 0.156\n",
            "13800/17600 (epoch 39), train_loss = 1.298, time/batch = 0.190\n",
            "13801/17600 (epoch 39), train_loss = 1.370, time/batch = 0.156\n",
            "13802/17600 (epoch 39), train_loss = 1.299, time/batch = 0.162\n",
            "13803/17600 (epoch 39), train_loss = 1.302, time/batch = 0.148\n",
            "13804/17600 (epoch 39), train_loss = 1.257, time/batch = 0.156\n",
            "13805/17600 (epoch 39), train_loss = 1.312, time/batch = 0.156\n",
            "13806/17600 (epoch 39), train_loss = 1.319, time/batch = 0.174\n",
            "13807/17600 (epoch 39), train_loss = 1.394, time/batch = 0.152\n",
            "13808/17600 (epoch 39), train_loss = 1.363, time/batch = 0.161\n",
            "13809/17600 (epoch 39), train_loss = 1.357, time/batch = 0.159\n",
            "13810/17600 (epoch 39), train_loss = 1.338, time/batch = 0.169\n",
            "13811/17600 (epoch 39), train_loss = 1.309, time/batch = 0.146\n",
            "13812/17600 (epoch 39), train_loss = 1.286, time/batch = 0.168\n",
            "13813/17600 (epoch 39), train_loss = 1.321, time/batch = 0.174\n",
            "13814/17600 (epoch 39), train_loss = 1.312, time/batch = 0.159\n",
            "13815/17600 (epoch 39), train_loss = 1.252, time/batch = 0.156\n",
            "13816/17600 (epoch 39), train_loss = 1.291, time/batch = 0.154\n",
            "13817/17600 (epoch 39), train_loss = 1.232, time/batch = 0.145\n",
            "13818/17600 (epoch 39), train_loss = 1.296, time/batch = 0.167\n",
            "13819/17600 (epoch 39), train_loss = 1.283, time/batch = 0.168\n",
            "13820/17600 (epoch 39), train_loss = 1.261, time/batch = 0.166\n",
            "13821/17600 (epoch 39), train_loss = 1.315, time/batch = 0.157\n",
            "13822/17600 (epoch 39), train_loss = 1.298, time/batch = 0.170\n",
            "13823/17600 (epoch 39), train_loss = 1.262, time/batch = 0.163\n",
            "13824/17600 (epoch 39), train_loss = 1.238, time/batch = 0.170\n",
            "13825/17600 (epoch 39), train_loss = 1.287, time/batch = 0.170\n",
            "13826/17600 (epoch 39), train_loss = 1.321, time/batch = 0.166\n",
            "13827/17600 (epoch 39), train_loss = 1.280, time/batch = 0.164\n",
            "13828/17600 (epoch 39), train_loss = 1.253, time/batch = 0.177\n",
            "13829/17600 (epoch 39), train_loss = 1.338, time/batch = 0.163\n",
            "13830/17600 (epoch 39), train_loss = 1.278, time/batch = 0.160\n",
            "13831/17600 (epoch 39), train_loss = 1.388, time/batch = 0.224\n",
            "13832/17600 (epoch 39), train_loss = 1.314, time/batch = 0.282\n",
            "13833/17600 (epoch 39), train_loss = 1.288, time/batch = 0.279\n",
            "13834/17600 (epoch 39), train_loss = 1.330, time/batch = 0.268\n",
            "13835/17600 (epoch 39), train_loss = 1.315, time/batch = 0.312\n",
            "13836/17600 (epoch 39), train_loss = 1.319, time/batch = 0.282\n",
            "13837/17600 (epoch 39), train_loss = 1.303, time/batch = 0.283\n",
            "13838/17600 (epoch 39), train_loss = 1.259, time/batch = 0.284\n",
            "13839/17600 (epoch 39), train_loss = 1.282, time/batch = 0.317\n",
            "13840/17600 (epoch 39), train_loss = 1.313, time/batch = 0.310\n",
            "13841/17600 (epoch 39), train_loss = 1.207, time/batch = 0.183\n",
            "13842/17600 (epoch 39), train_loss = 1.282, time/batch = 0.164\n",
            "13843/17600 (epoch 39), train_loss = 1.313, time/batch = 0.160\n",
            "13844/17600 (epoch 39), train_loss = 1.282, time/batch = 0.176\n",
            "13845/17600 (epoch 39), train_loss = 1.307, time/batch = 0.148\n",
            "13846/17600 (epoch 39), train_loss = 1.250, time/batch = 0.159\n",
            "13847/17600 (epoch 39), train_loss = 1.352, time/batch = 0.152\n",
            "13848/17600 (epoch 39), train_loss = 1.259, time/batch = 0.179\n",
            "13849/17600 (epoch 39), train_loss = 1.320, time/batch = 0.164\n",
            "13850/17600 (epoch 39), train_loss = 1.345, time/batch = 0.173\n",
            "13851/17600 (epoch 39), train_loss = 1.331, time/batch = 0.158\n",
            "13852/17600 (epoch 39), train_loss = 1.247, time/batch = 0.167\n",
            "13853/17600 (epoch 39), train_loss = 1.313, time/batch = 0.154\n",
            "13854/17600 (epoch 39), train_loss = 1.310, time/batch = 0.167\n",
            "13855/17600 (epoch 39), train_loss = 1.297, time/batch = 0.152\n",
            "13856/17600 (epoch 39), train_loss = 1.279, time/batch = 0.183\n",
            "13857/17600 (epoch 39), train_loss = 1.312, time/batch = 0.143\n",
            "13858/17600 (epoch 39), train_loss = 1.326, time/batch = 0.157\n",
            "13859/17600 (epoch 39), train_loss = 1.329, time/batch = 0.166\n",
            "13860/17600 (epoch 39), train_loss = 1.299, time/batch = 0.172\n",
            "13861/17600 (epoch 39), train_loss = 1.271, time/batch = 0.166\n",
            "13862/17600 (epoch 39), train_loss = 1.281, time/batch = 0.193\n",
            "13863/17600 (epoch 39), train_loss = 1.253, time/batch = 0.161\n",
            "13864/17600 (epoch 39), train_loss = 1.323, time/batch = 0.153\n",
            "13865/17600 (epoch 39), train_loss = 1.347, time/batch = 0.171\n",
            "13866/17600 (epoch 39), train_loss = 1.292, time/batch = 0.152\n",
            "13867/17600 (epoch 39), train_loss = 1.344, time/batch = 0.163\n",
            "13868/17600 (epoch 39), train_loss = 1.372, time/batch = 0.156\n",
            "13869/17600 (epoch 39), train_loss = 1.282, time/batch = 0.185\n",
            "13870/17600 (epoch 39), train_loss = 1.308, time/batch = 0.146\n",
            "13871/17600 (epoch 39), train_loss = 1.273, time/batch = 0.159\n",
            "13872/17600 (epoch 39), train_loss = 1.327, time/batch = 0.162\n",
            "13873/17600 (epoch 39), train_loss = 1.310, time/batch = 0.165\n",
            "13874/17600 (epoch 39), train_loss = 1.286, time/batch = 0.151\n",
            "13875/17600 (epoch 39), train_loss = 1.294, time/batch = 0.179\n",
            "13876/17600 (epoch 39), train_loss = 1.276, time/batch = 0.145\n",
            "13877/17600 (epoch 39), train_loss = 1.339, time/batch = 0.147\n",
            "13878/17600 (epoch 39), train_loss = 1.227, time/batch = 0.175\n",
            "13879/17600 (epoch 39), train_loss = 1.300, time/batch = 0.164\n",
            "13880/17600 (epoch 39), train_loss = 1.325, time/batch = 0.166\n",
            "13881/17600 (epoch 39), train_loss = 1.293, time/batch = 0.164\n",
            "13882/17600 (epoch 39), train_loss = 1.311, time/batch = 0.168\n",
            "13883/17600 (epoch 39), train_loss = 1.247, time/batch = 0.160\n",
            "13884/17600 (epoch 39), train_loss = 1.245, time/batch = 0.169\n",
            "13885/17600 (epoch 39), train_loss = 1.263, time/batch = 0.153\n",
            "13886/17600 (epoch 39), train_loss = 1.324, time/batch = 0.163\n",
            "13887/17600 (epoch 39), train_loss = 1.368, time/batch = 0.159\n",
            "13888/17600 (epoch 39), train_loss = 1.322, time/batch = 0.175\n",
            "13889/17600 (epoch 39), train_loss = 1.323, time/batch = 0.166\n",
            "13890/17600 (epoch 39), train_loss = 1.253, time/batch = 0.163\n",
            "13891/17600 (epoch 39), train_loss = 1.283, time/batch = 0.152\n",
            "13892/17600 (epoch 39), train_loss = 1.304, time/batch = 0.162\n",
            "13893/17600 (epoch 39), train_loss = 1.291, time/batch = 0.153\n",
            "13894/17600 (epoch 39), train_loss = 1.304, time/batch = 0.174\n",
            "13895/17600 (epoch 39), train_loss = 1.330, time/batch = 0.156\n",
            "13896/17600 (epoch 39), train_loss = 1.321, time/batch = 0.157\n",
            "13897/17600 (epoch 39), train_loss = 1.301, time/batch = 0.163\n",
            "13898/17600 (epoch 39), train_loss = 1.236, time/batch = 0.153\n",
            "13899/17600 (epoch 39), train_loss = 1.248, time/batch = 0.170\n",
            "13900/17600 (epoch 39), train_loss = 1.336, time/batch = 0.174\n",
            "13901/17600 (epoch 39), train_loss = 1.304, time/batch = 0.161\n",
            "13902/17600 (epoch 39), train_loss = 1.295, time/batch = 0.211\n",
            "13903/17600 (epoch 39), train_loss = 1.298, time/batch = 0.262\n",
            "13904/17600 (epoch 39), train_loss = 1.213, time/batch = 0.235\n",
            "13905/17600 (epoch 39), train_loss = 1.276, time/batch = 0.273\n",
            "13906/17600 (epoch 39), train_loss = 1.276, time/batch = 0.287\n",
            "13907/17600 (epoch 39), train_loss = 1.254, time/batch = 0.274\n",
            "13908/17600 (epoch 39), train_loss = 1.253, time/batch = 0.264\n",
            "13909/17600 (epoch 39), train_loss = 1.306, time/batch = 0.279\n",
            "13910/17600 (epoch 39), train_loss = 1.363, time/batch = 0.276\n",
            "13911/17600 (epoch 39), train_loss = 1.279, time/batch = 0.284\n",
            "13912/17600 (epoch 39), train_loss = 1.279, time/batch = 0.263\n",
            "13913/17600 (epoch 39), train_loss = 1.243, time/batch = 0.190\n",
            "13914/17600 (epoch 39), train_loss = 1.319, time/batch = 0.176\n",
            "13915/17600 (epoch 39), train_loss = 1.328, time/batch = 0.160\n",
            "13916/17600 (epoch 39), train_loss = 1.247, time/batch = 0.156\n",
            "13917/17600 (epoch 39), train_loss = 1.343, time/batch = 0.150\n",
            "13918/17600 (epoch 39), train_loss = 1.307, time/batch = 0.161\n",
            "13919/17600 (epoch 39), train_loss = 1.289, time/batch = 0.167\n",
            "13920/17600 (epoch 39), train_loss = 1.297, time/batch = 0.163\n",
            "13921/17600 (epoch 39), train_loss = 1.315, time/batch = 0.144\n",
            "13922/17600 (epoch 39), train_loss = 1.316, time/batch = 0.160\n",
            "13923/17600 (epoch 39), train_loss = 1.326, time/batch = 0.157\n",
            "13924/17600 (epoch 39), train_loss = 1.283, time/batch = 0.154\n",
            "13925/17600 (epoch 39), train_loss = 1.334, time/batch = 0.155\n",
            "13926/17600 (epoch 39), train_loss = 1.281, time/batch = 0.173\n",
            "13927/17600 (epoch 39), train_loss = 1.315, time/batch = 0.153\n",
            "13928/17600 (epoch 39), train_loss = 1.318, time/batch = 0.161\n",
            "13929/17600 (epoch 39), train_loss = 1.284, time/batch = 0.149\n",
            "13930/17600 (epoch 39), train_loss = 1.275, time/batch = 0.163\n",
            "13931/17600 (epoch 39), train_loss = 1.303, time/batch = 0.153\n",
            "13932/17600 (epoch 39), train_loss = 1.338, time/batch = 0.191\n",
            "13933/17600 (epoch 39), train_loss = 1.385, time/batch = 0.192\n",
            "13934/17600 (epoch 39), train_loss = 1.300, time/batch = 0.177\n",
            "13935/17600 (epoch 39), train_loss = 1.300, time/batch = 0.160\n",
            "13936/17600 (epoch 39), train_loss = 1.301, time/batch = 0.160\n",
            "13937/17600 (epoch 39), train_loss = 1.281, time/batch = 0.149\n",
            "13938/17600 (epoch 39), train_loss = 1.325, time/batch = 0.183\n",
            "13939/17600 (epoch 39), train_loss = 1.316, time/batch = 0.165\n",
            "13940/17600 (epoch 39), train_loss = 1.284, time/batch = 0.163\n",
            "13941/17600 (epoch 39), train_loss = 1.359, time/batch = 0.157\n",
            "13942/17600 (epoch 39), train_loss = 1.400, time/batch = 0.166\n",
            "13943/17600 (epoch 39), train_loss = 1.289, time/batch = 0.154\n",
            "13944/17600 (epoch 39), train_loss = 1.311, time/batch = 0.184\n",
            "13945/17600 (epoch 39), train_loss = 1.265, time/batch = 0.158\n",
            "13946/17600 (epoch 39), train_loss = 1.301, time/batch = 0.173\n",
            "13947/17600 (epoch 39), train_loss = 1.283, time/batch = 0.163\n",
            "13948/17600 (epoch 39), train_loss = 1.274, time/batch = 0.165\n",
            "13949/17600 (epoch 39), train_loss = 1.300, time/batch = 0.156\n",
            "13950/17600 (epoch 39), train_loss = 1.367, time/batch = 0.168\n",
            "13951/17600 (epoch 39), train_loss = 1.340, time/batch = 0.158\n",
            "13952/17600 (epoch 39), train_loss = 1.257, time/batch = 0.178\n",
            "13953/17600 (epoch 39), train_loss = 1.307, time/batch = 0.159\n",
            "13954/17600 (epoch 39), train_loss = 1.311, time/batch = 0.163\n",
            "13955/17600 (epoch 39), train_loss = 1.291, time/batch = 0.156\n",
            "13956/17600 (epoch 39), train_loss = 1.321, time/batch = 0.154\n",
            "13957/17600 (epoch 39), train_loss = 1.286, time/batch = 0.176\n",
            "13958/17600 (epoch 39), train_loss = 1.275, time/batch = 0.166\n",
            "13959/17600 (epoch 39), train_loss = 1.341, time/batch = 0.143\n",
            "13960/17600 (epoch 39), train_loss = 1.337, time/batch = 0.159\n",
            "13961/17600 (epoch 39), train_loss = 1.352, time/batch = 0.148\n",
            "13962/17600 (epoch 39), train_loss = 1.266, time/batch = 0.162\n",
            "13963/17600 (epoch 39), train_loss = 1.298, time/batch = 0.174\n",
            "13964/17600 (epoch 39), train_loss = 1.286, time/batch = 0.173\n",
            "13965/17600 (epoch 39), train_loss = 1.356, time/batch = 0.166\n",
            "13966/17600 (epoch 39), train_loss = 1.352, time/batch = 0.152\n",
            "13967/17600 (epoch 39), train_loss = 1.247, time/batch = 0.170\n",
            "13968/17600 (epoch 39), train_loss = 1.339, time/batch = 0.157\n",
            "13969/17600 (epoch 39), train_loss = 1.287, time/batch = 0.192\n",
            "13970/17600 (epoch 39), train_loss = 1.322, time/batch = 0.165\n",
            "13971/17600 (epoch 39), train_loss = 1.314, time/batch = 0.165\n",
            "13972/17600 (epoch 39), train_loss = 1.283, time/batch = 0.151\n",
            "13973/17600 (epoch 39), train_loss = 1.323, time/batch = 0.165\n",
            "13974/17600 (epoch 39), train_loss = 1.315, time/batch = 0.269\n",
            "13975/17600 (epoch 39), train_loss = 1.288, time/batch = 0.273\n",
            "13976/17600 (epoch 39), train_loss = 1.265, time/batch = 0.287\n",
            "13977/17600 (epoch 39), train_loss = 1.244, time/batch = 0.269\n",
            "13978/17600 (epoch 39), train_loss = 1.379, time/batch = 0.262\n",
            "13979/17600 (epoch 39), train_loss = 1.317, time/batch = 0.277\n",
            "13980/17600 (epoch 39), train_loss = 1.311, time/batch = 0.275\n",
            "13981/17600 (epoch 39), train_loss = 1.317, time/batch = 0.296\n",
            "13982/17600 (epoch 39), train_loss = 1.335, time/batch = 0.284\n",
            "13983/17600 (epoch 39), train_loss = 1.344, time/batch = 0.296\n",
            "13984/17600 (epoch 39), train_loss = 1.391, time/batch = 0.203\n",
            "13985/17600 (epoch 39), train_loss = 1.359, time/batch = 0.180\n",
            "13986/17600 (epoch 39), train_loss = 1.310, time/batch = 0.169\n",
            "13987/17600 (epoch 39), train_loss = 1.312, time/batch = 0.158\n",
            "13988/17600 (epoch 39), train_loss = 1.290, time/batch = 0.177\n",
            "13989/17600 (epoch 39), train_loss = 1.296, time/batch = 0.170\n",
            "13990/17600 (epoch 39), train_loss = 1.278, time/batch = 0.147\n",
            "13991/17600 (epoch 39), train_loss = 1.236, time/batch = 0.157\n",
            "13992/17600 (epoch 39), train_loss = 1.286, time/batch = 0.165\n",
            "13993/17600 (epoch 39), train_loss = 1.293, time/batch = 0.171\n",
            "13994/17600 (epoch 39), train_loss = 1.246, time/batch = 0.163\n",
            "13995/17600 (epoch 39), train_loss = 1.250, time/batch = 0.157\n",
            "13996/17600 (epoch 39), train_loss = 1.283, time/batch = 0.147\n",
            "13997/17600 (epoch 39), train_loss = 1.317, time/batch = 0.171\n",
            "13998/17600 (epoch 39), train_loss = 1.241, time/batch = 0.162\n",
            "13999/17600 (epoch 39), train_loss = 1.309, time/batch = 0.173\n",
            "14000/17600 (epoch 39), train_loss = 1.276, time/batch = 0.178\n",
            "Model saved to checkpoints/dracula/dracula!\n",
            "14001/17600 (epoch 39), train_loss = 1.300, time/batch = 0.162\n",
            "14002/17600 (epoch 39), train_loss = 1.284, time/batch = 0.177\n",
            "14003/17600 (epoch 39), train_loss = 1.340, time/batch = 0.167\n",
            "14004/17600 (epoch 39), train_loss = 1.304, time/batch = 0.165\n",
            "14005/17600 (epoch 39), train_loss = 1.330, time/batch = 0.155\n",
            "14006/17600 (epoch 39), train_loss = 1.297, time/batch = 0.161\n",
            "14007/17600 (epoch 39), train_loss = 1.303, time/batch = 0.159\n",
            "14008/17600 (epoch 39), train_loss = 1.325, time/batch = 0.169\n",
            "14009/17600 (epoch 39), train_loss = 1.372, time/batch = 0.164\n",
            "14010/17600 (epoch 39), train_loss = 1.343, time/batch = 0.159\n",
            "14011/17600 (epoch 39), train_loss = 1.342, time/batch = 0.152\n",
            "14012/17600 (epoch 39), train_loss = 1.360, time/batch = 0.166\n",
            "14013/17600 (epoch 39), train_loss = 1.372, time/batch = 0.150\n",
            "14014/17600 (epoch 39), train_loss = 1.317, time/batch = 0.162\n",
            "14015/17600 (epoch 39), train_loss = 1.363, time/batch = 0.163\n",
            "14016/17600 (epoch 39), train_loss = 1.300, time/batch = 0.194\n",
            "14017/17600 (epoch 39), train_loss = 1.298, time/batch = 0.275\n",
            "14018/17600 (epoch 39), train_loss = 1.364, time/batch = 0.263\n",
            "14019/17600 (epoch 39), train_loss = 1.268, time/batch = 0.297\n",
            "14020/17600 (epoch 39), train_loss = 1.263, time/batch = 0.261\n",
            "14021/17600 (epoch 39), train_loss = 1.337, time/batch = 0.259\n",
            "14022/17600 (epoch 39), train_loss = 1.287, time/batch = 0.288\n",
            "14023/17600 (epoch 39), train_loss = 1.346, time/batch = 0.278\n",
            "14024/17600 (epoch 39), train_loss = 1.292, time/batch = 0.294\n",
            "14025/17600 (epoch 39), train_loss = 1.321, time/batch = 0.288\n",
            "14026/17600 (epoch 39), train_loss = 1.303, time/batch = 0.272\n",
            "14027/17600 (epoch 39), train_loss = 1.318, time/batch = 0.161\n",
            "14028/17600 (epoch 39), train_loss = 1.349, time/batch = 0.178\n",
            "14029/17600 (epoch 39), train_loss = 1.305, time/batch = 0.154\n",
            "14030/17600 (epoch 39), train_loss = 1.375, time/batch = 0.165\n",
            "14031/17600 (epoch 39), train_loss = 1.368, time/batch = 0.159\n",
            "14032/17600 (epoch 39), train_loss = 1.378, time/batch = 0.166\n",
            "14033/17600 (epoch 39), train_loss = 1.292, time/batch = 0.152\n",
            "14034/17600 (epoch 39), train_loss = 1.329, time/batch = 0.179\n",
            "14035/17600 (epoch 39), train_loss = 1.333, time/batch = 0.156\n",
            "14036/17600 (epoch 39), train_loss = 1.334, time/batch = 0.161\n",
            "14037/17600 (epoch 39), train_loss = 1.335, time/batch = 0.155\n",
            "14038/17600 (epoch 39), train_loss = 1.372, time/batch = 0.163\n",
            "14039/17600 (epoch 39), train_loss = 1.320, time/batch = 0.147\n",
            "14040/17600 (epoch 39), train_loss = 1.352, time/batch = 0.180\n",
            "14041/17600 (epoch 39), train_loss = 1.289, time/batch = 0.163\n",
            "14042/17600 (epoch 39), train_loss = 1.304, time/batch = 0.175\n",
            "14043/17600 (epoch 39), train_loss = 1.272, time/batch = 0.160\n",
            "14044/17600 (epoch 39), train_loss = 1.370, time/batch = 0.168\n",
            "14045/17600 (epoch 39), train_loss = 1.357, time/batch = 0.155\n",
            "14046/17600 (epoch 39), train_loss = 1.356, time/batch = 0.154\n",
            "14047/17600 (epoch 39), train_loss = 1.337, time/batch = 0.168\n",
            "14048/17600 (epoch 39), train_loss = 1.378, time/batch = 0.162\n",
            "14049/17600 (epoch 39), train_loss = 1.344, time/batch = 0.156\n",
            "14050/17600 (epoch 39), train_loss = 1.399, time/batch = 0.159\n",
            "14051/17600 (epoch 39), train_loss = 1.368, time/batch = 0.160\n",
            "14052/17600 (epoch 39), train_loss = 1.350, time/batch = 0.173\n",
            "14053/17600 (epoch 39), train_loss = 1.303, time/batch = 0.172\n",
            "14054/17600 (epoch 39), train_loss = 1.378, time/batch = 0.169\n",
            "14055/17600 (epoch 39), train_loss = 1.358, time/batch = 0.151\n",
            "14056/17600 (epoch 39), train_loss = 1.394, time/batch = 0.165\n",
            "14057/17600 (epoch 39), train_loss = 1.362, time/batch = 0.158\n",
            "14058/17600 (epoch 39), train_loss = 1.285, time/batch = 0.154\n",
            "14059/17600 (epoch 39), train_loss = 1.312, time/batch = 0.165\n",
            "14060/17600 (epoch 39), train_loss = 1.362, time/batch = 0.162\n",
            "14061/17600 (epoch 39), train_loss = 1.371, time/batch = 0.152\n",
            "14062/17600 (epoch 39), train_loss = 1.327, time/batch = 0.162\n",
            "14063/17600 (epoch 39), train_loss = 1.320, time/batch = 0.154\n",
            "14064/17600 (epoch 39), train_loss = 1.329, time/batch = 0.165\n",
            "14065/17600 (epoch 39), train_loss = 1.365, time/batch = 0.154\n",
            "14066/17600 (epoch 39), train_loss = 1.313, time/batch = 0.177\n",
            "14067/17600 (epoch 39), train_loss = 1.344, time/batch = 0.158\n",
            "14068/17600 (epoch 39), train_loss = 1.309, time/batch = 0.162\n",
            "14069/17600 (epoch 39), train_loss = 1.352, time/batch = 0.165\n",
            "14070/17600 (epoch 39), train_loss = 1.361, time/batch = 0.162\n",
            "14071/17600 (epoch 39), train_loss = 1.325, time/batch = 0.153\n",
            "14072/17600 (epoch 39), train_loss = 1.305, time/batch = 0.176\n",
            "14073/17600 (epoch 39), train_loss = 1.323, time/batch = 0.176\n",
            "14074/17600 (epoch 39), train_loss = 1.345, time/batch = 0.171\n",
            "14075/17600 (epoch 39), train_loss = 1.318, time/batch = 0.160\n",
            "14076/17600 (epoch 39), train_loss = 1.342, time/batch = 0.165\n",
            "14077/17600 (epoch 39), train_loss = 1.326, time/batch = 0.150\n",
            "14078/17600 (epoch 39), train_loss = 1.317, time/batch = 0.176\n",
            "14079/17600 (epoch 39), train_loss = 1.310, time/batch = 0.155\n",
            "14080/17600 (epoch 40), train_loss = 1.123, time/batch = 0.158\n",
            "14081/17600 (epoch 40), train_loss = 1.339, time/batch = 0.142\n",
            "14082/17600 (epoch 40), train_loss = 1.334, time/batch = 0.165\n",
            "14083/17600 (epoch 40), train_loss = 1.279, time/batch = 0.172\n",
            "14084/17600 (epoch 40), train_loss = 1.373, time/batch = 0.163\n",
            "14085/17600 (epoch 40), train_loss = 1.349, time/batch = 0.154\n",
            "14086/17600 (epoch 40), train_loss = 1.297, time/batch = 0.232\n",
            "14087/17600 (epoch 40), train_loss = 1.284, time/batch = 0.282\n",
            "14088/17600 (epoch 40), train_loss = 1.324, time/batch = 0.266\n",
            "14089/17600 (epoch 40), train_loss = 1.379, time/batch = 0.293\n",
            "14090/17600 (epoch 40), train_loss = 1.306, time/batch = 0.268\n",
            "14091/17600 (epoch 40), train_loss = 1.322, time/batch = 0.270\n",
            "14092/17600 (epoch 40), train_loss = 1.302, time/batch = 0.300\n",
            "14093/17600 (epoch 40), train_loss = 1.287, time/batch = 0.269\n",
            "14094/17600 (epoch 40), train_loss = 1.268, time/batch = 0.256\n",
            "14095/17600 (epoch 40), train_loss = 1.332, time/batch = 0.308\n",
            "14096/17600 (epoch 40), train_loss = 1.344, time/batch = 0.259\n",
            "14097/17600 (epoch 40), train_loss = 1.299, time/batch = 0.178\n",
            "14098/17600 (epoch 40), train_loss = 1.312, time/batch = 0.159\n",
            "14099/17600 (epoch 40), train_loss = 1.356, time/batch = 0.163\n",
            "14100/17600 (epoch 40), train_loss = 1.279, time/batch = 0.154\n",
            "14101/17600 (epoch 40), train_loss = 1.333, time/batch = 0.176\n",
            "14102/17600 (epoch 40), train_loss = 1.299, time/batch = 0.159\n",
            "14103/17600 (epoch 40), train_loss = 1.272, time/batch = 0.169\n",
            "14104/17600 (epoch 40), train_loss = 1.358, time/batch = 0.153\n",
            "14105/17600 (epoch 40), train_loss = 1.288, time/batch = 0.169\n",
            "14106/17600 (epoch 40), train_loss = 1.336, time/batch = 0.153\n",
            "14107/17600 (epoch 40), train_loss = 1.318, time/batch = 0.176\n",
            "14108/17600 (epoch 40), train_loss = 1.297, time/batch = 0.152\n",
            "14109/17600 (epoch 40), train_loss = 1.230, time/batch = 0.166\n",
            "14110/17600 (epoch 40), train_loss = 1.338, time/batch = 0.154\n",
            "14111/17600 (epoch 40), train_loss = 1.318, time/batch = 0.158\n",
            "14112/17600 (epoch 40), train_loss = 1.294, time/batch = 0.152\n",
            "14113/17600 (epoch 40), train_loss = 1.298, time/batch = 0.171\n",
            "14114/17600 (epoch 40), train_loss = 1.301, time/batch = 0.168\n",
            "14115/17600 (epoch 40), train_loss = 1.315, time/batch = 0.175\n",
            "14116/17600 (epoch 40), train_loss = 1.341, time/batch = 0.155\n",
            "14117/17600 (epoch 40), train_loss = 1.338, time/batch = 0.158\n",
            "14118/17600 (epoch 40), train_loss = 1.318, time/batch = 0.164\n",
            "14119/17600 (epoch 40), train_loss = 1.296, time/batch = 0.169\n",
            "14120/17600 (epoch 40), train_loss = 1.343, time/batch = 0.166\n",
            "14121/17600 (epoch 40), train_loss = 1.278, time/batch = 0.159\n",
            "14122/17600 (epoch 40), train_loss = 1.334, time/batch = 0.154\n",
            "14123/17600 (epoch 40), train_loss = 1.254, time/batch = 0.163\n",
            "14124/17600 (epoch 40), train_loss = 1.323, time/batch = 0.160\n",
            "14125/17600 (epoch 40), train_loss = 1.294, time/batch = 0.159\n",
            "14126/17600 (epoch 40), train_loss = 1.250, time/batch = 0.188\n",
            "14127/17600 (epoch 40), train_loss = 1.309, time/batch = 0.151\n",
            "14128/17600 (epoch 40), train_loss = 1.259, time/batch = 0.168\n",
            "14129/17600 (epoch 40), train_loss = 1.278, time/batch = 0.163\n",
            "14130/17600 (epoch 40), train_loss = 1.301, time/batch = 0.162\n",
            "14131/17600 (epoch 40), train_loss = 1.292, time/batch = 0.160\n",
            "14132/17600 (epoch 40), train_loss = 1.261, time/batch = 0.161\n",
            "14133/17600 (epoch 40), train_loss = 1.289, time/batch = 0.165\n",
            "14134/17600 (epoch 40), train_loss = 1.335, time/batch = 0.174\n",
            "14135/17600 (epoch 40), train_loss = 1.312, time/batch = 0.166\n",
            "14136/17600 (epoch 40), train_loss = 1.314, time/batch = 0.149\n",
            "14137/17600 (epoch 40), train_loss = 1.307, time/batch = 0.164\n",
            "14138/17600 (epoch 40), train_loss = 1.287, time/batch = 0.164\n",
            "14139/17600 (epoch 40), train_loss = 1.313, time/batch = 0.176\n",
            "14140/17600 (epoch 40), train_loss = 1.309, time/batch = 0.172\n",
            "14141/17600 (epoch 40), train_loss = 1.251, time/batch = 0.143\n",
            "14142/17600 (epoch 40), train_loss = 1.253, time/batch = 0.158\n",
            "14143/17600 (epoch 40), train_loss = 1.272, time/batch = 0.141\n",
            "14144/17600 (epoch 40), train_loss = 1.303, time/batch = 0.164\n",
            "14145/17600 (epoch 40), train_loss = 1.270, time/batch = 0.165\n",
            "14146/17600 (epoch 40), train_loss = 1.307, time/batch = 0.171\n",
            "14147/17600 (epoch 40), train_loss = 1.265, time/batch = 0.156\n",
            "14148/17600 (epoch 40), train_loss = 1.313, time/batch = 0.173\n",
            "14149/17600 (epoch 40), train_loss = 1.244, time/batch = 0.154\n",
            "14150/17600 (epoch 40), train_loss = 1.290, time/batch = 0.168\n",
            "14151/17600 (epoch 40), train_loss = 1.335, time/batch = 0.161\n",
            "14152/17600 (epoch 40), train_loss = 1.296, time/batch = 0.178\n",
            "14153/17600 (epoch 40), train_loss = 1.368, time/batch = 0.159\n",
            "14154/17600 (epoch 40), train_loss = 1.297, time/batch = 0.153\n",
            "14155/17600 (epoch 40), train_loss = 1.300, time/batch = 0.158\n",
            "14156/17600 (epoch 40), train_loss = 1.255, time/batch = 0.172\n",
            "14157/17600 (epoch 40), train_loss = 1.310, time/batch = 0.187\n",
            "14158/17600 (epoch 40), train_loss = 1.317, time/batch = 0.285\n",
            "14159/17600 (epoch 40), train_loss = 1.392, time/batch = 0.257\n",
            "14160/17600 (epoch 40), train_loss = 1.361, time/batch = 0.276\n",
            "14161/17600 (epoch 40), train_loss = 1.355, time/batch = 0.269\n",
            "14162/17600 (epoch 40), train_loss = 1.336, time/batch = 0.284\n",
            "14163/17600 (epoch 40), train_loss = 1.307, time/batch = 0.263\n",
            "14164/17600 (epoch 40), train_loss = 1.284, time/batch = 0.256\n",
            "14165/17600 (epoch 40), train_loss = 1.319, time/batch = 0.304\n",
            "14166/17600 (epoch 40), train_loss = 1.310, time/batch = 0.277\n",
            "14167/17600 (epoch 40), train_loss = 1.250, time/batch = 0.278\n",
            "14168/17600 (epoch 40), train_loss = 1.289, time/batch = 0.212\n",
            "14169/17600 (epoch 40), train_loss = 1.229, time/batch = 0.160\n",
            "14170/17600 (epoch 40), train_loss = 1.294, time/batch = 0.172\n",
            "14171/17600 (epoch 40), train_loss = 1.281, time/batch = 0.163\n",
            "14172/17600 (epoch 40), train_loss = 1.258, time/batch = 0.154\n",
            "14173/17600 (epoch 40), train_loss = 1.313, time/batch = 0.172\n",
            "14174/17600 (epoch 40), train_loss = 1.296, time/batch = 0.159\n",
            "14175/17600 (epoch 40), train_loss = 1.260, time/batch = 0.172\n",
            "14176/17600 (epoch 40), train_loss = 1.235, time/batch = 0.171\n",
            "14177/17600 (epoch 40), train_loss = 1.285, time/batch = 0.164\n",
            "14178/17600 (epoch 40), train_loss = 1.319, time/batch = 0.156\n",
            "14179/17600 (epoch 40), train_loss = 1.278, time/batch = 0.175\n",
            "14180/17600 (epoch 40), train_loss = 1.251, time/batch = 0.153\n",
            "14181/17600 (epoch 40), train_loss = 1.336, time/batch = 0.168\n",
            "14182/17600 (epoch 40), train_loss = 1.276, time/batch = 0.164\n",
            "14183/17600 (epoch 40), train_loss = 1.385, time/batch = 0.177\n",
            "14184/17600 (epoch 40), train_loss = 1.312, time/batch = 0.156\n",
            "14185/17600 (epoch 40), train_loss = 1.286, time/batch = 0.169\n",
            "14186/17600 (epoch 40), train_loss = 1.327, time/batch = 0.147\n",
            "14187/17600 (epoch 40), train_loss = 1.313, time/batch = 0.175\n",
            "14188/17600 (epoch 40), train_loss = 1.317, time/batch = 0.157\n",
            "14189/17600 (epoch 40), train_loss = 1.301, time/batch = 0.168\n",
            "14190/17600 (epoch 40), train_loss = 1.257, time/batch = 0.162\n",
            "14191/17600 (epoch 40), train_loss = 1.280, time/batch = 0.162\n",
            "14192/17600 (epoch 40), train_loss = 1.311, time/batch = 0.144\n",
            "14193/17600 (epoch 40), train_loss = 1.205, time/batch = 0.154\n",
            "14194/17600 (epoch 40), train_loss = 1.280, time/batch = 0.158\n",
            "14195/17600 (epoch 40), train_loss = 1.310, time/batch = 0.188\n",
            "14196/17600 (epoch 40), train_loss = 1.278, time/batch = 0.153\n",
            "14197/17600 (epoch 40), train_loss = 1.304, time/batch = 0.169\n",
            "14198/17600 (epoch 40), train_loss = 1.247, time/batch = 0.146\n",
            "14199/17600 (epoch 40), train_loss = 1.350, time/batch = 0.149\n",
            "14200/17600 (epoch 40), train_loss = 1.257, time/batch = 0.158\n",
            "14201/17600 (epoch 40), train_loss = 1.318, time/batch = 0.153\n",
            "14202/17600 (epoch 40), train_loss = 1.343, time/batch = 0.174\n",
            "14203/17600 (epoch 40), train_loss = 1.329, time/batch = 0.147\n",
            "14204/17600 (epoch 40), train_loss = 1.245, time/batch = 0.164\n",
            "14205/17600 (epoch 40), train_loss = 1.311, time/batch = 0.154\n",
            "14206/17600 (epoch 40), train_loss = 1.308, time/batch = 0.160\n",
            "14207/17600 (epoch 40), train_loss = 1.295, time/batch = 0.153\n",
            "14208/17600 (epoch 40), train_loss = 1.277, time/batch = 0.169\n",
            "14209/17600 (epoch 40), train_loss = 1.310, time/batch = 0.152\n",
            "14210/17600 (epoch 40), train_loss = 1.323, time/batch = 0.167\n",
            "14211/17600 (epoch 40), train_loss = 1.327, time/batch = 0.150\n",
            "14212/17600 (epoch 40), train_loss = 1.297, time/batch = 0.163\n",
            "14213/17600 (epoch 40), train_loss = 1.269, time/batch = 0.154\n",
            "14214/17600 (epoch 40), train_loss = 1.280, time/batch = 0.159\n",
            "14215/17600 (epoch 40), train_loss = 1.251, time/batch = 0.169\n",
            "14216/17600 (epoch 40), train_loss = 1.320, time/batch = 0.153\n",
            "14217/17600 (epoch 40), train_loss = 1.345, time/batch = 0.144\n",
            "14218/17600 (epoch 40), train_loss = 1.290, time/batch = 0.153\n",
            "14219/17600 (epoch 40), train_loss = 1.342, time/batch = 0.148\n",
            "14220/17600 (epoch 40), train_loss = 1.370, time/batch = 0.159\n",
            "14221/17600 (epoch 40), train_loss = 1.280, time/batch = 0.167\n",
            "14222/17600 (epoch 40), train_loss = 1.306, time/batch = 0.175\n",
            "14223/17600 (epoch 40), train_loss = 1.271, time/batch = 0.149\n",
            "14224/17600 (epoch 40), train_loss = 1.325, time/batch = 0.167\n",
            "14225/17600 (epoch 40), train_loss = 1.308, time/batch = 0.158\n",
            "14226/17600 (epoch 40), train_loss = 1.284, time/batch = 0.168\n",
            "14227/17600 (epoch 40), train_loss = 1.292, time/batch = 0.158\n",
            "14228/17600 (epoch 40), train_loss = 1.274, time/batch = 0.174\n",
            "14229/17600 (epoch 40), train_loss = 1.337, time/batch = 0.154\n",
            "14230/17600 (epoch 40), train_loss = 1.225, time/batch = 0.244\n",
            "14231/17600 (epoch 40), train_loss = 1.298, time/batch = 0.283\n",
            "14232/17600 (epoch 40), train_loss = 1.323, time/batch = 0.293\n",
            "14233/17600 (epoch 40), train_loss = 1.291, time/batch = 0.278\n",
            "14234/17600 (epoch 40), train_loss = 1.308, time/batch = 0.274\n",
            "14235/17600 (epoch 40), train_loss = 1.245, time/batch = 0.265\n",
            "14236/17600 (epoch 40), train_loss = 1.243, time/batch = 0.298\n",
            "14237/17600 (epoch 40), train_loss = 1.261, time/batch = 0.316\n",
            "14238/17600 (epoch 40), train_loss = 1.322, time/batch = 0.299\n",
            "14239/17600 (epoch 40), train_loss = 1.365, time/batch = 0.301\n",
            "14240/17600 (epoch 40), train_loss = 1.320, time/batch = 0.245\n",
            "14241/17600 (epoch 40), train_loss = 1.320, time/batch = 0.169\n",
            "14242/17600 (epoch 40), train_loss = 1.251, time/batch = 0.160\n",
            "14243/17600 (epoch 40), train_loss = 1.280, time/batch = 0.162\n",
            "14244/17600 (epoch 40), train_loss = 1.302, time/batch = 0.153\n",
            "14245/17600 (epoch 40), train_loss = 1.289, time/batch = 0.180\n",
            "14246/17600 (epoch 40), train_loss = 1.301, time/batch = 0.157\n",
            "14247/17600 (epoch 40), train_loss = 1.328, time/batch = 0.157\n",
            "14248/17600 (epoch 40), train_loss = 1.319, time/batch = 0.157\n",
            "14249/17600 (epoch 40), train_loss = 1.299, time/batch = 0.159\n",
            "14250/17600 (epoch 40), train_loss = 1.234, time/batch = 0.147\n",
            "14251/17600 (epoch 40), train_loss = 1.245, time/batch = 0.157\n",
            "14252/17600 (epoch 40), train_loss = 1.334, time/batch = 0.173\n",
            "14253/17600 (epoch 40), train_loss = 1.302, time/batch = 0.173\n",
            "14254/17600 (epoch 40), train_loss = 1.292, time/batch = 0.141\n",
            "14255/17600 (epoch 40), train_loss = 1.296, time/batch = 0.164\n",
            "14256/17600 (epoch 40), train_loss = 1.211, time/batch = 0.145\n",
            "14257/17600 (epoch 40), train_loss = 1.274, time/batch = 0.151\n",
            "14258/17600 (epoch 40), train_loss = 1.274, time/batch = 0.170\n",
            "14259/17600 (epoch 40), train_loss = 1.252, time/batch = 0.178\n",
            "14260/17600 (epoch 40), train_loss = 1.251, time/batch = 0.157\n",
            "14261/17600 (epoch 40), train_loss = 1.304, time/batch = 0.155\n",
            "14262/17600 (epoch 40), train_loss = 1.361, time/batch = 0.164\n",
            "14263/17600 (epoch 40), train_loss = 1.277, time/batch = 0.155\n",
            "14264/17600 (epoch 40), train_loss = 1.277, time/batch = 0.202\n",
            "14265/17600 (epoch 40), train_loss = 1.241, time/batch = 0.179\n",
            "14266/17600 (epoch 40), train_loss = 1.317, time/batch = 0.169\n",
            "14267/17600 (epoch 40), train_loss = 1.325, time/batch = 0.165\n",
            "14268/17600 (epoch 40), train_loss = 1.245, time/batch = 0.167\n",
            "14269/17600 (epoch 40), train_loss = 1.341, time/batch = 0.149\n",
            "14270/17600 (epoch 40), train_loss = 1.306, time/batch = 0.161\n",
            "14271/17600 (epoch 40), train_loss = 1.286, time/batch = 0.171\n",
            "14272/17600 (epoch 40), train_loss = 1.295, time/batch = 0.164\n",
            "14273/17600 (epoch 40), train_loss = 1.313, time/batch = 0.156\n",
            "14274/17600 (epoch 40), train_loss = 1.314, time/batch = 0.163\n",
            "14275/17600 (epoch 40), train_loss = 1.324, time/batch = 0.154\n",
            "14276/17600 (epoch 40), train_loss = 1.280, time/batch = 0.161\n",
            "14277/17600 (epoch 40), train_loss = 1.332, time/batch = 0.175\n",
            "14278/17600 (epoch 40), train_loss = 1.279, time/batch = 0.176\n",
            "14279/17600 (epoch 40), train_loss = 1.312, time/batch = 0.177\n",
            "14280/17600 (epoch 40), train_loss = 1.316, time/batch = 0.160\n",
            "14281/17600 (epoch 40), train_loss = 1.282, time/batch = 0.160\n",
            "14282/17600 (epoch 40), train_loss = 1.273, time/batch = 0.156\n",
            "14283/17600 (epoch 40), train_loss = 1.301, time/batch = 0.183\n",
            "14284/17600 (epoch 40), train_loss = 1.335, time/batch = 0.152\n",
            "14285/17600 (epoch 40), train_loss = 1.382, time/batch = 0.169\n",
            "14286/17600 (epoch 40), train_loss = 1.297, time/batch = 0.168\n",
            "14287/17600 (epoch 40), train_loss = 1.298, time/batch = 0.172\n",
            "14288/17600 (epoch 40), train_loss = 1.299, time/batch = 0.190\n",
            "14289/17600 (epoch 40), train_loss = 1.278, time/batch = 0.176\n",
            "14290/17600 (epoch 40), train_loss = 1.323, time/batch = 0.171\n",
            "14291/17600 (epoch 40), train_loss = 1.314, time/batch = 0.175\n",
            "14292/17600 (epoch 40), train_loss = 1.283, time/batch = 0.175\n",
            "14293/17600 (epoch 40), train_loss = 1.357, time/batch = 0.157\n",
            "14294/17600 (epoch 40), train_loss = 1.397, time/batch = 0.168\n",
            "14295/17600 (epoch 40), train_loss = 1.286, time/batch = 0.172\n",
            "14296/17600 (epoch 40), train_loss = 1.308, time/batch = 0.172\n",
            "14297/17600 (epoch 40), train_loss = 1.263, time/batch = 0.154\n",
            "14298/17600 (epoch 40), train_loss = 1.299, time/batch = 0.160\n",
            "14299/17600 (epoch 40), train_loss = 1.281, time/batch = 0.144\n",
            "14300/17600 (epoch 40), train_loss = 1.272, time/batch = 0.210\n",
            "14301/17600 (epoch 40), train_loss = 1.298, time/batch = 0.292\n",
            "14302/17600 (epoch 40), train_loss = 1.365, time/batch = 0.280\n",
            "14303/17600 (epoch 40), train_loss = 1.338, time/batch = 0.291\n",
            "14304/17600 (epoch 40), train_loss = 1.255, time/batch = 0.282\n",
            "14305/17600 (epoch 40), train_loss = 1.304, time/batch = 0.303\n",
            "14306/17600 (epoch 40), train_loss = 1.309, time/batch = 0.271\n",
            "14307/17600 (epoch 40), train_loss = 1.288, time/batch = 0.282\n",
            "14308/17600 (epoch 40), train_loss = 1.319, time/batch = 0.279\n",
            "14309/17600 (epoch 40), train_loss = 1.283, time/batch = 0.302\n",
            "14310/17600 (epoch 40), train_loss = 1.273, time/batch = 0.296\n",
            "14311/17600 (epoch 40), train_loss = 1.338, time/batch = 0.164\n",
            "14312/17600 (epoch 40), train_loss = 1.335, time/batch = 0.161\n",
            "14313/17600 (epoch 40), train_loss = 1.350, time/batch = 0.167\n",
            "14314/17600 (epoch 40), train_loss = 1.264, time/batch = 0.162\n",
            "14315/17600 (epoch 40), train_loss = 1.295, time/batch = 0.146\n",
            "14316/17600 (epoch 40), train_loss = 1.284, time/batch = 0.184\n",
            "14317/17600 (epoch 40), train_loss = 1.354, time/batch = 0.172\n",
            "14318/17600 (epoch 40), train_loss = 1.350, time/batch = 0.173\n",
            "14319/17600 (epoch 40), train_loss = 1.245, time/batch = 0.173\n",
            "14320/17600 (epoch 40), train_loss = 1.337, time/batch = 0.161\n",
            "14321/17600 (epoch 40), train_loss = 1.285, time/batch = 0.160\n",
            "14322/17600 (epoch 40), train_loss = 1.320, time/batch = 0.165\n",
            "14323/17600 (epoch 40), train_loss = 1.312, time/batch = 0.157\n",
            "14324/17600 (epoch 40), train_loss = 1.281, time/batch = 0.155\n",
            "14325/17600 (epoch 40), train_loss = 1.321, time/batch = 0.144\n",
            "14326/17600 (epoch 40), train_loss = 1.313, time/batch = 0.180\n",
            "14327/17600 (epoch 40), train_loss = 1.286, time/batch = 0.153\n",
            "14328/17600 (epoch 40), train_loss = 1.263, time/batch = 0.182\n",
            "14329/17600 (epoch 40), train_loss = 1.242, time/batch = 0.153\n",
            "14330/17600 (epoch 40), train_loss = 1.377, time/batch = 0.177\n",
            "14331/17600 (epoch 40), train_loss = 1.314, time/batch = 0.147\n",
            "14332/17600 (epoch 40), train_loss = 1.308, time/batch = 0.181\n",
            "14333/17600 (epoch 40), train_loss = 1.315, time/batch = 0.152\n",
            "14334/17600 (epoch 40), train_loss = 1.333, time/batch = 0.177\n",
            "14335/17600 (epoch 40), train_loss = 1.341, time/batch = 0.150\n",
            "14336/17600 (epoch 40), train_loss = 1.389, time/batch = 0.163\n",
            "14337/17600 (epoch 40), train_loss = 1.357, time/batch = 0.152\n",
            "14338/17600 (epoch 40), train_loss = 1.308, time/batch = 0.176\n",
            "14339/17600 (epoch 40), train_loss = 1.310, time/batch = 0.146\n",
            "14340/17600 (epoch 40), train_loss = 1.287, time/batch = 0.165\n",
            "14341/17600 (epoch 40), train_loss = 1.294, time/batch = 0.165\n",
            "14342/17600 (epoch 40), train_loss = 1.276, time/batch = 0.168\n",
            "14343/17600 (epoch 40), train_loss = 1.234, time/batch = 0.140\n",
            "14344/17600 (epoch 40), train_loss = 1.284, time/batch = 0.154\n",
            "14345/17600 (epoch 40), train_loss = 1.291, time/batch = 0.176\n",
            "14346/17600 (epoch 40), train_loss = 1.244, time/batch = 0.158\n",
            "14347/17600 (epoch 40), train_loss = 1.248, time/batch = 0.158\n",
            "14348/17600 (epoch 40), train_loss = 1.280, time/batch = 0.157\n",
            "14349/17600 (epoch 40), train_loss = 1.314, time/batch = 0.154\n",
            "14350/17600 (epoch 40), train_loss = 1.239, time/batch = 0.170\n",
            "14351/17600 (epoch 40), train_loss = 1.307, time/batch = 0.173\n",
            "14352/17600 (epoch 40), train_loss = 1.274, time/batch = 0.167\n",
            "14353/17600 (epoch 40), train_loss = 1.298, time/batch = 0.151\n",
            "14354/17600 (epoch 40), train_loss = 1.282, time/batch = 0.165\n",
            "14355/17600 (epoch 40), train_loss = 1.338, time/batch = 0.157\n",
            "14356/17600 (epoch 40), train_loss = 1.301, time/batch = 0.177\n",
            "14357/17600 (epoch 40), train_loss = 1.328, time/batch = 0.168\n",
            "14358/17600 (epoch 40), train_loss = 1.295, time/batch = 0.162\n",
            "14359/17600 (epoch 40), train_loss = 1.300, time/batch = 0.162\n",
            "14360/17600 (epoch 40), train_loss = 1.323, time/batch = 0.166\n",
            "14361/17600 (epoch 40), train_loss = 1.370, time/batch = 0.152\n",
            "14362/17600 (epoch 40), train_loss = 1.341, time/batch = 0.177\n",
            "14363/17600 (epoch 40), train_loss = 1.339, time/batch = 0.154\n",
            "14364/17600 (epoch 40), train_loss = 1.358, time/batch = 0.174\n",
            "14365/17600 (epoch 40), train_loss = 1.370, time/batch = 0.158\n",
            "14366/17600 (epoch 40), train_loss = 1.315, time/batch = 0.162\n",
            "14367/17600 (epoch 40), train_loss = 1.361, time/batch = 0.149\n",
            "14368/17600 (epoch 40), train_loss = 1.299, time/batch = 0.170\n",
            "14369/17600 (epoch 40), train_loss = 1.296, time/batch = 0.161\n",
            "14370/17600 (epoch 40), train_loss = 1.361, time/batch = 0.177\n",
            "14371/17600 (epoch 40), train_loss = 1.266, time/batch = 0.182\n",
            "14372/17600 (epoch 40), train_loss = 1.261, time/batch = 0.277\n",
            "14373/17600 (epoch 40), train_loss = 1.335, time/batch = 0.285\n",
            "14374/17600 (epoch 40), train_loss = 1.284, time/batch = 0.305\n",
            "14375/17600 (epoch 40), train_loss = 1.344, time/batch = 0.276\n",
            "14376/17600 (epoch 40), train_loss = 1.290, time/batch = 0.272\n",
            "14377/17600 (epoch 40), train_loss = 1.318, time/batch = 0.276\n",
            "14378/17600 (epoch 40), train_loss = 1.302, time/batch = 0.290\n",
            "14379/17600 (epoch 40), train_loss = 1.316, time/batch = 0.286\n",
            "14380/17600 (epoch 40), train_loss = 1.347, time/batch = 0.282\n",
            "14381/17600 (epoch 40), train_loss = 1.304, time/batch = 0.241\n",
            "14382/17600 (epoch 40), train_loss = 1.373, time/batch = 0.176\n",
            "14383/17600 (epoch 40), train_loss = 1.366, time/batch = 0.154\n",
            "14384/17600 (epoch 40), train_loss = 1.376, time/batch = 0.166\n",
            "14385/17600 (epoch 40), train_loss = 1.290, time/batch = 0.144\n",
            "14386/17600 (epoch 40), train_loss = 1.327, time/batch = 0.165\n",
            "14387/17600 (epoch 40), train_loss = 1.331, time/batch = 0.159\n",
            "14388/17600 (epoch 40), train_loss = 1.331, time/batch = 0.174\n",
            "14389/17600 (epoch 40), train_loss = 1.333, time/batch = 0.165\n",
            "14390/17600 (epoch 40), train_loss = 1.370, time/batch = 0.150\n",
            "14391/17600 (epoch 40), train_loss = 1.319, time/batch = 0.164\n",
            "14392/17600 (epoch 40), train_loss = 1.351, time/batch = 0.154\n",
            "14393/17600 (epoch 40), train_loss = 1.287, time/batch = 0.165\n",
            "14394/17600 (epoch 40), train_loss = 1.302, time/batch = 0.142\n",
            "14395/17600 (epoch 40), train_loss = 1.270, time/batch = 0.187\n",
            "14396/17600 (epoch 40), train_loss = 1.367, time/batch = 0.148\n",
            "14397/17600 (epoch 40), train_loss = 1.355, time/batch = 0.164\n",
            "14398/17600 (epoch 40), train_loss = 1.354, time/batch = 0.145\n",
            "14399/17600 (epoch 40), train_loss = 1.335, time/batch = 0.174\n",
            "14400/17600 (epoch 40), train_loss = 1.376, time/batch = 0.165\n",
            "14401/17600 (epoch 40), train_loss = 1.341, time/batch = 0.180\n",
            "14402/17600 (epoch 40), train_loss = 1.396, time/batch = 0.153\n",
            "14403/17600 (epoch 40), train_loss = 1.366, time/batch = 0.169\n",
            "14404/17600 (epoch 40), train_loss = 1.349, time/batch = 0.150\n",
            "14405/17600 (epoch 40), train_loss = 1.301, time/batch = 0.173\n",
            "14406/17600 (epoch 40), train_loss = 1.377, time/batch = 0.155\n",
            "14407/17600 (epoch 40), train_loss = 1.356, time/batch = 0.182\n",
            "14408/17600 (epoch 40), train_loss = 1.392, time/batch = 0.164\n",
            "14409/17600 (epoch 40), train_loss = 1.361, time/batch = 0.167\n",
            "14410/17600 (epoch 40), train_loss = 1.283, time/batch = 0.166\n",
            "14411/17600 (epoch 40), train_loss = 1.310, time/batch = 0.165\n",
            "14412/17600 (epoch 40), train_loss = 1.359, time/batch = 0.159\n",
            "14413/17600 (epoch 40), train_loss = 1.369, time/batch = 0.177\n",
            "14414/17600 (epoch 40), train_loss = 1.326, time/batch = 0.166\n",
            "14415/17600 (epoch 40), train_loss = 1.318, time/batch = 0.161\n",
            "14416/17600 (epoch 40), train_loss = 1.326, time/batch = 0.148\n",
            "14417/17600 (epoch 40), train_loss = 1.363, time/batch = 0.176\n",
            "14418/17600 (epoch 40), train_loss = 1.311, time/batch = 0.142\n",
            "14419/17600 (epoch 40), train_loss = 1.341, time/batch = 0.164\n",
            "14420/17600 (epoch 40), train_loss = 1.307, time/batch = 0.175\n",
            "14421/17600 (epoch 40), train_loss = 1.350, time/batch = 0.169\n",
            "14422/17600 (epoch 40), train_loss = 1.359, time/batch = 0.158\n",
            "14423/17600 (epoch 40), train_loss = 1.323, time/batch = 0.173\n",
            "14424/17600 (epoch 40), train_loss = 1.302, time/batch = 0.157\n",
            "14425/17600 (epoch 40), train_loss = 1.320, time/batch = 0.156\n",
            "14426/17600 (epoch 40), train_loss = 1.343, time/batch = 0.173\n",
            "14427/17600 (epoch 40), train_loss = 1.316, time/batch = 0.176\n",
            "14428/17600 (epoch 40), train_loss = 1.340, time/batch = 0.157\n",
            "14429/17600 (epoch 40), train_loss = 1.324, time/batch = 0.169\n",
            "14430/17600 (epoch 40), train_loss = 1.315, time/batch = 0.173\n",
            "14431/17600 (epoch 40), train_loss = 1.308, time/batch = 0.167\n",
            "14432/17600 (epoch 41), train_loss = 1.114, time/batch = 0.157\n",
            "14433/17600 (epoch 41), train_loss = 1.337, time/batch = 0.149\n",
            "14434/17600 (epoch 41), train_loss = 1.332, time/batch = 0.178\n",
            "14435/17600 (epoch 41), train_loss = 1.277, time/batch = 0.151\n",
            "14436/17600 (epoch 41), train_loss = 1.372, time/batch = 0.170\n",
            "14437/17600 (epoch 41), train_loss = 1.348, time/batch = 0.166\n",
            "14438/17600 (epoch 41), train_loss = 1.295, time/batch = 0.159\n",
            "14439/17600 (epoch 41), train_loss = 1.281, time/batch = 0.163\n",
            "14440/17600 (epoch 41), train_loss = 1.322, time/batch = 0.209\n",
            "14441/17600 (epoch 41), train_loss = 1.376, time/batch = 0.265\n",
            "14442/17600 (epoch 41), train_loss = 1.304, time/batch = 0.271\n",
            "14443/17600 (epoch 41), train_loss = 1.318, time/batch = 0.292\n",
            "14444/17600 (epoch 41), train_loss = 1.300, time/batch = 0.267\n",
            "14445/17600 (epoch 41), train_loss = 1.285, time/batch = 0.277\n",
            "14446/17600 (epoch 41), train_loss = 1.266, time/batch = 0.276\n",
            "14447/17600 (epoch 41), train_loss = 1.330, time/batch = 0.275\n",
            "14448/17600 (epoch 41), train_loss = 1.342, time/batch = 0.283\n",
            "14449/17600 (epoch 41), train_loss = 1.297, time/batch = 0.300\n",
            "14450/17600 (epoch 41), train_loss = 1.310, time/batch = 0.293\n",
            "14451/17600 (epoch 41), train_loss = 1.354, time/batch = 0.181\n",
            "14452/17600 (epoch 41), train_loss = 1.277, time/batch = 0.163\n",
            "14453/17600 (epoch 41), train_loss = 1.331, time/batch = 0.160\n",
            "14454/17600 (epoch 41), train_loss = 1.297, time/batch = 0.175\n",
            "14455/17600 (epoch 41), train_loss = 1.270, time/batch = 0.192\n",
            "14456/17600 (epoch 41), train_loss = 1.355, time/batch = 0.174\n",
            "14457/17600 (epoch 41), train_loss = 1.286, time/batch = 0.167\n",
            "14458/17600 (epoch 41), train_loss = 1.334, time/batch = 0.174\n",
            "14459/17600 (epoch 41), train_loss = 1.317, time/batch = 0.167\n",
            "14460/17600 (epoch 41), train_loss = 1.295, time/batch = 0.167\n",
            "14461/17600 (epoch 41), train_loss = 1.228, time/batch = 0.178\n",
            "14462/17600 (epoch 41), train_loss = 1.337, time/batch = 0.166\n",
            "14463/17600 (epoch 41), train_loss = 1.315, time/batch = 0.166\n",
            "14464/17600 (epoch 41), train_loss = 1.291, time/batch = 0.176\n",
            "14465/17600 (epoch 41), train_loss = 1.296, time/batch = 0.166\n",
            "14466/17600 (epoch 41), train_loss = 1.299, time/batch = 0.164\n",
            "14467/17600 (epoch 41), train_loss = 1.313, time/batch = 0.174\n",
            "14468/17600 (epoch 41), train_loss = 1.340, time/batch = 0.171\n",
            "14469/17600 (epoch 41), train_loss = 1.336, time/batch = 0.159\n",
            "14470/17600 (epoch 41), train_loss = 1.316, time/batch = 0.153\n",
            "14471/17600 (epoch 41), train_loss = 1.294, time/batch = 0.151\n",
            "14472/17600 (epoch 41), train_loss = 1.341, time/batch = 0.165\n",
            "14473/17600 (epoch 41), train_loss = 1.276, time/batch = 0.168\n",
            "14474/17600 (epoch 41), train_loss = 1.332, time/batch = 0.162\n",
            "14475/17600 (epoch 41), train_loss = 1.252, time/batch = 0.156\n",
            "14476/17600 (epoch 41), train_loss = 1.320, time/batch = 0.162\n",
            "14477/17600 (epoch 41), train_loss = 1.292, time/batch = 0.149\n",
            "14478/17600 (epoch 41), train_loss = 1.248, time/batch = 0.161\n",
            "14479/17600 (epoch 41), train_loss = 1.307, time/batch = 0.158\n",
            "14480/17600 (epoch 41), train_loss = 1.257, time/batch = 0.180\n",
            "14481/17600 (epoch 41), train_loss = 1.276, time/batch = 0.161\n",
            "14482/17600 (epoch 41), train_loss = 1.299, time/batch = 0.163\n",
            "14483/17600 (epoch 41), train_loss = 1.289, time/batch = 0.160\n",
            "14484/17600 (epoch 41), train_loss = 1.259, time/batch = 0.171\n",
            "14485/17600 (epoch 41), train_loss = 1.287, time/batch = 0.156\n",
            "14486/17600 (epoch 41), train_loss = 1.332, time/batch = 0.175\n",
            "14487/17600 (epoch 41), train_loss = 1.310, time/batch = 0.161\n",
            "14488/17600 (epoch 41), train_loss = 1.311, time/batch = 0.169\n",
            "14489/17600 (epoch 41), train_loss = 1.305, time/batch = 0.164\n",
            "14490/17600 (epoch 41), train_loss = 1.284, time/batch = 0.161\n",
            "14491/17600 (epoch 41), train_loss = 1.311, time/batch = 0.157\n",
            "14492/17600 (epoch 41), train_loss = 1.307, time/batch = 0.175\n",
            "14493/17600 (epoch 41), train_loss = 1.248, time/batch = 0.152\n",
            "14494/17600 (epoch 41), train_loss = 1.251, time/batch = 0.166\n",
            "14495/17600 (epoch 41), train_loss = 1.270, time/batch = 0.170\n",
            "14496/17600 (epoch 41), train_loss = 1.301, time/batch = 0.178\n",
            "14497/17600 (epoch 41), train_loss = 1.268, time/batch = 0.159\n",
            "14498/17600 (epoch 41), train_loss = 1.305, time/batch = 0.176\n",
            "14499/17600 (epoch 41), train_loss = 1.263, time/batch = 0.154\n",
            "14500/17600 (epoch 41), train_loss = 1.311, time/batch = 0.160\n",
            "14501/17600 (epoch 41), train_loss = 1.242, time/batch = 0.165\n",
            "14502/17600 (epoch 41), train_loss = 1.288, time/batch = 0.171\n",
            "14503/17600 (epoch 41), train_loss = 1.333, time/batch = 0.157\n",
            "14504/17600 (epoch 41), train_loss = 1.294, time/batch = 0.168\n",
            "14505/17600 (epoch 41), train_loss = 1.366, time/batch = 0.175\n",
            "14506/17600 (epoch 41), train_loss = 1.295, time/batch = 0.172\n",
            "14507/17600 (epoch 41), train_loss = 1.298, time/batch = 0.171\n",
            "14508/17600 (epoch 41), train_loss = 1.253, time/batch = 0.168\n",
            "14509/17600 (epoch 41), train_loss = 1.308, time/batch = 0.159\n",
            "14510/17600 (epoch 41), train_loss = 1.315, time/batch = 0.154\n",
            "14511/17600 (epoch 41), train_loss = 1.390, time/batch = 0.256\n",
            "14512/17600 (epoch 41), train_loss = 1.359, time/batch = 0.276\n",
            "14513/17600 (epoch 41), train_loss = 1.353, time/batch = 0.260\n",
            "14514/17600 (epoch 41), train_loss = 1.333, time/batch = 0.281\n",
            "14515/17600 (epoch 41), train_loss = 1.305, time/batch = 0.261\n",
            "14516/17600 (epoch 41), train_loss = 1.282, time/batch = 0.283\n",
            "14517/17600 (epoch 41), train_loss = 1.317, time/batch = 0.267\n",
            "14518/17600 (epoch 41), train_loss = 1.308, time/batch = 0.281\n",
            "14519/17600 (epoch 41), train_loss = 1.248, time/batch = 0.279\n",
            "14520/17600 (epoch 41), train_loss = 1.287, time/batch = 0.307\n",
            "14521/17600 (epoch 41), train_loss = 1.227, time/batch = 0.215\n",
            "14522/17600 (epoch 41), train_loss = 1.292, time/batch = 0.157\n",
            "14523/17600 (epoch 41), train_loss = 1.278, time/batch = 0.187\n",
            "14524/17600 (epoch 41), train_loss = 1.256, time/batch = 0.156\n",
            "14525/17600 (epoch 41), train_loss = 1.311, time/batch = 0.160\n",
            "14526/17600 (epoch 41), train_loss = 1.294, time/batch = 0.149\n",
            "14527/17600 (epoch 41), train_loss = 1.257, time/batch = 0.172\n",
            "14528/17600 (epoch 41), train_loss = 1.232, time/batch = 0.152\n",
            "14529/17600 (epoch 41), train_loss = 1.283, time/batch = 0.185\n",
            "14530/17600 (epoch 41), train_loss = 1.317, time/batch = 0.151\n",
            "14531/17600 (epoch 41), train_loss = 1.276, time/batch = 0.156\n",
            "14532/17600 (epoch 41), train_loss = 1.250, time/batch = 0.143\n",
            "14533/17600 (epoch 41), train_loss = 1.335, time/batch = 0.168\n",
            "14534/17600 (epoch 41), train_loss = 1.274, time/batch = 0.151\n",
            "14535/17600 (epoch 41), train_loss = 1.383, time/batch = 0.164\n",
            "14536/17600 (epoch 41), train_loss = 1.311, time/batch = 0.180\n",
            "14537/17600 (epoch 41), train_loss = 1.284, time/batch = 0.164\n",
            "14538/17600 (epoch 41), train_loss = 1.325, time/batch = 0.159\n",
            "14539/17600 (epoch 41), train_loss = 1.312, time/batch = 0.170\n",
            "14540/17600 (epoch 41), train_loss = 1.315, time/batch = 0.156\n",
            "14541/17600 (epoch 41), train_loss = 1.299, time/batch = 0.169\n",
            "14542/17600 (epoch 41), train_loss = 1.255, time/batch = 0.163\n",
            "14543/17600 (epoch 41), train_loss = 1.278, time/batch = 0.172\n",
            "14544/17600 (epoch 41), train_loss = 1.309, time/batch = 0.159\n",
            "14545/17600 (epoch 41), train_loss = 1.204, time/batch = 0.165\n",
            "14546/17600 (epoch 41), train_loss = 1.278, time/batch = 0.162\n",
            "14547/17600 (epoch 41), train_loss = 1.308, time/batch = 0.165\n",
            "14548/17600 (epoch 41), train_loss = 1.276, time/batch = 0.167\n",
            "14549/17600 (epoch 41), train_loss = 1.302, time/batch = 0.174\n",
            "14550/17600 (epoch 41), train_loss = 1.245, time/batch = 0.164\n",
            "14551/17600 (epoch 41), train_loss = 1.349, time/batch = 0.171\n",
            "14552/17600 (epoch 41), train_loss = 1.255, time/batch = 0.162\n",
            "14553/17600 (epoch 41), train_loss = 1.317, time/batch = 0.165\n",
            "14554/17600 (epoch 41), train_loss = 1.341, time/batch = 0.161\n",
            "14555/17600 (epoch 41), train_loss = 1.327, time/batch = 0.161\n",
            "14556/17600 (epoch 41), train_loss = 1.243, time/batch = 0.143\n",
            "14557/17600 (epoch 41), train_loss = 1.309, time/batch = 0.161\n",
            "14558/17600 (epoch 41), train_loss = 1.306, time/batch = 0.153\n",
            "14559/17600 (epoch 41), train_loss = 1.293, time/batch = 0.173\n",
            "14560/17600 (epoch 41), train_loss = 1.275, time/batch = 0.151\n",
            "14561/17600 (epoch 41), train_loss = 1.308, time/batch = 0.176\n",
            "14562/17600 (epoch 41), train_loss = 1.321, time/batch = 0.149\n",
            "14563/17600 (epoch 41), train_loss = 1.326, time/batch = 0.156\n",
            "14564/17600 (epoch 41), train_loss = 1.295, time/batch = 0.158\n",
            "14565/17600 (epoch 41), train_loss = 1.267, time/batch = 0.174\n",
            "14566/17600 (epoch 41), train_loss = 1.278, time/batch = 0.154\n",
            "14567/17600 (epoch 41), train_loss = 1.249, time/batch = 0.188\n",
            "14568/17600 (epoch 41), train_loss = 1.318, time/batch = 0.148\n",
            "14569/17600 (epoch 41), train_loss = 1.343, time/batch = 0.151\n",
            "14570/17600 (epoch 41), train_loss = 1.288, time/batch = 0.156\n",
            "14571/17600 (epoch 41), train_loss = 1.340, time/batch = 0.159\n",
            "14572/17600 (epoch 41), train_loss = 1.367, time/batch = 0.169\n",
            "14573/17600 (epoch 41), train_loss = 1.278, time/batch = 0.186\n",
            "14574/17600 (epoch 41), train_loss = 1.304, time/batch = 0.156\n",
            "14575/17600 (epoch 41), train_loss = 1.270, time/batch = 0.168\n",
            "14576/17600 (epoch 41), train_loss = 1.322, time/batch = 0.173\n",
            "14577/17600 (epoch 41), train_loss = 1.306, time/batch = 0.165\n",
            "14578/17600 (epoch 41), train_loss = 1.282, time/batch = 0.148\n",
            "14579/17600 (epoch 41), train_loss = 1.290, time/batch = 0.184\n",
            "14580/17600 (epoch 41), train_loss = 1.272, time/batch = 0.181\n",
            "14581/17600 (epoch 41), train_loss = 1.335, time/batch = 0.171\n",
            "14582/17600 (epoch 41), train_loss = 1.222, time/batch = 0.234\n",
            "14583/17600 (epoch 41), train_loss = 1.296, time/batch = 0.268\n",
            "14584/17600 (epoch 41), train_loss = 1.321, time/batch = 0.274\n",
            "14585/17600 (epoch 41), train_loss = 1.289, time/batch = 0.271\n",
            "14586/17600 (epoch 41), train_loss = 1.306, time/batch = 0.220\n",
            "14587/17600 (epoch 41), train_loss = 1.244, time/batch = 0.283\n",
            "14588/17600 (epoch 41), train_loss = 1.241, time/batch = 0.280\n",
            "14589/17600 (epoch 41), train_loss = 1.259, time/batch = 0.266\n",
            "14590/17600 (epoch 41), train_loss = 1.320, time/batch = 0.287\n",
            "14591/17600 (epoch 41), train_loss = 1.363, time/batch = 0.292\n",
            "14592/17600 (epoch 41), train_loss = 1.318, time/batch = 0.290\n",
            "14593/17600 (epoch 41), train_loss = 1.318, time/batch = 0.172\n",
            "14594/17600 (epoch 41), train_loss = 1.249, time/batch = 0.163\n",
            "14595/17600 (epoch 41), train_loss = 1.278, time/batch = 0.157\n",
            "14596/17600 (epoch 41), train_loss = 1.300, time/batch = 0.169\n",
            "14597/17600 (epoch 41), train_loss = 1.287, time/batch = 0.151\n",
            "14598/17600 (epoch 41), train_loss = 1.299, time/batch = 0.177\n",
            "14599/17600 (epoch 41), train_loss = 1.326, time/batch = 0.152\n",
            "14600/17600 (epoch 41), train_loss = 1.317, time/batch = 0.174\n",
            "14601/17600 (epoch 41), train_loss = 1.298, time/batch = 0.159\n",
            "14602/17600 (epoch 41), train_loss = 1.232, time/batch = 0.161\n",
            "14603/17600 (epoch 41), train_loss = 1.243, time/batch = 0.152\n",
            "14604/17600 (epoch 41), train_loss = 1.332, time/batch = 0.175\n",
            "14605/17600 (epoch 41), train_loss = 1.300, time/batch = 0.153\n",
            "14606/17600 (epoch 41), train_loss = 1.290, time/batch = 0.160\n",
            "14607/17600 (epoch 41), train_loss = 1.294, time/batch = 0.154\n",
            "14608/17600 (epoch 41), train_loss = 1.209, time/batch = 0.163\n",
            "14609/17600 (epoch 41), train_loss = 1.271, time/batch = 0.160\n",
            "14610/17600 (epoch 41), train_loss = 1.271, time/batch = 0.168\n",
            "14611/17600 (epoch 41), train_loss = 1.250, time/batch = 0.164\n",
            "14612/17600 (epoch 41), train_loss = 1.249, time/batch = 0.158\n",
            "14613/17600 (epoch 41), train_loss = 1.303, time/batch = 0.174\n",
            "14614/17600 (epoch 41), train_loss = 1.359, time/batch = 0.166\n",
            "14615/17600 (epoch 41), train_loss = 1.275, time/batch = 0.159\n",
            "14616/17600 (epoch 41), train_loss = 1.275, time/batch = 0.169\n",
            "14617/17600 (epoch 41), train_loss = 1.239, time/batch = 0.173\n",
            "14618/17600 (epoch 41), train_loss = 1.315, time/batch = 0.168\n",
            "14619/17600 (epoch 41), train_loss = 1.323, time/batch = 0.160\n",
            "14620/17600 (epoch 41), train_loss = 1.243, time/batch = 0.167\n",
            "14621/17600 (epoch 41), train_loss = 1.338, time/batch = 0.158\n",
            "14622/17600 (epoch 41), train_loss = 1.304, time/batch = 0.152\n",
            "14623/17600 (epoch 41), train_loss = 1.285, time/batch = 0.175\n",
            "14624/17600 (epoch 41), train_loss = 1.292, time/batch = 0.160\n",
            "14625/17600 (epoch 41), train_loss = 1.311, time/batch = 0.159\n",
            "14626/17600 (epoch 41), train_loss = 1.312, time/batch = 0.168\n",
            "14627/17600 (epoch 41), train_loss = 1.321, time/batch = 0.165\n",
            "14628/17600 (epoch 41), train_loss = 1.278, time/batch = 0.159\n",
            "14629/17600 (epoch 41), train_loss = 1.330, time/batch = 0.166\n",
            "14630/17600 (epoch 41), train_loss = 1.277, time/batch = 0.174\n",
            "14631/17600 (epoch 41), train_loss = 1.311, time/batch = 0.165\n",
            "14632/17600 (epoch 41), train_loss = 1.314, time/batch = 0.172\n",
            "14633/17600 (epoch 41), train_loss = 1.280, time/batch = 0.155\n",
            "14634/17600 (epoch 41), train_loss = 1.271, time/batch = 0.159\n",
            "14635/17600 (epoch 41), train_loss = 1.299, time/batch = 0.162\n",
            "14636/17600 (epoch 41), train_loss = 1.333, time/batch = 0.167\n",
            "14637/17600 (epoch 41), train_loss = 1.380, time/batch = 0.162\n",
            "14638/17600 (epoch 41), train_loss = 1.295, time/batch = 0.173\n",
            "14639/17600 (epoch 41), train_loss = 1.296, time/batch = 0.160\n",
            "14640/17600 (epoch 41), train_loss = 1.298, time/batch = 0.162\n",
            "14641/17600 (epoch 41), train_loss = 1.277, time/batch = 0.170\n",
            "14642/17600 (epoch 41), train_loss = 1.321, time/batch = 0.189\n",
            "14643/17600 (epoch 41), train_loss = 1.312, time/batch = 0.161\n",
            "14644/17600 (epoch 41), train_loss = 1.281, time/batch = 0.164\n",
            "14645/17600 (epoch 41), train_loss = 1.356, time/batch = 0.160\n",
            "14646/17600 (epoch 41), train_loss = 1.395, time/batch = 0.155\n",
            "14647/17600 (epoch 41), train_loss = 1.284, time/batch = 0.151\n",
            "14648/17600 (epoch 41), train_loss = 1.306, time/batch = 0.180\n",
            "14649/17600 (epoch 41), train_loss = 1.261, time/batch = 0.174\n",
            "14650/17600 (epoch 41), train_loss = 1.296, time/batch = 0.190\n",
            "14651/17600 (epoch 41), train_loss = 1.278, time/batch = 0.178\n",
            "14652/17600 (epoch 41), train_loss = 1.270, time/batch = 0.163\n",
            "14653/17600 (epoch 41), train_loss = 1.295, time/batch = 0.238\n",
            "14654/17600 (epoch 41), train_loss = 1.363, time/batch = 0.289\n",
            "14655/17600 (epoch 41), train_loss = 1.336, time/batch = 0.268\n",
            "14656/17600 (epoch 41), train_loss = 1.253, time/batch = 0.267\n",
            "14657/17600 (epoch 41), train_loss = 1.301, time/batch = 0.278\n",
            "14658/17600 (epoch 41), train_loss = 1.307, time/batch = 0.292\n",
            "14659/17600 (epoch 41), train_loss = 1.286, time/batch = 0.268\n",
            "14660/17600 (epoch 41), train_loss = 1.317, time/batch = 0.266\n",
            "14661/17600 (epoch 41), train_loss = 1.281, time/batch = 0.262\n",
            "14662/17600 (epoch 41), train_loss = 1.270, time/batch = 0.279\n",
            "14663/17600 (epoch 41), train_loss = 1.335, time/batch = 0.287\n",
            "14664/17600 (epoch 41), train_loss = 1.333, time/batch = 0.154\n",
            "14665/17600 (epoch 41), train_loss = 1.347, time/batch = 0.158\n",
            "14666/17600 (epoch 41), train_loss = 1.262, time/batch = 0.175\n",
            "14667/17600 (epoch 41), train_loss = 1.293, time/batch = 0.164\n",
            "14668/17600 (epoch 41), train_loss = 1.283, time/batch = 0.152\n",
            "14669/17600 (epoch 41), train_loss = 1.352, time/batch = 0.186\n",
            "14670/17600 (epoch 41), train_loss = 1.348, time/batch = 0.167\n",
            "14671/17600 (epoch 41), train_loss = 1.243, time/batch = 0.176\n",
            "14672/17600 (epoch 41), train_loss = 1.335, time/batch = 0.174\n",
            "14673/17600 (epoch 41), train_loss = 1.283, time/batch = 0.164\n",
            "14674/17600 (epoch 41), train_loss = 1.318, time/batch = 0.168\n",
            "14675/17600 (epoch 41), train_loss = 1.310, time/batch = 0.165\n",
            "14676/17600 (epoch 41), train_loss = 1.279, time/batch = 0.180\n",
            "14677/17600 (epoch 41), train_loss = 1.319, time/batch = 0.173\n",
            "14678/17600 (epoch 41), train_loss = 1.311, time/batch = 0.178\n",
            "14679/17600 (epoch 41), train_loss = 1.284, time/batch = 0.168\n",
            "14680/17600 (epoch 41), train_loss = 1.261, time/batch = 0.160\n",
            "14681/17600 (epoch 41), train_loss = 1.240, time/batch = 0.166\n",
            "14682/17600 (epoch 41), train_loss = 1.375, time/batch = 0.159\n",
            "14683/17600 (epoch 41), train_loss = 1.312, time/batch = 0.176\n",
            "14684/17600 (epoch 41), train_loss = 1.306, time/batch = 0.185\n",
            "14685/17600 (epoch 41), train_loss = 1.313, time/batch = 0.169\n",
            "14686/17600 (epoch 41), train_loss = 1.331, time/batch = 0.156\n",
            "14687/17600 (epoch 41), train_loss = 1.339, time/batch = 0.164\n",
            "14688/17600 (epoch 41), train_loss = 1.386, time/batch = 0.152\n",
            "14689/17600 (epoch 41), train_loss = 1.355, time/batch = 0.170\n",
            "14690/17600 (epoch 41), train_loss = 1.306, time/batch = 0.171\n",
            "14691/17600 (epoch 41), train_loss = 1.308, time/batch = 0.152\n",
            "14692/17600 (epoch 41), train_loss = 1.285, time/batch = 0.164\n",
            "14693/17600 (epoch 41), train_loss = 1.293, time/batch = 0.168\n",
            "14694/17600 (epoch 41), train_loss = 1.274, time/batch = 0.168\n",
            "14695/17600 (epoch 41), train_loss = 1.231, time/batch = 0.166\n",
            "14696/17600 (epoch 41), train_loss = 1.282, time/batch = 0.191\n",
            "14697/17600 (epoch 41), train_loss = 1.289, time/batch = 0.172\n",
            "14698/17600 (epoch 41), train_loss = 1.242, time/batch = 0.155\n",
            "14699/17600 (epoch 41), train_loss = 1.246, time/batch = 0.178\n",
            "14700/17600 (epoch 41), train_loss = 1.278, time/batch = 0.166\n",
            "14701/17600 (epoch 41), train_loss = 1.312, time/batch = 0.187\n",
            "14702/17600 (epoch 41), train_loss = 1.236, time/batch = 0.196\n",
            "14703/17600 (epoch 41), train_loss = 1.305, time/batch = 0.163\n",
            "14704/17600 (epoch 41), train_loss = 1.272, time/batch = 0.164\n",
            "14705/17600 (epoch 41), train_loss = 1.297, time/batch = 0.163\n",
            "14706/17600 (epoch 41), train_loss = 1.280, time/batch = 0.167\n",
            "14707/17600 (epoch 41), train_loss = 1.336, time/batch = 0.156\n",
            "14708/17600 (epoch 41), train_loss = 1.300, time/batch = 0.184\n",
            "14709/17600 (epoch 41), train_loss = 1.326, time/batch = 0.156\n",
            "14710/17600 (epoch 41), train_loss = 1.292, time/batch = 0.158\n",
            "14711/17600 (epoch 41), train_loss = 1.299, time/batch = 0.168\n",
            "14712/17600 (epoch 41), train_loss = 1.321, time/batch = 0.168\n",
            "14713/17600 (epoch 41), train_loss = 1.368, time/batch = 0.163\n",
            "14714/17600 (epoch 41), train_loss = 1.339, time/batch = 0.172\n",
            "14715/17600 (epoch 41), train_loss = 1.337, time/batch = 0.177\n",
            "14716/17600 (epoch 41), train_loss = 1.355, time/batch = 0.161\n",
            "14717/17600 (epoch 41), train_loss = 1.368, time/batch = 0.158\n",
            "14718/17600 (epoch 41), train_loss = 1.313, time/batch = 0.161\n",
            "14719/17600 (epoch 41), train_loss = 1.360, time/batch = 0.157\n",
            "14720/17600 (epoch 41), train_loss = 1.297, time/batch = 0.170\n",
            "14721/17600 (epoch 41), train_loss = 1.293, time/batch = 0.177\n",
            "14722/17600 (epoch 41), train_loss = 1.359, time/batch = 0.179\n",
            "14723/17600 (epoch 41), train_loss = 1.264, time/batch = 0.238\n",
            "14724/17600 (epoch 41), train_loss = 1.259, time/batch = 0.308\n",
            "14725/17600 (epoch 41), train_loss = 1.333, time/batch = 0.289\n",
            "14726/17600 (epoch 41), train_loss = 1.282, time/batch = 0.294\n",
            "14727/17600 (epoch 41), train_loss = 1.342, time/batch = 0.309\n",
            "14728/17600 (epoch 41), train_loss = 1.287, time/batch = 0.304\n",
            "14729/17600 (epoch 41), train_loss = 1.316, time/batch = 0.312\n",
            "14730/17600 (epoch 41), train_loss = 1.300, time/batch = 0.288\n",
            "14731/17600 (epoch 41), train_loss = 1.314, time/batch = 0.277\n",
            "14732/17600 (epoch 41), train_loss = 1.345, time/batch = 0.293\n",
            "14733/17600 (epoch 41), train_loss = 1.302, time/batch = 0.171\n",
            "14734/17600 (epoch 41), train_loss = 1.371, time/batch = 0.169\n",
            "14735/17600 (epoch 41), train_loss = 1.364, time/batch = 0.155\n",
            "14736/17600 (epoch 41), train_loss = 1.374, time/batch = 0.173\n",
            "14737/17600 (epoch 41), train_loss = 1.288, time/batch = 0.166\n",
            "14738/17600 (epoch 41), train_loss = 1.324, time/batch = 0.167\n",
            "14739/17600 (epoch 41), train_loss = 1.329, time/batch = 0.175\n",
            "14740/17600 (epoch 41), train_loss = 1.328, time/batch = 0.160\n",
            "14741/17600 (epoch 41), train_loss = 1.331, time/batch = 0.160\n",
            "14742/17600 (epoch 41), train_loss = 1.367, time/batch = 0.171\n",
            "14743/17600 (epoch 41), train_loss = 1.317, time/batch = 0.168\n",
            "14744/17600 (epoch 41), train_loss = 1.349, time/batch = 0.168\n",
            "14745/17600 (epoch 41), train_loss = 1.285, time/batch = 0.177\n",
            "14746/17600 (epoch 41), train_loss = 1.300, time/batch = 0.168\n",
            "14747/17600 (epoch 41), train_loss = 1.268, time/batch = 0.153\n",
            "14748/17600 (epoch 41), train_loss = 1.366, time/batch = 0.162\n",
            "14749/17600 (epoch 41), train_loss = 1.353, time/batch = 0.162\n",
            "14750/17600 (epoch 41), train_loss = 1.352, time/batch = 0.168\n",
            "14751/17600 (epoch 41), train_loss = 1.333, time/batch = 0.179\n",
            "14752/17600 (epoch 41), train_loss = 1.373, time/batch = 0.148\n",
            "14753/17600 (epoch 41), train_loss = 1.339, time/batch = 0.158\n",
            "14754/17600 (epoch 41), train_loss = 1.394, time/batch = 0.174\n",
            "14755/17600 (epoch 41), train_loss = 1.364, time/batch = 0.149\n",
            "14756/17600 (epoch 41), train_loss = 1.348, time/batch = 0.157\n",
            "14757/17600 (epoch 41), train_loss = 1.299, time/batch = 0.171\n",
            "14758/17600 (epoch 41), train_loss = 1.376, time/batch = 0.180\n",
            "14759/17600 (epoch 41), train_loss = 1.354, time/batch = 0.163\n",
            "14760/17600 (epoch 41), train_loss = 1.390, time/batch = 0.159\n",
            "14761/17600 (epoch 41), train_loss = 1.358, time/batch = 0.165\n",
            "14762/17600 (epoch 41), train_loss = 1.280, time/batch = 0.162\n",
            "14763/17600 (epoch 41), train_loss = 1.308, time/batch = 0.165\n",
            "14764/17600 (epoch 41), train_loss = 1.357, time/batch = 0.159\n",
            "14765/17600 (epoch 41), train_loss = 1.367, time/batch = 0.165\n",
            "14766/17600 (epoch 41), train_loss = 1.324, time/batch = 0.150\n",
            "14767/17600 (epoch 41), train_loss = 1.316, time/batch = 0.177\n",
            "14768/17600 (epoch 41), train_loss = 1.324, time/batch = 0.152\n",
            "14769/17600 (epoch 41), train_loss = 1.362, time/batch = 0.166\n",
            "14770/17600 (epoch 41), train_loss = 1.309, time/batch = 0.171\n",
            "14771/17600 (epoch 41), train_loss = 1.339, time/batch = 0.154\n",
            "14772/17600 (epoch 41), train_loss = 1.306, time/batch = 0.155\n",
            "14773/17600 (epoch 41), train_loss = 1.348, time/batch = 0.165\n",
            "14774/17600 (epoch 41), train_loss = 1.357, time/batch = 0.160\n",
            "14775/17600 (epoch 41), train_loss = 1.321, time/batch = 0.164\n",
            "14776/17600 (epoch 41), train_loss = 1.300, time/batch = 0.172\n",
            "14777/17600 (epoch 41), train_loss = 1.318, time/batch = 0.172\n",
            "14778/17600 (epoch 41), train_loss = 1.341, time/batch = 0.161\n",
            "14779/17600 (epoch 41), train_loss = 1.314, time/batch = 0.166\n",
            "14780/17600 (epoch 41), train_loss = 1.338, time/batch = 0.160\n",
            "14781/17600 (epoch 41), train_loss = 1.322, time/batch = 0.167\n",
            "14782/17600 (epoch 41), train_loss = 1.313, time/batch = 0.165\n",
            "14783/17600 (epoch 41), train_loss = 1.305, time/batch = 0.171\n",
            "14784/17600 (epoch 42), train_loss = 1.108, time/batch = 0.177\n",
            "14785/17600 (epoch 42), train_loss = 1.336, time/batch = 0.162\n",
            "14786/17600 (epoch 42), train_loss = 1.330, time/batch = 0.166\n",
            "14787/17600 (epoch 42), train_loss = 1.275, time/batch = 0.189\n",
            "14788/17600 (epoch 42), train_loss = 1.370, time/batch = 0.207\n",
            "14789/17600 (epoch 42), train_loss = 1.346, time/batch = 0.168\n",
            "14790/17600 (epoch 42), train_loss = 1.292, time/batch = 0.165\n",
            "14791/17600 (epoch 42), train_loss = 1.279, time/batch = 0.190\n",
            "14792/17600 (epoch 42), train_loss = 1.320, time/batch = 0.259\n",
            "14793/17600 (epoch 42), train_loss = 1.373, time/batch = 0.323\n",
            "14794/17600 (epoch 42), train_loss = 1.302, time/batch = 0.275\n",
            "14795/17600 (epoch 42), train_loss = 1.316, time/batch = 0.284\n",
            "14796/17600 (epoch 42), train_loss = 1.298, time/batch = 0.271\n",
            "14797/17600 (epoch 42), train_loss = 1.283, time/batch = 0.278\n",
            "14798/17600 (epoch 42), train_loss = 1.264, time/batch = 0.277\n",
            "14799/17600 (epoch 42), train_loss = 1.328, time/batch = 0.268\n",
            "14800/17600 (epoch 42), train_loss = 1.340, time/batch = 0.279\n",
            "14801/17600 (epoch 42), train_loss = 1.295, time/batch = 0.291\n",
            "14802/17600 (epoch 42), train_loss = 1.307, time/batch = 0.204\n",
            "14803/17600 (epoch 42), train_loss = 1.352, time/batch = 0.172\n",
            "14804/17600 (epoch 42), train_loss = 1.275, time/batch = 0.190\n",
            "14805/17600 (epoch 42), train_loss = 1.329, time/batch = 0.171\n",
            "14806/17600 (epoch 42), train_loss = 1.295, time/batch = 0.164\n",
            "14807/17600 (epoch 42), train_loss = 1.268, time/batch = 0.172\n",
            "14808/17600 (epoch 42), train_loss = 1.353, time/batch = 0.152\n",
            "14809/17600 (epoch 42), train_loss = 1.284, time/batch = 0.166\n",
            "14810/17600 (epoch 42), train_loss = 1.331, time/batch = 0.177\n",
            "14811/17600 (epoch 42), train_loss = 1.315, time/batch = 0.173\n",
            "14812/17600 (epoch 42), train_loss = 1.293, time/batch = 0.183\n",
            "14813/17600 (epoch 42), train_loss = 1.227, time/batch = 0.170\n",
            "14814/17600 (epoch 42), train_loss = 1.335, time/batch = 0.164\n",
            "14815/17600 (epoch 42), train_loss = 1.313, time/batch = 0.172\n",
            "14816/17600 (epoch 42), train_loss = 1.289, time/batch = 0.169\n",
            "14817/17600 (epoch 42), train_loss = 1.293, time/batch = 0.163\n",
            "14818/17600 (epoch 42), train_loss = 1.297, time/batch = 0.171\n",
            "14819/17600 (epoch 42), train_loss = 1.311, time/batch = 0.165\n",
            "14820/17600 (epoch 42), train_loss = 1.338, time/batch = 0.166\n",
            "14821/17600 (epoch 42), train_loss = 1.334, time/batch = 0.159\n",
            "14822/17600 (epoch 42), train_loss = 1.314, time/batch = 0.176\n",
            "14823/17600 (epoch 42), train_loss = 1.292, time/batch = 0.168\n",
            "14824/17600 (epoch 42), train_loss = 1.338, time/batch = 0.182\n",
            "14825/17600 (epoch 42), train_loss = 1.274, time/batch = 0.159\n",
            "14826/17600 (epoch 42), train_loss = 1.330, time/batch = 0.176\n",
            "14827/17600 (epoch 42), train_loss = 1.250, time/batch = 0.165\n",
            "14828/17600 (epoch 42), train_loss = 1.318, time/batch = 0.177\n",
            "14829/17600 (epoch 42), train_loss = 1.290, time/batch = 0.155\n",
            "14830/17600 (epoch 42), train_loss = 1.245, time/batch = 0.179\n",
            "14831/17600 (epoch 42), train_loss = 1.305, time/batch = 0.155\n",
            "14832/17600 (epoch 42), train_loss = 1.255, time/batch = 0.169\n",
            "14833/17600 (epoch 42), train_loss = 1.273, time/batch = 0.162\n",
            "14834/17600 (epoch 42), train_loss = 1.297, time/batch = 0.177\n",
            "14835/17600 (epoch 42), train_loss = 1.287, time/batch = 0.169\n",
            "14836/17600 (epoch 42), train_loss = 1.258, time/batch = 0.171\n",
            "14837/17600 (epoch 42), train_loss = 1.286, time/batch = 0.162\n",
            "14838/17600 (epoch 42), train_loss = 1.330, time/batch = 0.181\n",
            "14839/17600 (epoch 42), train_loss = 1.308, time/batch = 0.154\n",
            "14840/17600 (epoch 42), train_loss = 1.309, time/batch = 0.169\n",
            "14841/17600 (epoch 42), train_loss = 1.303, time/batch = 0.174\n",
            "14842/17600 (epoch 42), train_loss = 1.282, time/batch = 0.165\n",
            "14843/17600 (epoch 42), train_loss = 1.309, time/batch = 0.154\n",
            "14844/17600 (epoch 42), train_loss = 1.305, time/batch = 0.167\n",
            "14845/17600 (epoch 42), train_loss = 1.247, time/batch = 0.155\n",
            "14846/17600 (epoch 42), train_loss = 1.249, time/batch = 0.168\n",
            "14847/17600 (epoch 42), train_loss = 1.268, time/batch = 0.171\n",
            "14848/17600 (epoch 42), train_loss = 1.299, time/batch = 0.169\n",
            "14849/17600 (epoch 42), train_loss = 1.265, time/batch = 0.158\n",
            "14850/17600 (epoch 42), train_loss = 1.303, time/batch = 0.162\n",
            "14851/17600 (epoch 42), train_loss = 1.261, time/batch = 0.159\n",
            "14852/17600 (epoch 42), train_loss = 1.309, time/batch = 0.166\n",
            "14853/17600 (epoch 42), train_loss = 1.241, time/batch = 0.168\n",
            "14854/17600 (epoch 42), train_loss = 1.286, time/batch = 0.171\n",
            "14855/17600 (epoch 42), train_loss = 1.331, time/batch = 0.154\n",
            "14856/17600 (epoch 42), train_loss = 1.293, time/batch = 0.165\n",
            "14857/17600 (epoch 42), train_loss = 1.363, time/batch = 0.164\n",
            "14858/17600 (epoch 42), train_loss = 1.293, time/batch = 0.169\n",
            "14859/17600 (epoch 42), train_loss = 1.295, time/batch = 0.164\n",
            "14860/17600 (epoch 42), train_loss = 1.251, time/batch = 0.185\n",
            "14861/17600 (epoch 42), train_loss = 1.306, time/batch = 0.200\n",
            "14862/17600 (epoch 42), train_loss = 1.313, time/batch = 0.271\n",
            "14863/17600 (epoch 42), train_loss = 1.389, time/batch = 0.251\n",
            "14864/17600 (epoch 42), train_loss = 1.358, time/batch = 0.292\n",
            "14865/17600 (epoch 42), train_loss = 1.351, time/batch = 0.261\n",
            "14866/17600 (epoch 42), train_loss = 1.332, time/batch = 0.266\n",
            "14867/17600 (epoch 42), train_loss = 1.304, time/batch = 0.272\n",
            "14868/17600 (epoch 42), train_loss = 1.281, time/batch = 0.300\n",
            "14869/17600 (epoch 42), train_loss = 1.315, time/batch = 0.276\n",
            "14870/17600 (epoch 42), train_loss = 1.306, time/batch = 0.276\n",
            "14871/17600 (epoch 42), train_loss = 1.245, time/batch = 0.295\n",
            "14872/17600 (epoch 42), train_loss = 1.285, time/batch = 0.199\n",
            "14873/17600 (epoch 42), train_loss = 1.225, time/batch = 0.175\n",
            "14874/17600 (epoch 42), train_loss = 1.290, time/batch = 0.153\n",
            "14875/17600 (epoch 42), train_loss = 1.276, time/batch = 0.171\n",
            "14876/17600 (epoch 42), train_loss = 1.254, time/batch = 0.157\n",
            "14877/17600 (epoch 42), train_loss = 1.309, time/batch = 0.178\n",
            "14878/17600 (epoch 42), train_loss = 1.292, time/batch = 0.152\n",
            "14879/17600 (epoch 42), train_loss = 1.255, time/batch = 0.151\n",
            "14880/17600 (epoch 42), train_loss = 1.230, time/batch = 0.150\n",
            "14881/17600 (epoch 42), train_loss = 1.281, time/batch = 0.159\n",
            "14882/17600 (epoch 42), train_loss = 1.315, time/batch = 0.162\n",
            "14883/17600 (epoch 42), train_loss = 1.274, time/batch = 0.152\n",
            "14884/17600 (epoch 42), train_loss = 1.248, time/batch = 0.177\n",
            "14885/17600 (epoch 42), train_loss = 1.333, time/batch = 0.150\n",
            "14886/17600 (epoch 42), train_loss = 1.272, time/batch = 0.172\n",
            "14887/17600 (epoch 42), train_loss = 1.381, time/batch = 0.162\n",
            "14888/17600 (epoch 42), train_loss = 1.309, time/batch = 0.168\n",
            "14889/17600 (epoch 42), train_loss = 1.283, time/batch = 0.156\n",
            "14890/17600 (epoch 42), train_loss = 1.324, time/batch = 0.179\n",
            "14891/17600 (epoch 42), train_loss = 1.310, time/batch = 0.150\n",
            "14892/17600 (epoch 42), train_loss = 1.313, time/batch = 0.168\n",
            "14893/17600 (epoch 42), train_loss = 1.297, time/batch = 0.157\n",
            "14894/17600 (epoch 42), train_loss = 1.253, time/batch = 0.158\n",
            "14895/17600 (epoch 42), train_loss = 1.276, time/batch = 0.143\n",
            "14896/17600 (epoch 42), train_loss = 1.307, time/batch = 0.164\n",
            "14897/17600 (epoch 42), train_loss = 1.202, time/batch = 0.175\n",
            "14898/17600 (epoch 42), train_loss = 1.277, time/batch = 0.179\n",
            "14899/17600 (epoch 42), train_loss = 1.306, time/batch = 0.157\n",
            "14900/17600 (epoch 42), train_loss = 1.274, time/batch = 0.162\n",
            "14901/17600 (epoch 42), train_loss = 1.300, time/batch = 0.156\n",
            "14902/17600 (epoch 42), train_loss = 1.243, time/batch = 0.164\n",
            "14903/17600 (epoch 42), train_loss = 1.347, time/batch = 0.180\n",
            "14904/17600 (epoch 42), train_loss = 1.253, time/batch = 0.146\n",
            "14905/17600 (epoch 42), train_loss = 1.315, time/batch = 0.170\n",
            "14906/17600 (epoch 42), train_loss = 1.339, time/batch = 0.150\n",
            "14907/17600 (epoch 42), train_loss = 1.325, time/batch = 0.164\n",
            "14908/17600 (epoch 42), train_loss = 1.241, time/batch = 0.161\n",
            "14909/17600 (epoch 42), train_loss = 1.307, time/batch = 0.182\n",
            "14910/17600 (epoch 42), train_loss = 1.304, time/batch = 0.160\n",
            "14911/17600 (epoch 42), train_loss = 1.291, time/batch = 0.158\n",
            "14912/17600 (epoch 42), train_loss = 1.273, time/batch = 0.158\n",
            "14913/17600 (epoch 42), train_loss = 1.306, time/batch = 0.171\n",
            "14914/17600 (epoch 42), train_loss = 1.319, time/batch = 0.154\n",
            "14915/17600 (epoch 42), train_loss = 1.324, time/batch = 0.156\n",
            "14916/17600 (epoch 42), train_loss = 1.293, time/batch = 0.186\n",
            "14917/17600 (epoch 42), train_loss = 1.265, time/batch = 0.163\n",
            "14918/17600 (epoch 42), train_loss = 1.276, time/batch = 0.174\n",
            "14919/17600 (epoch 42), train_loss = 1.247, time/batch = 0.154\n",
            "14920/17600 (epoch 42), train_loss = 1.315, time/batch = 0.158\n",
            "14921/17600 (epoch 42), train_loss = 1.341, time/batch = 0.159\n",
            "14922/17600 (epoch 42), train_loss = 1.286, time/batch = 0.170\n",
            "14923/17600 (epoch 42), train_loss = 1.338, time/batch = 0.145\n",
            "14924/17600 (epoch 42), train_loss = 1.365, time/batch = 0.155\n",
            "14925/17600 (epoch 42), train_loss = 1.276, time/batch = 0.152\n",
            "14926/17600 (epoch 42), train_loss = 1.302, time/batch = 0.159\n",
            "14927/17600 (epoch 42), train_loss = 1.268, time/batch = 0.146\n",
            "14928/17600 (epoch 42), train_loss = 1.320, time/batch = 0.156\n",
            "14929/17600 (epoch 42), train_loss = 1.304, time/batch = 0.171\n",
            "14930/17600 (epoch 42), train_loss = 1.280, time/batch = 0.161\n",
            "14931/17600 (epoch 42), train_loss = 1.288, time/batch = 0.150\n",
            "14932/17600 (epoch 42), train_loss = 1.270, time/batch = 0.161\n",
            "14933/17600 (epoch 42), train_loss = 1.333, time/batch = 0.180\n",
            "14934/17600 (epoch 42), train_loss = 1.220, time/batch = 0.290\n",
            "14935/17600 (epoch 42), train_loss = 1.294, time/batch = 0.277\n",
            "14936/17600 (epoch 42), train_loss = 1.319, time/batch = 0.271\n",
            "14937/17600 (epoch 42), train_loss = 1.287, time/batch = 0.260\n",
            "14938/17600 (epoch 42), train_loss = 1.304, time/batch = 0.267\n",
            "14939/17600 (epoch 42), train_loss = 1.242, time/batch = 0.280\n",
            "14940/17600 (epoch 42), train_loss = 1.239, time/batch = 0.301\n",
            "14941/17600 (epoch 42), train_loss = 1.257, time/batch = 0.270\n",
            "14942/17600 (epoch 42), train_loss = 1.318, time/batch = 0.317\n",
            "14943/17600 (epoch 42), train_loss = 1.361, time/batch = 0.284\n",
            "14944/17600 (epoch 42), train_loss = 1.316, time/batch = 0.166\n",
            "14945/17600 (epoch 42), train_loss = 1.316, time/batch = 0.172\n",
            "14946/17600 (epoch 42), train_loss = 1.247, time/batch = 0.161\n",
            "14947/17600 (epoch 42), train_loss = 1.276, time/batch = 0.181\n",
            "14948/17600 (epoch 42), train_loss = 1.298, time/batch = 0.143\n",
            "14949/17600 (epoch 42), train_loss = 1.285, time/batch = 0.158\n",
            "14950/17600 (epoch 42), train_loss = 1.297, time/batch = 0.168\n",
            "14951/17600 (epoch 42), train_loss = 1.323, time/batch = 0.153\n",
            "14952/17600 (epoch 42), train_loss = 1.316, time/batch = 0.153\n",
            "14953/17600 (epoch 42), train_loss = 1.296, time/batch = 0.157\n",
            "14954/17600 (epoch 42), train_loss = 1.229, time/batch = 0.162\n",
            "14955/17600 (epoch 42), train_loss = 1.241, time/batch = 0.156\n",
            "14956/17600 (epoch 42), train_loss = 1.330, time/batch = 0.169\n",
            "14957/17600 (epoch 42), train_loss = 1.298, time/batch = 0.151\n",
            "14958/17600 (epoch 42), train_loss = 1.289, time/batch = 0.171\n",
            "14959/17600 (epoch 42), train_loss = 1.292, time/batch = 0.151\n",
            "14960/17600 (epoch 42), train_loss = 1.206, time/batch = 0.171\n",
            "14961/17600 (epoch 42), train_loss = 1.269, time/batch = 0.159\n",
            "14962/17600 (epoch 42), train_loss = 1.269, time/batch = 0.163\n",
            "14963/17600 (epoch 42), train_loss = 1.248, time/batch = 0.163\n",
            "14964/17600 (epoch 42), train_loss = 1.247, time/batch = 0.160\n",
            "14965/17600 (epoch 42), train_loss = 1.301, time/batch = 0.153\n",
            "14966/17600 (epoch 42), train_loss = 1.357, time/batch = 0.176\n",
            "14967/17600 (epoch 42), train_loss = 1.273, time/batch = 0.157\n",
            "14968/17600 (epoch 42), train_loss = 1.273, time/batch = 0.155\n",
            "14969/17600 (epoch 42), train_loss = 1.237, time/batch = 0.167\n",
            "14970/17600 (epoch 42), train_loss = 1.312, time/batch = 0.153\n",
            "14971/17600 (epoch 42), train_loss = 1.322, time/batch = 0.144\n",
            "14972/17600 (epoch 42), train_loss = 1.241, time/batch = 0.174\n",
            "14973/17600 (epoch 42), train_loss = 1.336, time/batch = 0.193\n",
            "14974/17600 (epoch 42), train_loss = 1.303, time/batch = 0.164\n",
            "14975/17600 (epoch 42), train_loss = 1.283, time/batch = 0.164\n",
            "14976/17600 (epoch 42), train_loss = 1.291, time/batch = 0.163\n",
            "14977/17600 (epoch 42), train_loss = 1.309, time/batch = 0.150\n",
            "14978/17600 (epoch 42), train_loss = 1.310, time/batch = 0.160\n",
            "14979/17600 (epoch 42), train_loss = 1.320, time/batch = 0.163\n",
            "14980/17600 (epoch 42), train_loss = 1.276, time/batch = 0.181\n",
            "14981/17600 (epoch 42), train_loss = 1.328, time/batch = 0.161\n",
            "14982/17600 (epoch 42), train_loss = 1.275, time/batch = 0.160\n",
            "14983/17600 (epoch 42), train_loss = 1.309, time/batch = 0.156\n",
            "14984/17600 (epoch 42), train_loss = 1.312, time/batch = 0.170\n",
            "14985/17600 (epoch 42), train_loss = 1.278, time/batch = 0.178\n",
            "14986/17600 (epoch 42), train_loss = 1.268, time/batch = 0.165\n",
            "14987/17600 (epoch 42), train_loss = 1.297, time/batch = 0.160\n",
            "14988/17600 (epoch 42), train_loss = 1.331, time/batch = 0.177\n",
            "14989/17600 (epoch 42), train_loss = 1.378, time/batch = 0.147\n",
            "14990/17600 (epoch 42), train_loss = 1.293, time/batch = 0.165\n",
            "14991/17600 (epoch 42), train_loss = 1.294, time/batch = 0.167\n",
            "14992/17600 (epoch 42), train_loss = 1.296, time/batch = 0.168\n",
            "14993/17600 (epoch 42), train_loss = 1.275, time/batch = 0.157\n",
            "14994/17600 (epoch 42), train_loss = 1.319, time/batch = 0.172\n",
            "14995/17600 (epoch 42), train_loss = 1.311, time/batch = 0.140\n",
            "14996/17600 (epoch 42), train_loss = 1.280, time/batch = 0.163\n",
            "14997/17600 (epoch 42), train_loss = 1.354, time/batch = 0.166\n",
            "14998/17600 (epoch 42), train_loss = 1.393, time/batch = 0.183\n",
            "14999/17600 (epoch 42), train_loss = 1.282, time/batch = 0.153\n",
            "15000/17600 (epoch 42), train_loss = 1.303, time/batch = 0.160\n",
            "Model saved to checkpoints/dracula/dracula!\n",
            "15001/17600 (epoch 42), train_loss = 1.259, time/batch = 0.178\n",
            "15002/17600 (epoch 42), train_loss = 1.293, time/batch = 0.162\n",
            "15003/17600 (epoch 42), train_loss = 1.276, time/batch = 0.155\n",
            "15004/17600 (epoch 42), train_loss = 1.268, time/batch = 0.150\n",
            "15005/17600 (epoch 42), train_loss = 1.293, time/batch = 0.158\n",
            "15006/17600 (epoch 42), train_loss = 1.361, time/batch = 0.161\n",
            "15007/17600 (epoch 42), train_loss = 1.334, time/batch = 0.172\n",
            "15008/17600 (epoch 42), train_loss = 1.251, time/batch = 0.171\n",
            "15009/17600 (epoch 42), train_loss = 1.299, time/batch = 0.146\n",
            "15010/17600 (epoch 42), train_loss = 1.304, time/batch = 0.158\n",
            "15011/17600 (epoch 42), train_loss = 1.283, time/batch = 0.143\n",
            "15012/17600 (epoch 42), train_loss = 1.314, time/batch = 0.167\n",
            "15013/17600 (epoch 42), train_loss = 1.279, time/batch = 0.157\n",
            "15014/17600 (epoch 42), train_loss = 1.268, time/batch = 0.175\n",
            "15015/17600 (epoch 42), train_loss = 1.333, time/batch = 0.161\n",
            "15016/17600 (epoch 42), train_loss = 1.331, time/batch = 0.152\n",
            "15017/17600 (epoch 42), train_loss = 1.345, time/batch = 0.168\n",
            "15018/17600 (epoch 42), train_loss = 1.260, time/batch = 0.164\n",
            "15019/17600 (epoch 42), train_loss = 1.291, time/batch = 0.163\n",
            "15020/17600 (epoch 42), train_loss = 1.281, time/batch = 0.171\n",
            "15021/17600 (epoch 42), train_loss = 1.350, time/batch = 0.162\n",
            "15022/17600 (epoch 42), train_loss = 1.347, time/batch = 0.153\n",
            "15023/17600 (epoch 42), train_loss = 1.241, time/batch = 0.180\n",
            "15024/17600 (epoch 42), train_loss = 1.333, time/batch = 0.146\n",
            "15025/17600 (epoch 42), train_loss = 1.281, time/batch = 0.166\n",
            "15026/17600 (epoch 42), train_loss = 1.317, time/batch = 0.160\n",
            "15027/17600 (epoch 42), train_loss = 1.309, time/batch = 0.158\n",
            "15028/17600 (epoch 42), train_loss = 1.277, time/batch = 0.158\n",
            "15029/17600 (epoch 42), train_loss = 1.317, time/batch = 0.172\n",
            "15030/17600 (epoch 42), train_loss = 1.309, time/batch = 0.161\n",
            "15031/17600 (epoch 42), train_loss = 1.282, time/batch = 0.169\n",
            "15032/17600 (epoch 42), train_loss = 1.259, time/batch = 0.172\n",
            "15033/17600 (epoch 42), train_loss = 1.238, time/batch = 0.170\n",
            "15034/17600 (epoch 42), train_loss = 1.373, time/batch = 0.161\n",
            "15035/17600 (epoch 42), train_loss = 1.311, time/batch = 0.152\n",
            "15036/17600 (epoch 42), train_loss = 1.304, time/batch = 0.172\n",
            "15037/17600 (epoch 42), train_loss = 1.311, time/batch = 0.157\n",
            "15038/17600 (epoch 42), train_loss = 1.329, time/batch = 0.162\n",
            "15039/17600 (epoch 42), train_loss = 1.337, time/batch = 0.167\n",
            "15040/17600 (epoch 42), train_loss = 1.384, time/batch = 0.161\n",
            "15041/17600 (epoch 42), train_loss = 1.353, time/batch = 0.152\n",
            "15042/17600 (epoch 42), train_loss = 1.305, time/batch = 0.166\n",
            "15043/17600 (epoch 42), train_loss = 1.307, time/batch = 0.157\n",
            "15044/17600 (epoch 42), train_loss = 1.283, time/batch = 0.157\n",
            "15045/17600 (epoch 42), train_loss = 1.291, time/batch = 0.168\n",
            "15046/17600 (epoch 42), train_loss = 1.272, time/batch = 0.159\n",
            "15047/17600 (epoch 42), train_loss = 1.229, time/batch = 0.219\n",
            "15048/17600 (epoch 42), train_loss = 1.280, time/batch = 0.262\n",
            "15049/17600 (epoch 42), train_loss = 1.287, time/batch = 0.286\n",
            "15050/17600 (epoch 42), train_loss = 1.240, time/batch = 0.295\n",
            "15051/17600 (epoch 42), train_loss = 1.244, time/batch = 0.284\n",
            "15052/17600 (epoch 42), train_loss = 1.276, time/batch = 0.291\n",
            "15053/17600 (epoch 42), train_loss = 1.310, time/batch = 0.278\n",
            "15054/17600 (epoch 42), train_loss = 1.234, time/batch = 0.296\n",
            "15055/17600 (epoch 42), train_loss = 1.303, time/batch = 0.328\n",
            "15056/17600 (epoch 42), train_loss = 1.270, time/batch = 0.277\n",
            "15057/17600 (epoch 42), train_loss = 1.295, time/batch = 0.221\n",
            "15058/17600 (epoch 42), train_loss = 1.279, time/batch = 0.165\n",
            "15059/17600 (epoch 42), train_loss = 1.333, time/batch = 0.151\n",
            "15060/17600 (epoch 42), train_loss = 1.298, time/batch = 0.168\n",
            "15061/17600 (epoch 42), train_loss = 1.324, time/batch = 0.151\n",
            "15062/17600 (epoch 42), train_loss = 1.290, time/batch = 0.177\n",
            "15063/17600 (epoch 42), train_loss = 1.296, time/batch = 0.167\n",
            "15064/17600 (epoch 42), train_loss = 1.319, time/batch = 0.171\n",
            "15065/17600 (epoch 42), train_loss = 1.366, time/batch = 0.154\n",
            "15066/17600 (epoch 42), train_loss = 1.337, time/batch = 0.168\n",
            "15067/17600 (epoch 42), train_loss = 1.336, time/batch = 0.143\n",
            "15068/17600 (epoch 42), train_loss = 1.353, time/batch = 0.177\n",
            "15069/17600 (epoch 42), train_loss = 1.366, time/batch = 0.150\n",
            "15070/17600 (epoch 42), train_loss = 1.312, time/batch = 0.186\n",
            "15071/17600 (epoch 42), train_loss = 1.358, time/batch = 0.152\n",
            "15072/17600 (epoch 42), train_loss = 1.295, time/batch = 0.161\n",
            "15073/17600 (epoch 42), train_loss = 1.291, time/batch = 0.154\n",
            "15074/17600 (epoch 42), train_loss = 1.357, time/batch = 0.168\n",
            "15075/17600 (epoch 42), train_loss = 1.262, time/batch = 0.154\n",
            "15076/17600 (epoch 42), train_loss = 1.257, time/batch = 0.191\n",
            "15077/17600 (epoch 42), train_loss = 1.331, time/batch = 0.166\n",
            "15078/17600 (epoch 42), train_loss = 1.280, time/batch = 0.161\n",
            "15079/17600 (epoch 42), train_loss = 1.340, time/batch = 0.157\n",
            "15080/17600 (epoch 42), train_loss = 1.285, time/batch = 0.168\n",
            "15081/17600 (epoch 42), train_loss = 1.313, time/batch = 0.150\n",
            "15082/17600 (epoch 42), train_loss = 1.298, time/batch = 0.187\n",
            "15083/17600 (epoch 42), train_loss = 1.312, time/batch = 0.160\n",
            "15084/17600 (epoch 42), train_loss = 1.343, time/batch = 0.167\n",
            "15085/17600 (epoch 42), train_loss = 1.300, time/batch = 0.158\n",
            "15086/17600 (epoch 42), train_loss = 1.369, time/batch = 0.183\n",
            "15087/17600 (epoch 42), train_loss = 1.362, time/batch = 0.160\n",
            "15088/17600 (epoch 42), train_loss = 1.372, time/batch = 0.198\n",
            "15089/17600 (epoch 42), train_loss = 1.286, time/batch = 0.152\n",
            "15090/17600 (epoch 42), train_loss = 1.322, time/batch = 0.166\n",
            "15091/17600 (epoch 42), train_loss = 1.327, time/batch = 0.169\n",
            "15092/17600 (epoch 42), train_loss = 1.326, time/batch = 0.173\n",
            "15093/17600 (epoch 42), train_loss = 1.329, time/batch = 0.152\n",
            "15094/17600 (epoch 42), train_loss = 1.365, time/batch = 0.184\n",
            "15095/17600 (epoch 42), train_loss = 1.315, time/batch = 0.156\n",
            "15096/17600 (epoch 42), train_loss = 1.347, time/batch = 0.165\n",
            "15097/17600 (epoch 42), train_loss = 1.283, time/batch = 0.150\n",
            "15098/17600 (epoch 42), train_loss = 1.298, time/batch = 0.178\n",
            "15099/17600 (epoch 42), train_loss = 1.267, time/batch = 0.158\n",
            "15100/17600 (epoch 42), train_loss = 1.363, time/batch = 0.160\n",
            "15101/17600 (epoch 42), train_loss = 1.350, time/batch = 0.160\n",
            "15102/17600 (epoch 42), train_loss = 1.349, time/batch = 0.170\n",
            "15103/17600 (epoch 42), train_loss = 1.331, time/batch = 0.148\n",
            "15104/17600 (epoch 42), train_loss = 1.371, time/batch = 0.169\n",
            "15105/17600 (epoch 42), train_loss = 1.337, time/batch = 0.156\n",
            "15106/17600 (epoch 42), train_loss = 1.392, time/batch = 0.156\n",
            "15107/17600 (epoch 42), train_loss = 1.362, time/batch = 0.159\n",
            "15108/17600 (epoch 42), train_loss = 1.346, time/batch = 0.156\n",
            "15109/17600 (epoch 42), train_loss = 1.297, time/batch = 0.148\n",
            "15110/17600 (epoch 42), train_loss = 1.374, time/batch = 0.159\n",
            "15111/17600 (epoch 42), train_loss = 1.352, time/batch = 0.157\n",
            "15112/17600 (epoch 42), train_loss = 1.389, time/batch = 0.170\n",
            "15113/17600 (epoch 42), train_loss = 1.357, time/batch = 0.178\n",
            "15114/17600 (epoch 42), train_loss = 1.278, time/batch = 0.165\n",
            "15115/17600 (epoch 42), train_loss = 1.306, time/batch = 0.149\n",
            "15116/17600 (epoch 42), train_loss = 1.355, time/batch = 0.167\n",
            "15117/17600 (epoch 42), train_loss = 1.366, time/batch = 0.153\n",
            "15118/17600 (epoch 42), train_loss = 1.322, time/batch = 0.246\n",
            "15119/17600 (epoch 42), train_loss = 1.314, time/batch = 0.293\n",
            "15120/17600 (epoch 42), train_loss = 1.322, time/batch = 0.265\n",
            "15121/17600 (epoch 42), train_loss = 1.361, time/batch = 0.287\n",
            "15122/17600 (epoch 42), train_loss = 1.307, time/batch = 0.267\n",
            "15123/17600 (epoch 42), train_loss = 1.337, time/batch = 0.298\n",
            "15124/17600 (epoch 42), train_loss = 1.304, time/batch = 0.291\n",
            "15125/17600 (epoch 42), train_loss = 1.347, time/batch = 0.268\n",
            "15126/17600 (epoch 42), train_loss = 1.355, time/batch = 0.252\n",
            "15127/17600 (epoch 42), train_loss = 1.318, time/batch = 0.294\n",
            "15128/17600 (epoch 42), train_loss = 1.297, time/batch = 0.271\n",
            "15129/17600 (epoch 42), train_loss = 1.316, time/batch = 0.147\n",
            "15130/17600 (epoch 42), train_loss = 1.339, time/batch = 0.171\n",
            "15131/17600 (epoch 42), train_loss = 1.312, time/batch = 0.157\n",
            "15132/17600 (epoch 42), train_loss = 1.336, time/batch = 0.180\n",
            "15133/17600 (epoch 42), train_loss = 1.320, time/batch = 0.147\n",
            "15134/17600 (epoch 42), train_loss = 1.311, time/batch = 0.156\n",
            "15135/17600 (epoch 42), train_loss = 1.303, time/batch = 0.152\n",
            "15136/17600 (epoch 43), train_loss = 1.103, time/batch = 0.157\n",
            "15137/17600 (epoch 43), train_loss = 1.334, time/batch = 0.170\n",
            "15138/17600 (epoch 43), train_loss = 1.328, time/batch = 0.164\n",
            "15139/17600 (epoch 43), train_loss = 1.273, time/batch = 0.168\n",
            "15140/17600 (epoch 43), train_loss = 1.369, time/batch = 0.163\n",
            "15141/17600 (epoch 43), train_loss = 1.345, time/batch = 0.150\n",
            "15142/17600 (epoch 43), train_loss = 1.290, time/batch = 0.165\n",
            "15143/17600 (epoch 43), train_loss = 1.277, time/batch = 0.172\n",
            "15144/17600 (epoch 43), train_loss = 1.318, time/batch = 0.169\n",
            "15145/17600 (epoch 43), train_loss = 1.371, time/batch = 0.164\n",
            "15146/17600 (epoch 43), train_loss = 1.300, time/batch = 0.159\n",
            "15147/17600 (epoch 43), train_loss = 1.313, time/batch = 0.155\n",
            "15148/17600 (epoch 43), train_loss = 1.296, time/batch = 0.159\n",
            "15149/17600 (epoch 43), train_loss = 1.281, time/batch = 0.163\n",
            "15150/17600 (epoch 43), train_loss = 1.262, time/batch = 0.168\n",
            "15151/17600 (epoch 43), train_loss = 1.326, time/batch = 0.160\n",
            "15152/17600 (epoch 43), train_loss = 1.338, time/batch = 0.162\n",
            "15153/17600 (epoch 43), train_loss = 1.292, time/batch = 0.156\n",
            "15154/17600 (epoch 43), train_loss = 1.305, time/batch = 0.164\n",
            "15155/17600 (epoch 43), train_loss = 1.349, time/batch = 0.155\n",
            "15156/17600 (epoch 43), train_loss = 1.273, time/batch = 0.178\n",
            "15157/17600 (epoch 43), train_loss = 1.327, time/batch = 0.160\n",
            "15158/17600 (epoch 43), train_loss = 1.293, time/batch = 0.163\n",
            "15159/17600 (epoch 43), train_loss = 1.265, time/batch = 0.148\n",
            "15160/17600 (epoch 43), train_loss = 1.350, time/batch = 0.159\n",
            "15161/17600 (epoch 43), train_loss = 1.282, time/batch = 0.150\n",
            "15162/17600 (epoch 43), train_loss = 1.330, time/batch = 0.175\n",
            "15163/17600 (epoch 43), train_loss = 1.313, time/batch = 0.158\n",
            "15164/17600 (epoch 43), train_loss = 1.291, time/batch = 0.167\n",
            "15165/17600 (epoch 43), train_loss = 1.225, time/batch = 0.161\n",
            "15166/17600 (epoch 43), train_loss = 1.333, time/batch = 0.159\n",
            "15167/17600 (epoch 43), train_loss = 1.311, time/batch = 0.152\n",
            "15168/17600 (epoch 43), train_loss = 1.287, time/batch = 0.173\n",
            "15169/17600 (epoch 43), train_loss = 1.291, time/batch = 0.163\n",
            "15170/17600 (epoch 43), train_loss = 1.295, time/batch = 0.173\n",
            "15171/17600 (epoch 43), train_loss = 1.309, time/batch = 0.155\n",
            "15172/17600 (epoch 43), train_loss = 1.337, time/batch = 0.161\n",
            "15173/17600 (epoch 43), train_loss = 1.332, time/batch = 0.160\n",
            "15174/17600 (epoch 43), train_loss = 1.312, time/batch = 0.163\n",
            "15175/17600 (epoch 43), train_loss = 1.290, time/batch = 0.179\n",
            "15176/17600 (epoch 43), train_loss = 1.336, time/batch = 0.160\n",
            "15177/17600 (epoch 43), train_loss = 1.272, time/batch = 0.160\n",
            "15178/17600 (epoch 43), train_loss = 1.328, time/batch = 0.161\n",
            "15179/17600 (epoch 43), train_loss = 1.249, time/batch = 0.156\n",
            "15180/17600 (epoch 43), train_loss = 1.316, time/batch = 0.158\n",
            "15181/17600 (epoch 43), train_loss = 1.288, time/batch = 0.169\n",
            "15182/17600 (epoch 43), train_loss = 1.243, time/batch = 0.168\n",
            "15183/17600 (epoch 43), train_loss = 1.303, time/batch = 0.152\n",
            "15184/17600 (epoch 43), train_loss = 1.254, time/batch = 0.159\n",
            "15185/17600 (epoch 43), train_loss = 1.271, time/batch = 0.158\n",
            "15186/17600 (epoch 43), train_loss = 1.295, time/batch = 0.161\n",
            "15187/17600 (epoch 43), train_loss = 1.285, time/batch = 0.169\n",
            "15188/17600 (epoch 43), train_loss = 1.256, time/batch = 0.205\n",
            "15189/17600 (epoch 43), train_loss = 1.284, time/batch = 0.247\n",
            "15190/17600 (epoch 43), train_loss = 1.328, time/batch = 0.276\n",
            "15191/17600 (epoch 43), train_loss = 1.306, time/batch = 0.309\n",
            "15192/17600 (epoch 43), train_loss = 1.306, time/batch = 0.282\n",
            "15193/17600 (epoch 43), train_loss = 1.301, time/batch = 0.283\n",
            "15194/17600 (epoch 43), train_loss = 1.280, time/batch = 0.279\n",
            "15195/17600 (epoch 43), train_loss = 1.307, time/batch = 0.304\n",
            "15196/17600 (epoch 43), train_loss = 1.303, time/batch = 0.298\n",
            "15197/17600 (epoch 43), train_loss = 1.245, time/batch = 0.283\n",
            "15198/17600 (epoch 43), train_loss = 1.247, time/batch = 0.268\n",
            "15199/17600 (epoch 43), train_loss = 1.266, time/batch = 0.176\n",
            "15200/17600 (epoch 43), train_loss = 1.297, time/batch = 0.149\n",
            "15201/17600 (epoch 43), train_loss = 1.263, time/batch = 0.171\n",
            "15202/17600 (epoch 43), train_loss = 1.300, time/batch = 0.156\n",
            "15203/17600 (epoch 43), train_loss = 1.259, time/batch = 0.169\n",
            "15204/17600 (epoch 43), train_loss = 1.308, time/batch = 0.162\n",
            "15205/17600 (epoch 43), train_loss = 1.239, time/batch = 0.168\n",
            "15206/17600 (epoch 43), train_loss = 1.284, time/batch = 0.151\n",
            "15207/17600 (epoch 43), train_loss = 1.329, time/batch = 0.159\n",
            "15208/17600 (epoch 43), train_loss = 1.291, time/batch = 0.148\n",
            "15209/17600 (epoch 43), train_loss = 1.361, time/batch = 0.166\n",
            "15210/17600 (epoch 43), train_loss = 1.292, time/batch = 0.151\n",
            "15211/17600 (epoch 43), train_loss = 1.293, time/batch = 0.168\n",
            "15212/17600 (epoch 43), train_loss = 1.249, time/batch = 0.170\n",
            "15213/17600 (epoch 43), train_loss = 1.305, time/batch = 0.161\n",
            "15214/17600 (epoch 43), train_loss = 1.311, time/batch = 0.158\n",
            "15215/17600 (epoch 43), train_loss = 1.387, time/batch = 0.158\n",
            "15216/17600 (epoch 43), train_loss = 1.356, time/batch = 0.154\n",
            "15217/17600 (epoch 43), train_loss = 1.349, time/batch = 0.170\n",
            "15218/17600 (epoch 43), train_loss = 1.330, time/batch = 0.173\n",
            "15219/17600 (epoch 43), train_loss = 1.302, time/batch = 0.173\n",
            "15220/17600 (epoch 43), train_loss = 1.279, time/batch = 0.155\n",
            "15221/17600 (epoch 43), train_loss = 1.313, time/batch = 0.162\n",
            "15222/17600 (epoch 43), train_loss = 1.305, time/batch = 0.152\n",
            "15223/17600 (epoch 43), train_loss = 1.244, time/batch = 0.158\n",
            "15224/17600 (epoch 43), train_loss = 1.284, time/batch = 0.156\n",
            "15225/17600 (epoch 43), train_loss = 1.224, time/batch = 0.172\n",
            "15226/17600 (epoch 43), train_loss = 1.288, time/batch = 0.156\n",
            "15227/17600 (epoch 43), train_loss = 1.274, time/batch = 0.162\n",
            "15228/17600 (epoch 43), train_loss = 1.252, time/batch = 0.173\n",
            "15229/17600 (epoch 43), train_loss = 1.307, time/batch = 0.167\n",
            "15230/17600 (epoch 43), train_loss = 1.290, time/batch = 0.167\n",
            "15231/17600 (epoch 43), train_loss = 1.253, time/batch = 0.156\n",
            "15232/17600 (epoch 43), train_loss = 1.228, time/batch = 0.157\n",
            "15233/17600 (epoch 43), train_loss = 1.279, time/batch = 0.156\n",
            "15234/17600 (epoch 43), train_loss = 1.314, time/batch = 0.157\n",
            "15235/17600 (epoch 43), train_loss = 1.272, time/batch = 0.155\n",
            "15236/17600 (epoch 43), train_loss = 1.247, time/batch = 0.167\n",
            "15237/17600 (epoch 43), train_loss = 1.331, time/batch = 0.164\n",
            "15238/17600 (epoch 43), train_loss = 1.269, time/batch = 0.172\n",
            "15239/17600 (epoch 43), train_loss = 1.379, time/batch = 0.149\n",
            "15240/17600 (epoch 43), train_loss = 1.307, time/batch = 0.154\n",
            "15241/17600 (epoch 43), train_loss = 1.281, time/batch = 0.160\n",
            "15242/17600 (epoch 43), train_loss = 1.322, time/batch = 0.170\n",
            "15243/17600 (epoch 43), train_loss = 1.309, time/batch = 0.175\n",
            "15244/17600 (epoch 43), train_loss = 1.311, time/batch = 0.177\n",
            "15245/17600 (epoch 43), train_loss = 1.295, time/batch = 0.163\n",
            "15246/17600 (epoch 43), train_loss = 1.251, time/batch = 0.149\n",
            "15247/17600 (epoch 43), train_loss = 1.273, time/batch = 0.161\n",
            "15248/17600 (epoch 43), train_loss = 1.305, time/batch = 0.161\n",
            "15249/17600 (epoch 43), train_loss = 1.201, time/batch = 0.165\n",
            "15250/17600 (epoch 43), train_loss = 1.275, time/batch = 0.171\n",
            "15251/17600 (epoch 43), train_loss = 1.305, time/batch = 0.173\n",
            "15252/17600 (epoch 43), train_loss = 1.271, time/batch = 0.152\n",
            "15253/17600 (epoch 43), train_loss = 1.299, time/batch = 0.160\n",
            "15254/17600 (epoch 43), train_loss = 1.241, time/batch = 0.158\n",
            "15255/17600 (epoch 43), train_loss = 1.345, time/batch = 0.164\n",
            "15256/17600 (epoch 43), train_loss = 1.251, time/batch = 0.178\n",
            "15257/17600 (epoch 43), train_loss = 1.313, time/batch = 0.170\n",
            "15258/17600 (epoch 43), train_loss = 1.337, time/batch = 0.145\n",
            "15259/17600 (epoch 43), train_loss = 1.323, time/batch = 0.165\n",
            "15260/17600 (epoch 43), train_loss = 1.239, time/batch = 0.260\n",
            "15261/17600 (epoch 43), train_loss = 1.305, time/batch = 0.269\n",
            "15262/17600 (epoch 43), train_loss = 1.302, time/batch = 0.277\n",
            "15263/17600 (epoch 43), train_loss = 1.289, time/batch = 0.297\n",
            "15264/17600 (epoch 43), train_loss = 1.271, time/batch = 0.270\n",
            "15265/17600 (epoch 43), train_loss = 1.304, time/batch = 0.268\n",
            "15266/17600 (epoch 43), train_loss = 1.317, time/batch = 0.278\n",
            "15267/17600 (epoch 43), train_loss = 1.322, time/batch = 0.273\n",
            "15268/17600 (epoch 43), train_loss = 1.291, time/batch = 0.277\n",
            "15269/17600 (epoch 43), train_loss = 1.263, time/batch = 0.291\n",
            "15270/17600 (epoch 43), train_loss = 1.274, time/batch = 0.250\n",
            "15271/17600 (epoch 43), train_loss = 1.245, time/batch = 0.158\n",
            "15272/17600 (epoch 43), train_loss = 1.313, time/batch = 0.180\n",
            "15273/17600 (epoch 43), train_loss = 1.339, time/batch = 0.164\n",
            "15274/17600 (epoch 43), train_loss = 1.285, time/batch = 0.189\n",
            "15275/17600 (epoch 43), train_loss = 1.337, time/batch = 0.165\n",
            "15276/17600 (epoch 43), train_loss = 1.362, time/batch = 0.168\n",
            "15277/17600 (epoch 43), train_loss = 1.275, time/batch = 0.156\n",
            "15278/17600 (epoch 43), train_loss = 1.300, time/batch = 0.166\n",
            "15279/17600 (epoch 43), train_loss = 1.266, time/batch = 0.159\n",
            "15280/17600 (epoch 43), train_loss = 1.318, time/batch = 0.181\n",
            "15281/17600 (epoch 43), train_loss = 1.303, time/batch = 0.167\n",
            "15282/17600 (epoch 43), train_loss = 1.278, time/batch = 0.164\n",
            "15283/17600 (epoch 43), train_loss = 1.286, time/batch = 0.178\n",
            "15284/17600 (epoch 43), train_loss = 1.268, time/batch = 0.154\n",
            "15285/17600 (epoch 43), train_loss = 1.331, time/batch = 0.173\n",
            "15286/17600 (epoch 43), train_loss = 1.218, time/batch = 0.173\n",
            "15287/17600 (epoch 43), train_loss = 1.292, time/batch = 0.153\n",
            "15288/17600 (epoch 43), train_loss = 1.317, time/batch = 0.160\n",
            "15289/17600 (epoch 43), train_loss = 1.286, time/batch = 0.175\n",
            "15290/17600 (epoch 43), train_loss = 1.302, time/batch = 0.159\n",
            "15291/17600 (epoch 43), train_loss = 1.241, time/batch = 0.159\n",
            "15292/17600 (epoch 43), train_loss = 1.237, time/batch = 0.159\n",
            "15293/17600 (epoch 43), train_loss = 1.256, time/batch = 0.191\n",
            "15294/17600 (epoch 43), train_loss = 1.316, time/batch = 0.156\n",
            "15295/17600 (epoch 43), train_loss = 1.359, time/batch = 0.163\n",
            "15296/17600 (epoch 43), train_loss = 1.315, time/batch = 0.152\n",
            "15297/17600 (epoch 43), train_loss = 1.314, time/batch = 0.172\n",
            "15298/17600 (epoch 43), train_loss = 1.245, time/batch = 0.166\n",
            "15299/17600 (epoch 43), train_loss = 1.273, time/batch = 0.183\n",
            "15300/17600 (epoch 43), train_loss = 1.296, time/batch = 0.143\n",
            "15301/17600 (epoch 43), train_loss = 1.282, time/batch = 0.173\n",
            "15302/17600 (epoch 43), train_loss = 1.295, time/batch = 0.155\n",
            "15303/17600 (epoch 43), train_loss = 1.321, time/batch = 0.166\n",
            "15304/17600 (epoch 43), train_loss = 1.314, time/batch = 0.152\n",
            "15305/17600 (epoch 43), train_loss = 1.294, time/batch = 0.182\n",
            "15306/17600 (epoch 43), train_loss = 1.227, time/batch = 0.164\n",
            "15307/17600 (epoch 43), train_loss = 1.239, time/batch = 0.181\n",
            "15308/17600 (epoch 43), train_loss = 1.328, time/batch = 0.157\n",
            "15309/17600 (epoch 43), train_loss = 1.296, time/batch = 0.166\n",
            "15310/17600 (epoch 43), train_loss = 1.287, time/batch = 0.157\n",
            "15311/17600 (epoch 43), train_loss = 1.290, time/batch = 0.164\n",
            "15312/17600 (epoch 43), train_loss = 1.204, time/batch = 0.155\n",
            "15313/17600 (epoch 43), train_loss = 1.267, time/batch = 0.164\n",
            "15314/17600 (epoch 43), train_loss = 1.267, time/batch = 0.155\n",
            "15315/17600 (epoch 43), train_loss = 1.247, time/batch = 0.163\n",
            "15316/17600 (epoch 43), train_loss = 1.245, time/batch = 0.150\n",
            "15317/17600 (epoch 43), train_loss = 1.299, time/batch = 0.169\n",
            "15318/17600 (epoch 43), train_loss = 1.355, time/batch = 0.171\n",
            "15319/17600 (epoch 43), train_loss = 1.271, time/batch = 0.160\n",
            "15320/17600 (epoch 43), train_loss = 1.272, time/batch = 0.156\n",
            "15321/17600 (epoch 43), train_loss = 1.235, time/batch = 0.161\n",
            "15322/17600 (epoch 43), train_loss = 1.310, time/batch = 0.152\n",
            "15323/17600 (epoch 43), train_loss = 1.320, time/batch = 0.164\n",
            "15324/17600 (epoch 43), train_loss = 1.240, time/batch = 0.170\n",
            "15325/17600 (epoch 43), train_loss = 1.334, time/batch = 0.167\n",
            "15326/17600 (epoch 43), train_loss = 1.301, time/batch = 0.153\n",
            "15327/17600 (epoch 43), train_loss = 1.281, time/batch = 0.166\n",
            "15328/17600 (epoch 43), train_loss = 1.289, time/batch = 0.155\n",
            "15329/17600 (epoch 43), train_loss = 1.308, time/batch = 0.147\n",
            "15330/17600 (epoch 43), train_loss = 1.309, time/batch = 0.170\n",
            "15331/17600 (epoch 43), train_loss = 1.318, time/batch = 0.249\n",
            "15332/17600 (epoch 43), train_loss = 1.273, time/batch = 0.282\n",
            "15333/17600 (epoch 43), train_loss = 1.326, time/batch = 0.267\n",
            "15334/17600 (epoch 43), train_loss = 1.273, time/batch = 0.271\n",
            "15335/17600 (epoch 43), train_loss = 1.307, time/batch = 0.264\n",
            "15336/17600 (epoch 43), train_loss = 1.310, time/batch = 0.268\n",
            "15337/17600 (epoch 43), train_loss = 1.276, time/batch = 0.259\n",
            "15338/17600 (epoch 43), train_loss = 1.266, time/batch = 0.280\n",
            "15339/17600 (epoch 43), train_loss = 1.295, time/batch = 0.248\n",
            "15340/17600 (epoch 43), train_loss = 1.329, time/batch = 0.306\n",
            "15341/17600 (epoch 43), train_loss = 1.376, time/batch = 0.238\n",
            "15342/17600 (epoch 43), train_loss = 1.291, time/batch = 0.168\n",
            "15343/17600 (epoch 43), train_loss = 1.292, time/batch = 0.181\n",
            "15344/17600 (epoch 43), train_loss = 1.295, time/batch = 0.158\n",
            "15345/17600 (epoch 43), train_loss = 1.273, time/batch = 0.159\n",
            "15346/17600 (epoch 43), train_loss = 1.317, time/batch = 0.155\n",
            "15347/17600 (epoch 43), train_loss = 1.309, time/batch = 0.164\n",
            "15348/17600 (epoch 43), train_loss = 1.278, time/batch = 0.165\n",
            "15349/17600 (epoch 43), train_loss = 1.352, time/batch = 0.167\n",
            "15350/17600 (epoch 43), train_loss = 1.391, time/batch = 0.162\n",
            "15351/17600 (epoch 43), train_loss = 1.280, time/batch = 0.154\n",
            "15352/17600 (epoch 43), train_loss = 1.301, time/batch = 0.168\n",
            "15353/17600 (epoch 43), train_loss = 1.258, time/batch = 0.148\n",
            "15354/17600 (epoch 43), train_loss = 1.291, time/batch = 0.167\n",
            "15355/17600 (epoch 43), train_loss = 1.273, time/batch = 0.168\n",
            "15356/17600 (epoch 43), train_loss = 1.267, time/batch = 0.179\n",
            "15357/17600 (epoch 43), train_loss = 1.291, time/batch = 0.160\n",
            "15358/17600 (epoch 43), train_loss = 1.361, time/batch = 0.156\n",
            "15359/17600 (epoch 43), train_loss = 1.332, time/batch = 0.160\n",
            "15360/17600 (epoch 43), train_loss = 1.249, time/batch = 0.163\n",
            "15361/17600 (epoch 43), train_loss = 1.297, time/batch = 0.163\n",
            "15362/17600 (epoch 43), train_loss = 1.302, time/batch = 0.161\n",
            "15363/17600 (epoch 43), train_loss = 1.281, time/batch = 0.161\n",
            "15364/17600 (epoch 43), train_loss = 1.312, time/batch = 0.161\n",
            "15365/17600 (epoch 43), train_loss = 1.277, time/batch = 0.153\n",
            "15366/17600 (epoch 43), train_loss = 1.266, time/batch = 0.177\n",
            "15367/17600 (epoch 43), train_loss = 1.330, time/batch = 0.162\n",
            "15368/17600 (epoch 43), train_loss = 1.329, time/batch = 0.183\n",
            "15369/17600 (epoch 43), train_loss = 1.343, time/batch = 0.162\n",
            "15370/17600 (epoch 43), train_loss = 1.258, time/batch = 0.175\n",
            "15371/17600 (epoch 43), train_loss = 1.289, time/batch = 0.157\n",
            "15372/17600 (epoch 43), train_loss = 1.279, time/batch = 0.157\n",
            "15373/17600 (epoch 43), train_loss = 1.348, time/batch = 0.159\n",
            "15374/17600 (epoch 43), train_loss = 1.345, time/batch = 0.170\n",
            "15375/17600 (epoch 43), train_loss = 1.239, time/batch = 0.156\n",
            "15376/17600 (epoch 43), train_loss = 1.331, time/batch = 0.157\n",
            "15377/17600 (epoch 43), train_loss = 1.279, time/batch = 0.153\n",
            "15378/17600 (epoch 43), train_loss = 1.315, time/batch = 0.158\n",
            "15379/17600 (epoch 43), train_loss = 1.307, time/batch = 0.153\n",
            "15380/17600 (epoch 43), train_loss = 1.275, time/batch = 0.178\n",
            "15381/17600 (epoch 43), train_loss = 1.315, time/batch = 0.151\n",
            "15382/17600 (epoch 43), train_loss = 1.307, time/batch = 0.165\n",
            "15383/17600 (epoch 43), train_loss = 1.280, time/batch = 0.154\n",
            "15384/17600 (epoch 43), train_loss = 1.258, time/batch = 0.172\n",
            "15385/17600 (epoch 43), train_loss = 1.237, time/batch = 0.150\n",
            "15386/17600 (epoch 43), train_loss = 1.371, time/batch = 0.166\n",
            "15387/17600 (epoch 43), train_loss = 1.309, time/batch = 0.171\n",
            "15388/17600 (epoch 43), train_loss = 1.301, time/batch = 0.156\n",
            "15389/17600 (epoch 43), train_loss = 1.309, time/batch = 0.153\n",
            "15390/17600 (epoch 43), train_loss = 1.327, time/batch = 0.166\n",
            "15391/17600 (epoch 43), train_loss = 1.335, time/batch = 0.143\n",
            "15392/17600 (epoch 43), train_loss = 1.382, time/batch = 0.161\n",
            "15393/17600 (epoch 43), train_loss = 1.351, time/batch = 0.172\n",
            "15394/17600 (epoch 43), train_loss = 1.303, time/batch = 0.160\n",
            "15395/17600 (epoch 43), train_loss = 1.305, time/batch = 0.154\n",
            "15396/17600 (epoch 43), train_loss = 1.281, time/batch = 0.168\n",
            "15397/17600 (epoch 43), train_loss = 1.289, time/batch = 0.156\n",
            "15398/17600 (epoch 43), train_loss = 1.269, time/batch = 0.166\n",
            "15399/17600 (epoch 43), train_loss = 1.227, time/batch = 0.147\n",
            "15400/17600 (epoch 43), train_loss = 1.278, time/batch = 0.185\n",
            "15401/17600 (epoch 43), train_loss = 1.285, time/batch = 0.148\n",
            "15402/17600 (epoch 43), train_loss = 1.239, time/batch = 0.164\n",
            "15403/17600 (epoch 43), train_loss = 1.242, time/batch = 0.258\n",
            "15404/17600 (epoch 43), train_loss = 1.274, time/batch = 0.267\n",
            "15405/17600 (epoch 43), train_loss = 1.308, time/batch = 0.281\n",
            "15406/17600 (epoch 43), train_loss = 1.231, time/batch = 0.272\n",
            "15407/17600 (epoch 43), train_loss = 1.301, time/batch = 0.272\n",
            "15408/17600 (epoch 43), train_loss = 1.268, time/batch = 0.305\n",
            "15409/17600 (epoch 43), train_loss = 1.294, time/batch = 0.300\n",
            "15410/17600 (epoch 43), train_loss = 1.277, time/batch = 0.281\n",
            "15411/17600 (epoch 43), train_loss = 1.332, time/batch = 0.291\n",
            "15412/17600 (epoch 43), train_loss = 1.296, time/batch = 0.294\n",
            "15413/17600 (epoch 43), train_loss = 1.322, time/batch = 0.185\n",
            "15414/17600 (epoch 43), train_loss = 1.288, time/batch = 0.154\n",
            "15415/17600 (epoch 43), train_loss = 1.294, time/batch = 0.159\n",
            "15416/17600 (epoch 43), train_loss = 1.317, time/batch = 0.157\n",
            "15417/17600 (epoch 43), train_loss = 1.364, time/batch = 0.162\n",
            "15418/17600 (epoch 43), train_loss = 1.336, time/batch = 0.175\n",
            "15419/17600 (epoch 43), train_loss = 1.334, time/batch = 0.159\n",
            "15420/17600 (epoch 43), train_loss = 1.351, time/batch = 0.161\n",
            "15421/17600 (epoch 43), train_loss = 1.364, time/batch = 0.169\n",
            "15422/17600 (epoch 43), train_loss = 1.310, time/batch = 0.162\n",
            "15423/17600 (epoch 43), train_loss = 1.357, time/batch = 0.167\n",
            "15424/17600 (epoch 43), train_loss = 1.294, time/batch = 0.182\n",
            "15425/17600 (epoch 43), train_loss = 1.289, time/batch = 0.153\n",
            "15426/17600 (epoch 43), train_loss = 1.355, time/batch = 0.157\n",
            "15427/17600 (epoch 43), train_loss = 1.259, time/batch = 0.151\n",
            "15428/17600 (epoch 43), train_loss = 1.255, time/batch = 0.160\n",
            "15429/17600 (epoch 43), train_loss = 1.329, time/batch = 0.165\n",
            "15430/17600 (epoch 43), train_loss = 1.278, time/batch = 0.161\n",
            "15431/17600 (epoch 43), train_loss = 1.337, time/batch = 0.171\n",
            "15432/17600 (epoch 43), train_loss = 1.283, time/batch = 0.167\n",
            "15433/17600 (epoch 43), train_loss = 1.311, time/batch = 0.150\n",
            "15434/17600 (epoch 43), train_loss = 1.297, time/batch = 0.173\n",
            "15435/17600 (epoch 43), train_loss = 1.309, time/batch = 0.153\n",
            "15436/17600 (epoch 43), train_loss = 1.340, time/batch = 0.160\n",
            "15437/17600 (epoch 43), train_loss = 1.299, time/batch = 0.174\n",
            "15438/17600 (epoch 43), train_loss = 1.367, time/batch = 0.165\n",
            "15439/17600 (epoch 43), train_loss = 1.359, time/batch = 0.153\n",
            "15440/17600 (epoch 43), train_loss = 1.370, time/batch = 0.165\n",
            "15441/17600 (epoch 43), train_loss = 1.284, time/batch = 0.175\n",
            "15442/17600 (epoch 43), train_loss = 1.319, time/batch = 0.174\n",
            "15443/17600 (epoch 43), train_loss = 1.326, time/batch = 0.171\n",
            "15444/17600 (epoch 43), train_loss = 1.324, time/batch = 0.166\n",
            "15445/17600 (epoch 43), train_loss = 1.326, time/batch = 0.148\n",
            "15446/17600 (epoch 43), train_loss = 1.363, time/batch = 0.171\n",
            "15447/17600 (epoch 43), train_loss = 1.313, time/batch = 0.149\n",
            "15448/17600 (epoch 43), train_loss = 1.346, time/batch = 0.159\n",
            "15449/17600 (epoch 43), train_loss = 1.281, time/batch = 0.144\n",
            "15450/17600 (epoch 43), train_loss = 1.296, time/batch = 0.182\n",
            "15451/17600 (epoch 43), train_loss = 1.265, time/batch = 0.155\n",
            "15452/17600 (epoch 43), train_loss = 1.361, time/batch = 0.154\n",
            "15453/17600 (epoch 43), train_loss = 1.349, time/batch = 0.153\n",
            "15454/17600 (epoch 43), train_loss = 1.347, time/batch = 0.168\n",
            "15455/17600 (epoch 43), train_loss = 1.330, time/batch = 0.158\n",
            "15456/17600 (epoch 43), train_loss = 1.369, time/batch = 0.180\n",
            "15457/17600 (epoch 43), train_loss = 1.334, time/batch = 0.154\n",
            "15458/17600 (epoch 43), train_loss = 1.390, time/batch = 0.174\n",
            "15459/17600 (epoch 43), train_loss = 1.360, time/batch = 0.177\n",
            "15460/17600 (epoch 43), train_loss = 1.345, time/batch = 0.172\n",
            "15461/17600 (epoch 43), train_loss = 1.296, time/batch = 0.153\n",
            "15462/17600 (epoch 43), train_loss = 1.372, time/batch = 0.174\n",
            "15463/17600 (epoch 43), train_loss = 1.350, time/batch = 0.154\n",
            "15464/17600 (epoch 43), train_loss = 1.387, time/batch = 0.160\n",
            "15465/17600 (epoch 43), train_loss = 1.356, time/batch = 0.149\n",
            "15466/17600 (epoch 43), train_loss = 1.276, time/batch = 0.168\n",
            "15467/17600 (epoch 43), train_loss = 1.305, time/batch = 0.160\n",
            "15468/17600 (epoch 43), train_loss = 1.352, time/batch = 0.162\n",
            "15469/17600 (epoch 43), train_loss = 1.364, time/batch = 0.177\n",
            "15470/17600 (epoch 43), train_loss = 1.320, time/batch = 0.164\n",
            "15471/17600 (epoch 43), train_loss = 1.312, time/batch = 0.163\n",
            "15472/17600 (epoch 43), train_loss = 1.319, time/batch = 0.184\n",
            "15473/17600 (epoch 43), train_loss = 1.359, time/batch = 0.159\n",
            "15474/17600 (epoch 43), train_loss = 1.305, time/batch = 0.280\n",
            "15475/17600 (epoch 43), train_loss = 1.334, time/batch = 0.298\n",
            "15476/17600 (epoch 43), train_loss = 1.302, time/batch = 0.286\n",
            "15477/17600 (epoch 43), train_loss = 1.345, time/batch = 0.265\n",
            "15478/17600 (epoch 43), train_loss = 1.353, time/batch = 0.285\n",
            "15479/17600 (epoch 43), train_loss = 1.316, time/batch = 0.288\n",
            "15480/17600 (epoch 43), train_loss = 1.295, time/batch = 0.273\n",
            "15481/17600 (epoch 43), train_loss = 1.314, time/batch = 0.272\n",
            "15482/17600 (epoch 43), train_loss = 1.337, time/batch = 0.280\n",
            "15483/17600 (epoch 43), train_loss = 1.310, time/batch = 0.286\n",
            "15484/17600 (epoch 43), train_loss = 1.334, time/batch = 0.185\n",
            "15485/17600 (epoch 43), train_loss = 1.319, time/batch = 0.164\n",
            "15486/17600 (epoch 43), train_loss = 1.310, time/batch = 0.158\n",
            "15487/17600 (epoch 43), train_loss = 1.302, time/batch = 0.177\n",
            "15488/17600 (epoch 44), train_loss = 1.100, time/batch = 0.169\n",
            "15489/17600 (epoch 44), train_loss = 1.333, time/batch = 0.146\n",
            "15490/17600 (epoch 44), train_loss = 1.326, time/batch = 0.158\n",
            "15491/17600 (epoch 44), train_loss = 1.271, time/batch = 0.146\n",
            "15492/17600 (epoch 44), train_loss = 1.368, time/batch = 0.183\n",
            "15493/17600 (epoch 44), train_loss = 1.343, time/batch = 0.156\n",
            "15494/17600 (epoch 44), train_loss = 1.288, time/batch = 0.164\n",
            "15495/17600 (epoch 44), train_loss = 1.275, time/batch = 0.162\n",
            "15496/17600 (epoch 44), train_loss = 1.316, time/batch = 0.150\n",
            "15497/17600 (epoch 44), train_loss = 1.368, time/batch = 0.163\n",
            "15498/17600 (epoch 44), train_loss = 1.297, time/batch = 0.169\n",
            "15499/17600 (epoch 44), train_loss = 1.311, time/batch = 0.166\n",
            "15500/17600 (epoch 44), train_loss = 1.294, time/batch = 0.159\n",
            "15501/17600 (epoch 44), train_loss = 1.279, time/batch = 0.149\n",
            "15502/17600 (epoch 44), train_loss = 1.259, time/batch = 0.152\n",
            "15503/17600 (epoch 44), train_loss = 1.325, time/batch = 0.167\n",
            "15504/17600 (epoch 44), train_loss = 1.337, time/batch = 0.167\n",
            "15505/17600 (epoch 44), train_loss = 1.291, time/batch = 0.167\n",
            "15506/17600 (epoch 44), train_loss = 1.304, time/batch = 0.149\n",
            "15507/17600 (epoch 44), train_loss = 1.347, time/batch = 0.161\n",
            "15508/17600 (epoch 44), train_loss = 1.271, time/batch = 0.160\n",
            "15509/17600 (epoch 44), train_loss = 1.325, time/batch = 0.175\n",
            "15510/17600 (epoch 44), train_loss = 1.291, time/batch = 0.152\n",
            "15511/17600 (epoch 44), train_loss = 1.264, time/batch = 0.173\n",
            "15512/17600 (epoch 44), train_loss = 1.349, time/batch = 0.154\n",
            "15513/17600 (epoch 44), train_loss = 1.281, time/batch = 0.164\n",
            "15514/17600 (epoch 44), train_loss = 1.327, time/batch = 0.162\n",
            "15515/17600 (epoch 44), train_loss = 1.311, time/batch = 0.161\n",
            "15516/17600 (epoch 44), train_loss = 1.289, time/batch = 0.153\n",
            "15517/17600 (epoch 44), train_loss = 1.224, time/batch = 0.179\n",
            "15518/17600 (epoch 44), train_loss = 1.331, time/batch = 0.156\n",
            "15519/17600 (epoch 44), train_loss = 1.310, time/batch = 0.162\n",
            "15520/17600 (epoch 44), train_loss = 1.285, time/batch = 0.164\n",
            "15521/17600 (epoch 44), train_loss = 1.288, time/batch = 0.161\n",
            "15522/17600 (epoch 44), train_loss = 1.293, time/batch = 0.152\n",
            "15523/17600 (epoch 44), train_loss = 1.307, time/batch = 0.188\n",
            "15524/17600 (epoch 44), train_loss = 1.335, time/batch = 0.166\n",
            "15525/17600 (epoch 44), train_loss = 1.331, time/batch = 0.160\n",
            "15526/17600 (epoch 44), train_loss = 1.310, time/batch = 0.154\n",
            "15527/17600 (epoch 44), train_loss = 1.288, time/batch = 0.155\n",
            "15528/17600 (epoch 44), train_loss = 1.334, time/batch = 0.145\n",
            "15529/17600 (epoch 44), train_loss = 1.270, time/batch = 0.161\n",
            "15530/17600 (epoch 44), train_loss = 1.326, time/batch = 0.167\n",
            "15531/17600 (epoch 44), train_loss = 1.247, time/batch = 0.174\n",
            "15532/17600 (epoch 44), train_loss = 1.314, time/batch = 0.155\n",
            "15533/17600 (epoch 44), train_loss = 1.287, time/batch = 0.162\n",
            "15534/17600 (epoch 44), train_loss = 1.241, time/batch = 0.168\n",
            "15535/17600 (epoch 44), train_loss = 1.301, time/batch = 0.158\n",
            "15536/17600 (epoch 44), train_loss = 1.252, time/batch = 0.179\n",
            "15537/17600 (epoch 44), train_loss = 1.268, time/batch = 0.158\n",
            "15538/17600 (epoch 44), train_loss = 1.293, time/batch = 0.164\n",
            "15539/17600 (epoch 44), train_loss = 1.283, time/batch = 0.159\n",
            "15540/17600 (epoch 44), train_loss = 1.255, time/batch = 0.160\n",
            "15541/17600 (epoch 44), train_loss = 1.282, time/batch = 0.153\n",
            "15542/17600 (epoch 44), train_loss = 1.327, time/batch = 0.188\n",
            "15543/17600 (epoch 44), train_loss = 1.304, time/batch = 0.152\n",
            "15544/17600 (epoch 44), train_loss = 1.304, time/batch = 0.233\n",
            "15545/17600 (epoch 44), train_loss = 1.298, time/batch = 0.282\n",
            "15546/17600 (epoch 44), train_loss = 1.278, time/batch = 0.279\n",
            "15547/17600 (epoch 44), train_loss = 1.305, time/batch = 0.293\n",
            "15548/17600 (epoch 44), train_loss = 1.302, time/batch = 0.262\n",
            "15549/17600 (epoch 44), train_loss = 1.242, time/batch = 0.280\n",
            "15550/17600 (epoch 44), train_loss = 1.245, time/batch = 0.259\n",
            "15551/17600 (epoch 44), train_loss = 1.264, time/batch = 0.237\n",
            "15552/17600 (epoch 44), train_loss = 1.295, time/batch = 0.266\n",
            "15553/17600 (epoch 44), train_loss = 1.261, time/batch = 0.289\n",
            "15554/17600 (epoch 44), train_loss = 1.298, time/batch = 0.281\n",
            "15555/17600 (epoch 44), train_loss = 1.258, time/batch = 0.162\n",
            "15556/17600 (epoch 44), train_loss = 1.306, time/batch = 0.160\n",
            "15557/17600 (epoch 44), train_loss = 1.237, time/batch = 0.156\n",
            "15558/17600 (epoch 44), train_loss = 1.282, time/batch = 0.156\n",
            "15559/17600 (epoch 44), train_loss = 1.328, time/batch = 0.159\n",
            "15560/17600 (epoch 44), train_loss = 1.289, time/batch = 0.160\n",
            "15561/17600 (epoch 44), train_loss = 1.359, time/batch = 0.180\n",
            "15562/17600 (epoch 44), train_loss = 1.290, time/batch = 0.168\n",
            "15563/17600 (epoch 44), train_loss = 1.291, time/batch = 0.150\n",
            "15564/17600 (epoch 44), train_loss = 1.248, time/batch = 0.170\n",
            "15565/17600 (epoch 44), train_loss = 1.303, time/batch = 0.158\n",
            "15566/17600 (epoch 44), train_loss = 1.309, time/batch = 0.168\n",
            "15567/17600 (epoch 44), train_loss = 1.385, time/batch = 0.176\n",
            "15568/17600 (epoch 44), train_loss = 1.354, time/batch = 0.164\n",
            "15569/17600 (epoch 44), train_loss = 1.347, time/batch = 0.157\n",
            "15570/17600 (epoch 44), train_loss = 1.328, time/batch = 0.171\n",
            "15571/17600 (epoch 44), train_loss = 1.301, time/batch = 0.168\n",
            "15572/17600 (epoch 44), train_loss = 1.277, time/batch = 0.170\n",
            "15573/17600 (epoch 44), train_loss = 1.311, time/batch = 0.170\n",
            "15574/17600 (epoch 44), train_loss = 1.303, time/batch = 0.165\n",
            "15575/17600 (epoch 44), train_loss = 1.242, time/batch = 0.155\n",
            "15576/17600 (epoch 44), train_loss = 1.282, time/batch = 0.171\n",
            "15577/17600 (epoch 44), train_loss = 1.222, time/batch = 0.175\n",
            "15578/17600 (epoch 44), train_loss = 1.286, time/batch = 0.152\n",
            "15579/17600 (epoch 44), train_loss = 1.273, time/batch = 0.175\n",
            "15580/17600 (epoch 44), train_loss = 1.250, time/batch = 0.167\n",
            "15581/17600 (epoch 44), train_loss = 1.305, time/batch = 0.171\n",
            "15582/17600 (epoch 44), train_loss = 1.288, time/batch = 0.146\n",
            "15583/17600 (epoch 44), train_loss = 1.251, time/batch = 0.167\n",
            "15584/17600 (epoch 44), train_loss = 1.225, time/batch = 0.163\n",
            "15585/17600 (epoch 44), train_loss = 1.277, time/batch = 0.168\n",
            "15586/17600 (epoch 44), train_loss = 1.312, time/batch = 0.170\n",
            "15587/17600 (epoch 44), train_loss = 1.271, time/batch = 0.163\n",
            "15588/17600 (epoch 44), train_loss = 1.245, time/batch = 0.158\n",
            "15589/17600 (epoch 44), train_loss = 1.329, time/batch = 0.166\n",
            "15590/17600 (epoch 44), train_loss = 1.268, time/batch = 0.158\n",
            "15591/17600 (epoch 44), train_loss = 1.377, time/batch = 0.165\n",
            "15592/17600 (epoch 44), train_loss = 1.305, time/batch = 0.166\n",
            "15593/17600 (epoch 44), train_loss = 1.280, time/batch = 0.161\n",
            "15594/17600 (epoch 44), train_loss = 1.320, time/batch = 0.144\n",
            "15595/17600 (epoch 44), train_loss = 1.307, time/batch = 0.159\n",
            "15596/17600 (epoch 44), train_loss = 1.309, time/batch = 0.157\n",
            "15597/17600 (epoch 44), train_loss = 1.293, time/batch = 0.166\n",
            "15598/17600 (epoch 44), train_loss = 1.250, time/batch = 0.180\n",
            "15599/17600 (epoch 44), train_loss = 1.272, time/batch = 0.159\n",
            "15600/17600 (epoch 44), train_loss = 1.303, time/batch = 0.159\n",
            "15601/17600 (epoch 44), train_loss = 1.199, time/batch = 0.171\n",
            "15602/17600 (epoch 44), train_loss = 1.274, time/batch = 0.150\n",
            "15603/17600 (epoch 44), train_loss = 1.303, time/batch = 0.157\n",
            "15604/17600 (epoch 44), train_loss = 1.269, time/batch = 0.155\n",
            "15605/17600 (epoch 44), train_loss = 1.297, time/batch = 0.173\n",
            "15606/17600 (epoch 44), train_loss = 1.239, time/batch = 0.160\n",
            "15607/17600 (epoch 44), train_loss = 1.344, time/batch = 0.165\n",
            "15608/17600 (epoch 44), train_loss = 1.250, time/batch = 0.173\n",
            "15609/17600 (epoch 44), train_loss = 1.311, time/batch = 0.196\n",
            "15610/17600 (epoch 44), train_loss = 1.336, time/batch = 0.162\n",
            "15611/17600 (epoch 44), train_loss = 1.321, time/batch = 0.187\n",
            "15612/17600 (epoch 44), train_loss = 1.237, time/batch = 0.155\n",
            "15613/17600 (epoch 44), train_loss = 1.304, time/batch = 0.168\n",
            "15614/17600 (epoch 44), train_loss = 1.301, time/batch = 0.153\n",
            "15615/17600 (epoch 44), train_loss = 1.287, time/batch = 0.232\n",
            "15616/17600 (epoch 44), train_loss = 1.270, time/batch = 0.277\n",
            "15617/17600 (epoch 44), train_loss = 1.302, time/batch = 0.246\n",
            "15618/17600 (epoch 44), train_loss = 1.315, time/batch = 0.311\n",
            "15619/17600 (epoch 44), train_loss = 1.320, time/batch = 0.280\n",
            "15620/17600 (epoch 44), train_loss = 1.290, time/batch = 0.285\n",
            "15621/17600 (epoch 44), train_loss = 1.261, time/batch = 0.267\n",
            "15622/17600 (epoch 44), train_loss = 1.273, time/batch = 0.262\n",
            "15623/17600 (epoch 44), train_loss = 1.243, time/batch = 0.255\n",
            "15624/17600 (epoch 44), train_loss = 1.311, time/batch = 0.279\n",
            "15625/17600 (epoch 44), train_loss = 1.337, time/batch = 0.314\n",
            "15626/17600 (epoch 44), train_loss = 1.283, time/batch = 0.177\n",
            "15627/17600 (epoch 44), train_loss = 1.335, time/batch = 0.168\n",
            "15628/17600 (epoch 44), train_loss = 1.361, time/batch = 0.171\n",
            "15629/17600 (epoch 44), train_loss = 1.273, time/batch = 0.166\n",
            "15630/17600 (epoch 44), train_loss = 1.298, time/batch = 0.170\n",
            "15631/17600 (epoch 44), train_loss = 1.265, time/batch = 0.156\n",
            "15632/17600 (epoch 44), train_loss = 1.316, time/batch = 0.162\n",
            "15633/17600 (epoch 44), train_loss = 1.301, time/batch = 0.160\n",
            "15634/17600 (epoch 44), train_loss = 1.276, time/batch = 0.159\n",
            "15635/17600 (epoch 44), train_loss = 1.284, time/batch = 0.175\n",
            "15636/17600 (epoch 44), train_loss = 1.266, time/batch = 0.166\n",
            "15637/17600 (epoch 44), train_loss = 1.329, time/batch = 0.157\n",
            "15638/17600 (epoch 44), train_loss = 1.216, time/batch = 0.165\n",
            "15639/17600 (epoch 44), train_loss = 1.290, time/batch = 0.159\n",
            "15640/17600 (epoch 44), train_loss = 1.315, time/batch = 0.153\n",
            "15641/17600 (epoch 44), train_loss = 1.284, time/batch = 0.167\n",
            "15642/17600 (epoch 44), train_loss = 1.300, time/batch = 0.161\n",
            "15643/17600 (epoch 44), train_loss = 1.239, time/batch = 0.151\n",
            "15644/17600 (epoch 44), train_loss = 1.236, time/batch = 0.161\n",
            "15645/17600 (epoch 44), train_loss = 1.254, time/batch = 0.158\n",
            "15646/17600 (epoch 44), train_loss = 1.314, time/batch = 0.166\n",
            "15647/17600 (epoch 44), train_loss = 1.358, time/batch = 0.148\n",
            "15648/17600 (epoch 44), train_loss = 1.313, time/batch = 0.177\n",
            "15649/17600 (epoch 44), train_loss = 1.312, time/batch = 0.154\n",
            "15650/17600 (epoch 44), train_loss = 1.243, time/batch = 0.167\n",
            "15651/17600 (epoch 44), train_loss = 1.272, time/batch = 0.147\n",
            "15652/17600 (epoch 44), train_loss = 1.294, time/batch = 0.151\n",
            "15653/17600 (epoch 44), train_loss = 1.281, time/batch = 0.157\n",
            "15654/17600 (epoch 44), train_loss = 1.294, time/batch = 0.168\n",
            "15655/17600 (epoch 44), train_loss = 1.319, time/batch = 0.151\n",
            "15656/17600 (epoch 44), train_loss = 1.312, time/batch = 0.162\n",
            "15657/17600 (epoch 44), train_loss = 1.292, time/batch = 0.153\n",
            "15658/17600 (epoch 44), train_loss = 1.225, time/batch = 0.162\n",
            "15659/17600 (epoch 44), train_loss = 1.237, time/batch = 0.153\n",
            "15660/17600 (epoch 44), train_loss = 1.326, time/batch = 0.169\n",
            "15661/17600 (epoch 44), train_loss = 1.295, time/batch = 0.173\n",
            "15662/17600 (epoch 44), train_loss = 1.285, time/batch = 0.165\n",
            "15663/17600 (epoch 44), train_loss = 1.288, time/batch = 0.164\n",
            "15664/17600 (epoch 44), train_loss = 1.202, time/batch = 0.172\n",
            "15665/17600 (epoch 44), train_loss = 1.265, time/batch = 0.151\n",
            "15666/17600 (epoch 44), train_loss = 1.265, time/batch = 0.158\n",
            "15667/17600 (epoch 44), train_loss = 1.245, time/batch = 0.179\n",
            "15668/17600 (epoch 44), train_loss = 1.243, time/batch = 0.170\n",
            "15669/17600 (epoch 44), train_loss = 1.297, time/batch = 0.154\n",
            "15670/17600 (epoch 44), train_loss = 1.354, time/batch = 0.161\n",
            "15671/17600 (epoch 44), train_loss = 1.269, time/batch = 0.143\n",
            "15672/17600 (epoch 44), train_loss = 1.270, time/batch = 0.162\n",
            "15673/17600 (epoch 44), train_loss = 1.234, time/batch = 0.156\n",
            "15674/17600 (epoch 44), train_loss = 1.308, time/batch = 0.183\n",
            "15675/17600 (epoch 44), train_loss = 1.318, time/batch = 0.147\n",
            "15676/17600 (epoch 44), train_loss = 1.238, time/batch = 0.173\n",
            "15677/17600 (epoch 44), train_loss = 1.332, time/batch = 0.153\n",
            "15678/17600 (epoch 44), train_loss = 1.300, time/batch = 0.172\n",
            "15679/17600 (epoch 44), train_loss = 1.279, time/batch = 0.146\n",
            "15680/17600 (epoch 44), train_loss = 1.287, time/batch = 0.177\n",
            "15681/17600 (epoch 44), train_loss = 1.306, time/batch = 0.160\n",
            "15682/17600 (epoch 44), train_loss = 1.307, time/batch = 0.170\n",
            "15683/17600 (epoch 44), train_loss = 1.316, time/batch = 0.152\n",
            "15684/17600 (epoch 44), train_loss = 1.271, time/batch = 0.156\n",
            "15685/17600 (epoch 44), train_loss = 1.324, time/batch = 0.155\n",
            "15686/17600 (epoch 44), train_loss = 1.271, time/batch = 0.166\n",
            "15687/17600 (epoch 44), train_loss = 1.305, time/batch = 0.203\n",
            "15688/17600 (epoch 44), train_loss = 1.308, time/batch = 0.281\n",
            "15689/17600 (epoch 44), train_loss = 1.274, time/batch = 0.274\n",
            "15690/17600 (epoch 44), train_loss = 1.264, time/batch = 0.304\n",
            "15691/17600 (epoch 44), train_loss = 1.293, time/batch = 0.294\n",
            "15692/17600 (epoch 44), train_loss = 1.327, time/batch = 0.278\n",
            "15693/17600 (epoch 44), train_loss = 1.374, time/batch = 0.286\n",
            "15694/17600 (epoch 44), train_loss = 1.290, time/batch = 0.270\n",
            "15695/17600 (epoch 44), train_loss = 1.290, time/batch = 0.288\n",
            "15696/17600 (epoch 44), train_loss = 1.293, time/batch = 0.260\n",
            "15697/17600 (epoch 44), train_loss = 1.271, time/batch = 0.288\n",
            "15698/17600 (epoch 44), train_loss = 1.316, time/batch = 0.195\n",
            "15699/17600 (epoch 44), train_loss = 1.308, time/batch = 0.164\n",
            "15700/17600 (epoch 44), train_loss = 1.277, time/batch = 0.149\n",
            "15701/17600 (epoch 44), train_loss = 1.351, time/batch = 0.169\n",
            "15702/17600 (epoch 44), train_loss = 1.389, time/batch = 0.160\n",
            "15703/17600 (epoch 44), train_loss = 1.278, time/batch = 0.173\n",
            "15704/17600 (epoch 44), train_loss = 1.298, time/batch = 0.169\n",
            "15705/17600 (epoch 44), train_loss = 1.256, time/batch = 0.177\n",
            "15706/17600 (epoch 44), train_loss = 1.289, time/batch = 0.152\n",
            "15707/17600 (epoch 44), train_loss = 1.271, time/batch = 0.165\n",
            "15708/17600 (epoch 44), train_loss = 1.265, time/batch = 0.154\n",
            "15709/17600 (epoch 44), train_loss = 1.288, time/batch = 0.166\n",
            "15710/17600 (epoch 44), train_loss = 1.358, time/batch = 0.165\n",
            "15711/17600 (epoch 44), train_loss = 1.330, time/batch = 0.167\n",
            "15712/17600 (epoch 44), train_loss = 1.247, time/batch = 0.175\n",
            "15713/17600 (epoch 44), train_loss = 1.294, time/batch = 0.161\n",
            "15714/17600 (epoch 44), train_loss = 1.300, time/batch = 0.155\n",
            "15715/17600 (epoch 44), train_loss = 1.278, time/batch = 0.181\n",
            "15716/17600 (epoch 44), train_loss = 1.310, time/batch = 0.176\n",
            "15717/17600 (epoch 44), train_loss = 1.275, time/batch = 0.180\n",
            "15718/17600 (epoch 44), train_loss = 1.264, time/batch = 0.164\n",
            "15719/17600 (epoch 44), train_loss = 1.328, time/batch = 0.160\n",
            "15720/17600 (epoch 44), train_loss = 1.326, time/batch = 0.154\n",
            "15721/17600 (epoch 44), train_loss = 1.341, time/batch = 0.157\n",
            "15722/17600 (epoch 44), train_loss = 1.256, time/batch = 0.144\n",
            "15723/17600 (epoch 44), train_loss = 1.287, time/batch = 0.181\n",
            "15724/17600 (epoch 44), train_loss = 1.278, time/batch = 0.160\n",
            "15725/17600 (epoch 44), train_loss = 1.346, time/batch = 0.156\n",
            "15726/17600 (epoch 44), train_loss = 1.343, time/batch = 0.158\n",
            "15727/17600 (epoch 44), train_loss = 1.237, time/batch = 0.180\n",
            "15728/17600 (epoch 44), train_loss = 1.329, time/batch = 0.151\n",
            "15729/17600 (epoch 44), train_loss = 1.278, time/batch = 0.176\n",
            "15730/17600 (epoch 44), train_loss = 1.314, time/batch = 0.164\n",
            "15731/17600 (epoch 44), train_loss = 1.306, time/batch = 0.158\n",
            "15732/17600 (epoch 44), train_loss = 1.274, time/batch = 0.167\n",
            "15733/17600 (epoch 44), train_loss = 1.314, time/batch = 0.154\n",
            "15734/17600 (epoch 44), train_loss = 1.306, time/batch = 0.152\n",
            "15735/17600 (epoch 44), train_loss = 1.278, time/batch = 0.164\n",
            "15736/17600 (epoch 44), train_loss = 1.256, time/batch = 0.176\n",
            "15737/17600 (epoch 44), train_loss = 1.235, time/batch = 0.166\n",
            "15738/17600 (epoch 44), train_loss = 1.369, time/batch = 0.156\n",
            "15739/17600 (epoch 44), train_loss = 1.307, time/batch = 0.171\n",
            "15740/17600 (epoch 44), train_loss = 1.299, time/batch = 0.159\n",
            "15741/17600 (epoch 44), train_loss = 1.307, time/batch = 0.178\n",
            "15742/17600 (epoch 44), train_loss = 1.325, time/batch = 0.152\n",
            "15743/17600 (epoch 44), train_loss = 1.333, time/batch = 0.162\n",
            "15744/17600 (epoch 44), train_loss = 1.380, time/batch = 0.163\n",
            "15745/17600 (epoch 44), train_loss = 1.349, time/batch = 0.155\n",
            "15746/17600 (epoch 44), train_loss = 1.301, time/batch = 0.154\n",
            "15747/17600 (epoch 44), train_loss = 1.303, time/batch = 0.160\n",
            "15748/17600 (epoch 44), train_loss = 1.279, time/batch = 0.169\n",
            "15749/17600 (epoch 44), train_loss = 1.288, time/batch = 0.173\n",
            "15750/17600 (epoch 44), train_loss = 1.267, time/batch = 0.161\n",
            "15751/17600 (epoch 44), train_loss = 1.226, time/batch = 0.179\n",
            "15752/17600 (epoch 44), train_loss = 1.277, time/batch = 0.165\n",
            "15753/17600 (epoch 44), train_loss = 1.283, time/batch = 0.149\n",
            "15754/17600 (epoch 44), train_loss = 1.237, time/batch = 0.183\n",
            "15755/17600 (epoch 44), train_loss = 1.241, time/batch = 0.162\n",
            "15756/17600 (epoch 44), train_loss = 1.272, time/batch = 0.163\n",
            "15757/17600 (epoch 44), train_loss = 1.306, time/batch = 0.169\n",
            "15758/17600 (epoch 44), train_loss = 1.229, time/batch = 0.202\n",
            "15759/17600 (epoch 44), train_loss = 1.299, time/batch = 0.308\n",
            "15760/17600 (epoch 44), train_loss = 1.266, time/batch = 0.271\n",
            "15761/17600 (epoch 44), train_loss = 1.293, time/batch = 0.282\n",
            "15762/17600 (epoch 44), train_loss = 1.275, time/batch = 0.285\n",
            "15763/17600 (epoch 44), train_loss = 1.330, time/batch = 0.284\n",
            "15764/17600 (epoch 44), train_loss = 1.294, time/batch = 0.279\n",
            "15765/17600 (epoch 44), train_loss = 1.320, time/batch = 0.272\n",
            "15766/17600 (epoch 44), train_loss = 1.285, time/batch = 0.263\n",
            "15767/17600 (epoch 44), train_loss = 1.292, time/batch = 0.304\n",
            "15768/17600 (epoch 44), train_loss = 1.315, time/batch = 0.291\n",
            "15769/17600 (epoch 44), train_loss = 1.362, time/batch = 0.174\n",
            "15770/17600 (epoch 44), train_loss = 1.334, time/batch = 0.153\n",
            "15771/17600 (epoch 44), train_loss = 1.332, time/batch = 0.156\n",
            "15772/17600 (epoch 44), train_loss = 1.349, time/batch = 0.177\n",
            "15773/17600 (epoch 44), train_loss = 1.362, time/batch = 0.154\n",
            "15774/17600 (epoch 44), train_loss = 1.308, time/batch = 0.159\n",
            "15775/17600 (epoch 44), train_loss = 1.355, time/batch = 0.151\n",
            "15776/17600 (epoch 44), train_loss = 1.292, time/batch = 0.169\n",
            "15777/17600 (epoch 44), train_loss = 1.287, time/batch = 0.144\n",
            "15778/17600 (epoch 44), train_loss = 1.353, time/batch = 0.158\n",
            "15779/17600 (epoch 44), train_loss = 1.257, time/batch = 0.163\n",
            "15780/17600 (epoch 44), train_loss = 1.253, time/batch = 0.159\n",
            "15781/17600 (epoch 44), train_loss = 1.328, time/batch = 0.155\n",
            "15782/17600 (epoch 44), train_loss = 1.276, time/batch = 0.166\n",
            "15783/17600 (epoch 44), train_loss = 1.335, time/batch = 0.153\n",
            "15784/17600 (epoch 44), train_loss = 1.282, time/batch = 0.165\n",
            "15785/17600 (epoch 44), train_loss = 1.309, time/batch = 0.172\n",
            "15786/17600 (epoch 44), train_loss = 1.295, time/batch = 0.167\n",
            "15787/17600 (epoch 44), train_loss = 1.308, time/batch = 0.154\n",
            "15788/17600 (epoch 44), train_loss = 1.338, time/batch = 0.191\n",
            "15789/17600 (epoch 44), train_loss = 1.297, time/batch = 0.155\n",
            "15790/17600 (epoch 44), train_loss = 1.365, time/batch = 0.155\n",
            "15791/17600 (epoch 44), train_loss = 1.357, time/batch = 0.162\n",
            "15792/17600 (epoch 44), train_loss = 1.368, time/batch = 0.166\n",
            "15793/17600 (epoch 44), train_loss = 1.283, time/batch = 0.169\n",
            "15794/17600 (epoch 44), train_loss = 1.317, time/batch = 0.152\n",
            "15795/17600 (epoch 44), train_loss = 1.324, time/batch = 0.171\n",
            "15796/17600 (epoch 44), train_loss = 1.322, time/batch = 0.152\n",
            "15797/17600 (epoch 44), train_loss = 1.324, time/batch = 0.168\n",
            "15798/17600 (epoch 44), train_loss = 1.361, time/batch = 0.189\n",
            "15799/17600 (epoch 44), train_loss = 1.311, time/batch = 0.172\n",
            "15800/17600 (epoch 44), train_loss = 1.344, time/batch = 0.147\n",
            "15801/17600 (epoch 44), train_loss = 1.279, time/batch = 0.157\n",
            "15802/17600 (epoch 44), train_loss = 1.294, time/batch = 0.158\n",
            "15803/17600 (epoch 44), train_loss = 1.264, time/batch = 0.179\n",
            "15804/17600 (epoch 44), train_loss = 1.359, time/batch = 0.169\n",
            "15805/17600 (epoch 44), train_loss = 1.347, time/batch = 0.156\n",
            "15806/17600 (epoch 44), train_loss = 1.345, time/batch = 0.162\n",
            "15807/17600 (epoch 44), train_loss = 1.328, time/batch = 0.158\n",
            "15808/17600 (epoch 44), train_loss = 1.367, time/batch = 0.162\n",
            "15809/17600 (epoch 44), train_loss = 1.332, time/batch = 0.171\n",
            "15810/17600 (epoch 44), train_loss = 1.389, time/batch = 0.166\n",
            "15811/17600 (epoch 44), train_loss = 1.358, time/batch = 0.161\n",
            "15812/17600 (epoch 44), train_loss = 1.344, time/batch = 0.162\n",
            "15813/17600 (epoch 44), train_loss = 1.294, time/batch = 0.168\n",
            "15814/17600 (epoch 44), train_loss = 1.371, time/batch = 0.156\n",
            "15815/17600 (epoch 44), train_loss = 1.348, time/batch = 0.162\n",
            "15816/17600 (epoch 44), train_loss = 1.385, time/batch = 0.167\n",
            "15817/17600 (epoch 44), train_loss = 1.354, time/batch = 0.163\n",
            "15818/17600 (epoch 44), train_loss = 1.274, time/batch = 0.152\n",
            "15819/17600 (epoch 44), train_loss = 1.303, time/batch = 0.169\n",
            "15820/17600 (epoch 44), train_loss = 1.350, time/batch = 0.147\n",
            "15821/17600 (epoch 44), train_loss = 1.362, time/batch = 0.167\n",
            "15822/17600 (epoch 44), train_loss = 1.320, time/batch = 0.149\n",
            "15823/17600 (epoch 44), train_loss = 1.310, time/batch = 0.179\n",
            "15824/17600 (epoch 44), train_loss = 1.317, time/batch = 0.154\n",
            "15825/17600 (epoch 44), train_loss = 1.358, time/batch = 0.159\n",
            "15826/17600 (epoch 44), train_loss = 1.303, time/batch = 0.152\n",
            "15827/17600 (epoch 44), train_loss = 1.332, time/batch = 0.164\n",
            "15828/17600 (epoch 44), train_loss = 1.300, time/batch = 0.149\n",
            "15829/17600 (epoch 44), train_loss = 1.344, time/batch = 0.173\n",
            "15830/17600 (epoch 44), train_loss = 1.351, time/batch = 0.220\n",
            "15831/17600 (epoch 44), train_loss = 1.314, time/batch = 0.262\n",
            "15832/17600 (epoch 44), train_loss = 1.292, time/batch = 0.256\n",
            "15833/17600 (epoch 44), train_loss = 1.312, time/batch = 0.276\n",
            "15834/17600 (epoch 44), train_loss = 1.335, time/batch = 0.275\n",
            "15835/17600 (epoch 44), train_loss = 1.308, time/batch = 0.276\n",
            "15836/17600 (epoch 44), train_loss = 1.331, time/batch = 0.277\n",
            "15837/17600 (epoch 44), train_loss = 1.317, time/batch = 0.272\n",
            "15838/17600 (epoch 44), train_loss = 1.308, time/batch = 0.266\n",
            "15839/17600 (epoch 44), train_loss = 1.300, time/batch = 0.306\n",
            "15840/17600 (epoch 45), train_loss = 1.096, time/batch = 0.180\n",
            "15841/17600 (epoch 45), train_loss = 1.331, time/batch = 0.153\n",
            "15842/17600 (epoch 45), train_loss = 1.324, time/batch = 0.164\n",
            "15843/17600 (epoch 45), train_loss = 1.270, time/batch = 0.156\n",
            "15844/17600 (epoch 45), train_loss = 1.367, time/batch = 0.173\n",
            "15845/17600 (epoch 45), train_loss = 1.341, time/batch = 0.148\n",
            "15846/17600 (epoch 45), train_loss = 1.286, time/batch = 0.179\n",
            "15847/17600 (epoch 45), train_loss = 1.272, time/batch = 0.164\n",
            "15848/17600 (epoch 45), train_loss = 1.314, time/batch = 0.158\n",
            "15849/17600 (epoch 45), train_loss = 1.366, time/batch = 0.157\n",
            "15850/17600 (epoch 45), train_loss = 1.295, time/batch = 0.156\n",
            "15851/17600 (epoch 45), train_loss = 1.308, time/batch = 0.159\n",
            "15852/17600 (epoch 45), train_loss = 1.292, time/batch = 0.181\n",
            "15853/17600 (epoch 45), train_loss = 1.278, time/batch = 0.178\n",
            "15854/17600 (epoch 45), train_loss = 1.258, time/batch = 0.171\n",
            "15855/17600 (epoch 45), train_loss = 1.323, time/batch = 0.150\n",
            "15856/17600 (epoch 45), train_loss = 1.335, time/batch = 0.157\n",
            "15857/17600 (epoch 45), train_loss = 1.289, time/batch = 0.145\n",
            "15858/17600 (epoch 45), train_loss = 1.302, time/batch = 0.161\n",
            "15859/17600 (epoch 45), train_loss = 1.346, time/batch = 0.166\n",
            "15860/17600 (epoch 45), train_loss = 1.272, time/batch = 0.170\n",
            "15861/17600 (epoch 45), train_loss = 1.324, time/batch = 0.166\n",
            "15862/17600 (epoch 45), train_loss = 1.290, time/batch = 0.162\n",
            "15863/17600 (epoch 45), train_loss = 1.262, time/batch = 0.153\n",
            "15864/17600 (epoch 45), train_loss = 1.346, time/batch = 0.162\n",
            "15865/17600 (epoch 45), train_loss = 1.279, time/batch = 0.171\n",
            "15866/17600 (epoch 45), train_loss = 1.325, time/batch = 0.170\n",
            "15867/17600 (epoch 45), train_loss = 1.309, time/batch = 0.153\n",
            "15868/17600 (epoch 45), train_loss = 1.287, time/batch = 0.175\n",
            "15869/17600 (epoch 45), train_loss = 1.222, time/batch = 0.164\n",
            "15870/17600 (epoch 45), train_loss = 1.330, time/batch = 0.164\n",
            "15871/17600 (epoch 45), train_loss = 1.307, time/batch = 0.172\n",
            "15872/17600 (epoch 45), train_loss = 1.284, time/batch = 0.172\n",
            "15873/17600 (epoch 45), train_loss = 1.287, time/batch = 0.157\n",
            "15874/17600 (epoch 45), train_loss = 1.291, time/batch = 0.163\n",
            "15875/17600 (epoch 45), train_loss = 1.305, time/batch = 0.166\n",
            "15876/17600 (epoch 45), train_loss = 1.333, time/batch = 0.156\n",
            "15877/17600 (epoch 45), train_loss = 1.329, time/batch = 0.171\n",
            "15878/17600 (epoch 45), train_loss = 1.308, time/batch = 0.181\n",
            "15879/17600 (epoch 45), train_loss = 1.286, time/batch = 0.170\n",
            "15880/17600 (epoch 45), train_loss = 1.331, time/batch = 0.161\n",
            "15881/17600 (epoch 45), train_loss = 1.269, time/batch = 0.156\n",
            "15882/17600 (epoch 45), train_loss = 1.324, time/batch = 0.177\n",
            "15883/17600 (epoch 45), train_loss = 1.246, time/batch = 0.163\n",
            "15884/17600 (epoch 45), train_loss = 1.312, time/batch = 0.181\n",
            "15885/17600 (epoch 45), train_loss = 1.285, time/batch = 0.160\n",
            "15886/17600 (epoch 45), train_loss = 1.239, time/batch = 0.160\n",
            "15887/17600 (epoch 45), train_loss = 1.299, time/batch = 0.151\n",
            "15888/17600 (epoch 45), train_loss = 1.250, time/batch = 0.165\n",
            "15889/17600 (epoch 45), train_loss = 1.266, time/batch = 0.146\n",
            "15890/17600 (epoch 45), train_loss = 1.292, time/batch = 0.167\n",
            "15891/17600 (epoch 45), train_loss = 1.281, time/batch = 0.161\n",
            "15892/17600 (epoch 45), train_loss = 1.254, time/batch = 0.163\n",
            "15893/17600 (epoch 45), train_loss = 1.281, time/batch = 0.153\n",
            "15894/17600 (epoch 45), train_loss = 1.325, time/batch = 0.173\n",
            "15895/17600 (epoch 45), train_loss = 1.303, time/batch = 0.152\n",
            "15896/17600 (epoch 45), train_loss = 1.302, time/batch = 0.175\n",
            "15897/17600 (epoch 45), train_loss = 1.297, time/batch = 0.151\n",
            "15898/17600 (epoch 45), train_loss = 1.276, time/batch = 0.170\n",
            "15899/17600 (epoch 45), train_loss = 1.303, time/batch = 0.150\n",
            "15900/17600 (epoch 45), train_loss = 1.300, time/batch = 0.240\n",
            "15901/17600 (epoch 45), train_loss = 1.241, time/batch = 0.272\n",
            "15902/17600 (epoch 45), train_loss = 1.243, time/batch = 0.280\n",
            "15903/17600 (epoch 45), train_loss = 1.263, time/batch = 0.267\n",
            "15904/17600 (epoch 45), train_loss = 1.293, time/batch = 0.248\n",
            "15905/17600 (epoch 45), train_loss = 1.259, time/batch = 0.261\n",
            "15906/17600 (epoch 45), train_loss = 1.296, time/batch = 0.250\n",
            "15907/17600 (epoch 45), train_loss = 1.256, time/batch = 0.261\n",
            "15908/17600 (epoch 45), train_loss = 1.304, time/batch = 0.252\n",
            "15909/17600 (epoch 45), train_loss = 1.235, time/batch = 0.276\n",
            "15910/17600 (epoch 45), train_loss = 1.280, time/batch = 0.298\n",
            "15911/17600 (epoch 45), train_loss = 1.326, time/batch = 0.151\n",
            "15912/17600 (epoch 45), train_loss = 1.288, time/batch = 0.161\n",
            "15913/17600 (epoch 45), train_loss = 1.357, time/batch = 0.152\n",
            "15914/17600 (epoch 45), train_loss = 1.288, time/batch = 0.167\n",
            "15915/17600 (epoch 45), train_loss = 1.290, time/batch = 0.150\n",
            "15916/17600 (epoch 45), train_loss = 1.246, time/batch = 0.177\n",
            "15917/17600 (epoch 45), train_loss = 1.301, time/batch = 0.155\n",
            "15918/17600 (epoch 45), train_loss = 1.307, time/batch = 0.155\n",
            "15919/17600 (epoch 45), train_loss = 1.383, time/batch = 0.152\n",
            "15920/17600 (epoch 45), train_loss = 1.353, time/batch = 0.158\n",
            "15921/17600 (epoch 45), train_loss = 1.345, time/batch = 0.152\n",
            "15922/17600 (epoch 45), train_loss = 1.326, time/batch = 0.185\n",
            "15923/17600 (epoch 45), train_loss = 1.299, time/batch = 0.159\n",
            "15924/17600 (epoch 45), train_loss = 1.275, time/batch = 0.156\n",
            "15925/17600 (epoch 45), train_loss = 1.309, time/batch = 0.161\n",
            "15926/17600 (epoch 45), train_loss = 1.301, time/batch = 0.161\n",
            "15927/17600 (epoch 45), train_loss = 1.240, time/batch = 0.158\n",
            "15928/17600 (epoch 45), train_loss = 1.281, time/batch = 0.183\n",
            "15929/17600 (epoch 45), train_loss = 1.220, time/batch = 0.161\n",
            "15930/17600 (epoch 45), train_loss = 1.284, time/batch = 0.164\n",
            "15931/17600 (epoch 45), train_loss = 1.271, time/batch = 0.161\n",
            "15932/17600 (epoch 45), train_loss = 1.248, time/batch = 0.165\n",
            "15933/17600 (epoch 45), train_loss = 1.303, time/batch = 0.160\n",
            "15934/17600 (epoch 45), train_loss = 1.287, time/batch = 0.163\n",
            "15935/17600 (epoch 45), train_loss = 1.249, time/batch = 0.161\n",
            "15936/17600 (epoch 45), train_loss = 1.223, time/batch = 0.180\n",
            "15937/17600 (epoch 45), train_loss = 1.276, time/batch = 0.182\n",
            "15938/17600 (epoch 45), train_loss = 1.310, time/batch = 0.178\n",
            "15939/17600 (epoch 45), train_loss = 1.269, time/batch = 0.156\n",
            "15940/17600 (epoch 45), train_loss = 1.244, time/batch = 0.161\n",
            "15941/17600 (epoch 45), train_loss = 1.328, time/batch = 0.173\n",
            "15942/17600 (epoch 45), train_loss = 1.266, time/batch = 0.164\n",
            "15943/17600 (epoch 45), train_loss = 1.374, time/batch = 0.152\n",
            "15944/17600 (epoch 45), train_loss = 1.304, time/batch = 0.169\n",
            "15945/17600 (epoch 45), train_loss = 1.278, time/batch = 0.163\n",
            "15946/17600 (epoch 45), train_loss = 1.318, time/batch = 0.172\n",
            "15947/17600 (epoch 45), train_loss = 1.306, time/batch = 0.167\n",
            "15948/17600 (epoch 45), train_loss = 1.307, time/batch = 0.168\n",
            "15949/17600 (epoch 45), train_loss = 1.291, time/batch = 0.166\n",
            "15950/17600 (epoch 45), train_loss = 1.248, time/batch = 0.180\n",
            "15951/17600 (epoch 45), train_loss = 1.270, time/batch = 0.155\n",
            "15952/17600 (epoch 45), train_loss = 1.301, time/batch = 0.170\n",
            "15953/17600 (epoch 45), train_loss = 1.198, time/batch = 0.174\n",
            "15954/17600 (epoch 45), train_loss = 1.272, time/batch = 0.159\n",
            "15955/17600 (epoch 45), train_loss = 1.301, time/batch = 0.186\n",
            "15956/17600 (epoch 45), train_loss = 1.267, time/batch = 0.164\n",
            "15957/17600 (epoch 45), train_loss = 1.295, time/batch = 0.147\n",
            "15958/17600 (epoch 45), train_loss = 1.237, time/batch = 0.160\n",
            "15959/17600 (epoch 45), train_loss = 1.342, time/batch = 0.163\n",
            "15960/17600 (epoch 45), train_loss = 1.248, time/batch = 0.162\n",
            "15961/17600 (epoch 45), train_loss = 1.310, time/batch = 0.158\n",
            "15962/17600 (epoch 45), train_loss = 1.334, time/batch = 0.172\n",
            "15963/17600 (epoch 45), train_loss = 1.319, time/batch = 0.157\n",
            "15964/17600 (epoch 45), train_loss = 1.235, time/batch = 0.165\n",
            "15965/17600 (epoch 45), train_loss = 1.302, time/batch = 0.156\n",
            "15966/17600 (epoch 45), train_loss = 1.299, time/batch = 0.177\n",
            "15967/17600 (epoch 45), train_loss = 1.286, time/batch = 0.167\n",
            "15968/17600 (epoch 45), train_loss = 1.268, time/batch = 0.160\n",
            "15969/17600 (epoch 45), train_loss = 1.301, time/batch = 0.161\n",
            "15970/17600 (epoch 45), train_loss = 1.313, time/batch = 0.166\n",
            "15971/17600 (epoch 45), train_loss = 1.318, time/batch = 0.219\n",
            "15972/17600 (epoch 45), train_loss = 1.289, time/batch = 0.240\n",
            "15973/17600 (epoch 45), train_loss = 1.259, time/batch = 0.267\n",
            "15974/17600 (epoch 45), train_loss = 1.271, time/batch = 0.283\n",
            "15975/17600 (epoch 45), train_loss = 1.242, time/batch = 0.288\n",
            "15976/17600 (epoch 45), train_loss = 1.309, time/batch = 0.271\n",
            "15977/17600 (epoch 45), train_loss = 1.335, time/batch = 0.284\n",
            "15978/17600 (epoch 45), train_loss = 1.281, time/batch = 0.296\n",
            "15979/17600 (epoch 45), train_loss = 1.334, time/batch = 0.271\n",
            "15980/17600 (epoch 45), train_loss = 1.359, time/batch = 0.272\n",
            "15981/17600 (epoch 45), train_loss = 1.271, time/batch = 0.287\n",
            "15982/17600 (epoch 45), train_loss = 1.297, time/batch = 0.186\n",
            "15983/17600 (epoch 45), train_loss = 1.263, time/batch = 0.169\n",
            "15984/17600 (epoch 45), train_loss = 1.314, time/batch = 0.150\n",
            "15985/17600 (epoch 45), train_loss = 1.300, time/batch = 0.157\n",
            "15986/17600 (epoch 45), train_loss = 1.274, time/batch = 0.149\n",
            "15987/17600 (epoch 45), train_loss = 1.282, time/batch = 0.166\n",
            "15988/17600 (epoch 45), train_loss = 1.264, time/batch = 0.162\n",
            "15989/17600 (epoch 45), train_loss = 1.327, time/batch = 0.162\n",
            "15990/17600 (epoch 45), train_loss = 1.214, time/batch = 0.172\n",
            "15991/17600 (epoch 45), train_loss = 1.289, time/batch = 0.165\n",
            "15992/17600 (epoch 45), train_loss = 1.313, time/batch = 0.153\n",
            "15993/17600 (epoch 45), train_loss = 1.283, time/batch = 0.156\n",
            "15994/17600 (epoch 45), train_loss = 1.298, time/batch = 0.163\n",
            "15995/17600 (epoch 45), train_loss = 1.238, time/batch = 0.149\n",
            "15996/17600 (epoch 45), train_loss = 1.234, time/batch = 0.176\n",
            "15997/17600 (epoch 45), train_loss = 1.253, time/batch = 0.140\n",
            "15998/17600 (epoch 45), train_loss = 1.313, time/batch = 0.158\n",
            "15999/17600 (epoch 45), train_loss = 1.356, time/batch = 0.156\n",
            "16000/17600 (epoch 45), train_loss = 1.312, time/batch = 0.156\n",
            "Model saved to checkpoints/dracula/dracula!\n",
            "16001/17600 (epoch 45), train_loss = 1.310, time/batch = 0.158\n",
            "16002/17600 (epoch 45), train_loss = 1.242, time/batch = 0.163\n",
            "16003/17600 (epoch 45), train_loss = 1.269, time/batch = 0.157\n",
            "16004/17600 (epoch 45), train_loss = 1.293, time/batch = 0.164\n",
            "16005/17600 (epoch 45), train_loss = 1.278, time/batch = 0.175\n",
            "16006/17600 (epoch 45), train_loss = 1.292, time/batch = 0.166\n",
            "16007/17600 (epoch 45), train_loss = 1.317, time/batch = 0.153\n",
            "16008/17600 (epoch 45), train_loss = 1.310, time/batch = 0.165\n",
            "16009/17600 (epoch 45), train_loss = 1.291, time/batch = 0.146\n",
            "16010/17600 (epoch 45), train_loss = 1.223, time/batch = 0.162\n",
            "16011/17600 (epoch 45), train_loss = 1.235, time/batch = 0.169\n",
            "16012/17600 (epoch 45), train_loss = 1.325, time/batch = 0.162\n",
            "16013/17600 (epoch 45), train_loss = 1.293, time/batch = 0.156\n",
            "16014/17600 (epoch 45), train_loss = 1.283, time/batch = 0.205\n",
            "16015/17600 (epoch 45), train_loss = 1.287, time/batch = 0.261\n",
            "16016/17600 (epoch 45), train_loss = 1.200, time/batch = 0.282\n",
            "16017/17600 (epoch 45), train_loss = 1.263, time/batch = 0.291\n",
            "16018/17600 (epoch 45), train_loss = 1.263, time/batch = 0.264\n",
            "16019/17600 (epoch 45), train_loss = 1.243, time/batch = 0.282\n",
            "16020/17600 (epoch 45), train_loss = 1.242, time/batch = 0.267\n",
            "16021/17600 (epoch 45), train_loss = 1.296, time/batch = 0.294\n",
            "16022/17600 (epoch 45), train_loss = 1.352, time/batch = 0.272\n",
            "16023/17600 (epoch 45), train_loss = 1.268, time/batch = 0.272\n",
            "16024/17600 (epoch 45), train_loss = 1.268, time/batch = 0.273\n",
            "16025/17600 (epoch 45), train_loss = 1.232, time/batch = 0.159\n",
            "16026/17600 (epoch 45), train_loss = 1.306, time/batch = 0.157\n",
            "16027/17600 (epoch 45), train_loss = 1.317, time/batch = 0.168\n",
            "16028/17600 (epoch 45), train_loss = 1.237, time/batch = 0.158\n",
            "16029/17600 (epoch 45), train_loss = 1.330, time/batch = 0.156\n",
            "16030/17600 (epoch 45), train_loss = 1.299, time/batch = 0.172\n",
            "16031/17600 (epoch 45), train_loss = 1.278, time/batch = 0.162\n",
            "16032/17600 (epoch 45), train_loss = 1.286, time/batch = 0.160\n",
            "16033/17600 (epoch 45), train_loss = 1.305, time/batch = 0.161\n",
            "16034/17600 (epoch 45), train_loss = 1.305, time/batch = 0.158\n",
            "16035/17600 (epoch 45), train_loss = 1.314, time/batch = 0.164\n",
            "16036/17600 (epoch 45), train_loss = 1.269, time/batch = 0.168\n",
            "16037/17600 (epoch 45), train_loss = 1.322, time/batch = 0.161\n",
            "16038/17600 (epoch 45), train_loss = 1.269, time/batch = 0.154\n",
            "16039/17600 (epoch 45), train_loss = 1.303, time/batch = 0.159\n",
            "16040/17600 (epoch 45), train_loss = 1.306, time/batch = 0.161\n",
            "16041/17600 (epoch 45), train_loss = 1.272, time/batch = 0.171\n",
            "16042/17600 (epoch 45), train_loss = 1.262, time/batch = 0.167\n",
            "16043/17600 (epoch 45), train_loss = 1.291, time/batch = 0.165\n",
            "16044/17600 (epoch 45), train_loss = 1.325, time/batch = 0.153\n",
            "16045/17600 (epoch 45), train_loss = 1.372, time/batch = 0.158\n",
            "16046/17600 (epoch 45), train_loss = 1.288, time/batch = 0.172\n",
            "16047/17600 (epoch 45), train_loss = 1.288, time/batch = 0.157\n",
            "16048/17600 (epoch 45), train_loss = 1.292, time/batch = 0.184\n",
            "16049/17600 (epoch 45), train_loss = 1.269, time/batch = 0.162\n",
            "16050/17600 (epoch 45), train_loss = 1.314, time/batch = 0.180\n",
            "16051/17600 (epoch 45), train_loss = 1.307, time/batch = 0.168\n",
            "16052/17600 (epoch 45), train_loss = 1.276, time/batch = 0.163\n",
            "16053/17600 (epoch 45), train_loss = 1.349, time/batch = 0.166\n",
            "16054/17600 (epoch 45), train_loss = 1.387, time/batch = 0.171\n",
            "16055/17600 (epoch 45), train_loss = 1.276, time/batch = 0.163\n",
            "16056/17600 (epoch 45), train_loss = 1.295, time/batch = 0.173\n",
            "16057/17600 (epoch 45), train_loss = 1.254, time/batch = 0.172\n",
            "16058/17600 (epoch 45), train_loss = 1.286, time/batch = 0.156\n",
            "16059/17600 (epoch 45), train_loss = 1.269, time/batch = 0.176\n",
            "16060/17600 (epoch 45), train_loss = 1.263, time/batch = 0.151\n",
            "16061/17600 (epoch 45), train_loss = 1.286, time/batch = 0.177\n",
            "16062/17600 (epoch 45), train_loss = 1.356, time/batch = 0.155\n",
            "16063/17600 (epoch 45), train_loss = 1.329, time/batch = 0.166\n",
            "16064/17600 (epoch 45), train_loss = 1.246, time/batch = 0.154\n",
            "16065/17600 (epoch 45), train_loss = 1.292, time/batch = 0.167\n",
            "16066/17600 (epoch 45), train_loss = 1.298, time/batch = 0.158\n",
            "16067/17600 (epoch 45), train_loss = 1.276, time/batch = 0.179\n",
            "16068/17600 (epoch 45), train_loss = 1.308, time/batch = 0.155\n",
            "16069/17600 (epoch 45), train_loss = 1.273, time/batch = 0.159\n",
            "16070/17600 (epoch 45), train_loss = 1.262, time/batch = 0.157\n",
            "16071/17600 (epoch 45), train_loss = 1.326, time/batch = 0.163\n",
            "16072/17600 (epoch 45), train_loss = 1.324, time/batch = 0.151\n",
            "16073/17600 (epoch 45), train_loss = 1.338, time/batch = 0.193\n",
            "16074/17600 (epoch 45), train_loss = 1.254, time/batch = 0.158\n",
            "16075/17600 (epoch 45), train_loss = 1.285, time/batch = 0.172\n",
            "16076/17600 (epoch 45), train_loss = 1.276, time/batch = 0.159\n",
            "16077/17600 (epoch 45), train_loss = 1.344, time/batch = 0.153\n",
            "16078/17600 (epoch 45), train_loss = 1.342, time/batch = 0.149\n",
            "16079/17600 (epoch 45), train_loss = 1.236, time/batch = 0.165\n",
            "16080/17600 (epoch 45), train_loss = 1.327, time/batch = 0.168\n",
            "16081/17600 (epoch 45), train_loss = 1.276, time/batch = 0.159\n",
            "16082/17600 (epoch 45), train_loss = 1.312, time/batch = 0.141\n",
            "16083/17600 (epoch 45), train_loss = 1.304, time/batch = 0.164\n",
            "16084/17600 (epoch 45), train_loss = 1.272, time/batch = 0.154\n",
            "16085/17600 (epoch 45), train_loss = 1.312, time/batch = 0.163\n",
            "16086/17600 (epoch 45), train_loss = 1.304, time/batch = 0.254\n",
            "16087/17600 (epoch 45), train_loss = 1.276, time/batch = 0.257\n",
            "16088/17600 (epoch 45), train_loss = 1.255, time/batch = 0.257\n",
            "16089/17600 (epoch 45), train_loss = 1.233, time/batch = 0.303\n",
            "16090/17600 (epoch 45), train_loss = 1.367, time/batch = 0.282\n",
            "16091/17600 (epoch 45), train_loss = 1.305, time/batch = 0.263\n",
            "16092/17600 (epoch 45), train_loss = 1.297, time/batch = 0.280\n",
            "16093/17600 (epoch 45), train_loss = 1.306, time/batch = 0.270\n",
            "16094/17600 (epoch 45), train_loss = 1.324, time/batch = 0.304\n",
            "16095/17600 (epoch 45), train_loss = 1.331, time/batch = 0.274\n",
            "16096/17600 (epoch 45), train_loss = 1.378, time/batch = 0.244\n",
            "16097/17600 (epoch 45), train_loss = 1.348, time/batch = 0.159\n",
            "16098/17600 (epoch 45), train_loss = 1.300, time/batch = 0.177\n",
            "16099/17600 (epoch 45), train_loss = 1.302, time/batch = 0.152\n",
            "16100/17600 (epoch 45), train_loss = 1.277, time/batch = 0.174\n",
            "16101/17600 (epoch 45), train_loss = 1.286, time/batch = 0.152\n",
            "16102/17600 (epoch 45), train_loss = 1.265, time/batch = 0.172\n",
            "16103/17600 (epoch 45), train_loss = 1.224, time/batch = 0.164\n",
            "16104/17600 (epoch 45), train_loss = 1.275, time/batch = 0.173\n",
            "16105/17600 (epoch 45), train_loss = 1.281, time/batch = 0.158\n",
            "16106/17600 (epoch 45), train_loss = 1.235, time/batch = 0.165\n",
            "16107/17600 (epoch 45), train_loss = 1.239, time/batch = 0.150\n",
            "16108/17600 (epoch 45), train_loss = 1.270, time/batch = 0.163\n",
            "16109/17600 (epoch 45), train_loss = 1.304, time/batch = 0.151\n",
            "16110/17600 (epoch 45), train_loss = 1.227, time/batch = 0.155\n",
            "16111/17600 (epoch 45), train_loss = 1.297, time/batch = 0.169\n",
            "16112/17600 (epoch 45), train_loss = 1.264, time/batch = 0.165\n",
            "16113/17600 (epoch 45), train_loss = 1.292, time/batch = 0.145\n",
            "16114/17600 (epoch 45), train_loss = 1.274, time/batch = 0.179\n",
            "16115/17600 (epoch 45), train_loss = 1.328, time/batch = 0.145\n",
            "16116/17600 (epoch 45), train_loss = 1.292, time/batch = 0.164\n",
            "16117/17600 (epoch 45), train_loss = 1.318, time/batch = 0.171\n",
            "16118/17600 (epoch 45), train_loss = 1.283, time/batch = 0.172\n",
            "16119/17600 (epoch 45), train_loss = 1.290, time/batch = 0.159\n",
            "16120/17600 (epoch 45), train_loss = 1.313, time/batch = 0.162\n",
            "16121/17600 (epoch 45), train_loss = 1.361, time/batch = 0.158\n",
            "16122/17600 (epoch 45), train_loss = 1.333, time/batch = 0.164\n",
            "16123/17600 (epoch 45), train_loss = 1.330, time/batch = 0.169\n",
            "16124/17600 (epoch 45), train_loss = 1.347, time/batch = 0.183\n",
            "16125/17600 (epoch 45), train_loss = 1.360, time/batch = 0.163\n",
            "16126/17600 (epoch 45), train_loss = 1.307, time/batch = 0.151\n",
            "16127/17600 (epoch 45), train_loss = 1.354, time/batch = 0.143\n",
            "16128/17600 (epoch 45), train_loss = 1.290, time/batch = 0.159\n",
            "16129/17600 (epoch 45), train_loss = 1.285, time/batch = 0.148\n",
            "16130/17600 (epoch 45), train_loss = 1.351, time/batch = 0.173\n",
            "16131/17600 (epoch 45), train_loss = 1.256, time/batch = 0.169\n",
            "16132/17600 (epoch 45), train_loss = 1.252, time/batch = 0.152\n",
            "16133/17600 (epoch 45), train_loss = 1.326, time/batch = 0.153\n",
            "16134/17600 (epoch 45), train_loss = 1.274, time/batch = 0.161\n",
            "16135/17600 (epoch 45), train_loss = 1.334, time/batch = 0.151\n",
            "16136/17600 (epoch 45), train_loss = 1.280, time/batch = 0.180\n",
            "16137/17600 (epoch 45), train_loss = 1.307, time/batch = 0.152\n",
            "16138/17600 (epoch 45), train_loss = 1.294, time/batch = 0.159\n",
            "16139/17600 (epoch 45), train_loss = 1.306, time/batch = 0.154\n",
            "16140/17600 (epoch 45), train_loss = 1.336, time/batch = 0.173\n",
            "16141/17600 (epoch 45), train_loss = 1.295, time/batch = 0.156\n",
            "16142/17600 (epoch 45), train_loss = 1.364, time/batch = 0.164\n",
            "16143/17600 (epoch 45), train_loss = 1.355, time/batch = 0.175\n",
            "16144/17600 (epoch 45), train_loss = 1.366, time/batch = 0.160\n",
            "16145/17600 (epoch 45), train_loss = 1.281, time/batch = 0.161\n",
            "16146/17600 (epoch 45), train_loss = 1.315, time/batch = 0.152\n",
            "16147/17600 (epoch 45), train_loss = 1.323, time/batch = 0.153\n",
            "16148/17600 (epoch 45), train_loss = 1.319, time/batch = 0.171\n",
            "16149/17600 (epoch 45), train_loss = 1.322, time/batch = 0.166\n",
            "16150/17600 (epoch 45), train_loss = 1.359, time/batch = 0.165\n",
            "16151/17600 (epoch 45), train_loss = 1.309, time/batch = 0.157\n",
            "16152/17600 (epoch 45), train_loss = 1.342, time/batch = 0.160\n",
            "16153/17600 (epoch 45), train_loss = 1.277, time/batch = 0.154\n",
            "16154/17600 (epoch 45), train_loss = 1.292, time/batch = 0.160\n",
            "16155/17600 (epoch 45), train_loss = 1.263, time/batch = 0.165\n",
            "16156/17600 (epoch 45), train_loss = 1.357, time/batch = 0.165\n",
            "16157/17600 (epoch 45), train_loss = 1.345, time/batch = 0.141\n",
            "16158/17600 (epoch 45), train_loss = 1.343, time/batch = 0.213\n",
            "16159/17600 (epoch 45), train_loss = 1.326, time/batch = 0.272\n",
            "16160/17600 (epoch 45), train_loss = 1.365, time/batch = 0.282\n",
            "16161/17600 (epoch 45), train_loss = 1.330, time/batch = 0.276\n",
            "16162/17600 (epoch 45), train_loss = 1.387, time/batch = 0.255\n",
            "16163/17600 (epoch 45), train_loss = 1.356, time/batch = 0.271\n",
            "16164/17600 (epoch 45), train_loss = 1.343, time/batch = 0.259\n",
            "16165/17600 (epoch 45), train_loss = 1.292, time/batch = 0.278\n",
            "16166/17600 (epoch 45), train_loss = 1.369, time/batch = 0.300\n",
            "16167/17600 (epoch 45), train_loss = 1.346, time/batch = 0.271\n",
            "16168/17600 (epoch 45), train_loss = 1.384, time/batch = 0.293\n",
            "16169/17600 (epoch 45), train_loss = 1.353, time/batch = 0.166\n",
            "16170/17600 (epoch 45), train_loss = 1.272, time/batch = 0.172\n",
            "16171/17600 (epoch 45), train_loss = 1.302, time/batch = 0.157\n",
            "16172/17600 (epoch 45), train_loss = 1.348, time/batch = 0.159\n",
            "16173/17600 (epoch 45), train_loss = 1.361, time/batch = 0.161\n",
            "16174/17600 (epoch 45), train_loss = 1.317, time/batch = 0.178\n",
            "16175/17600 (epoch 45), train_loss = 1.309, time/batch = 0.154\n",
            "16176/17600 (epoch 45), train_loss = 1.315, time/batch = 0.165\n",
            "16177/17600 (epoch 45), train_loss = 1.356, time/batch = 0.153\n",
            "16178/17600 (epoch 45), train_loss = 1.302, time/batch = 0.161\n",
            "16179/17600 (epoch 45), train_loss = 1.330, time/batch = 0.159\n",
            "16180/17600 (epoch 45), train_loss = 1.299, time/batch = 0.177\n",
            "16181/17600 (epoch 45), train_loss = 1.342, time/batch = 0.157\n",
            "16182/17600 (epoch 45), train_loss = 1.349, time/batch = 0.163\n",
            "16183/17600 (epoch 45), train_loss = 1.312, time/batch = 0.153\n",
            "16184/17600 (epoch 45), train_loss = 1.290, time/batch = 0.170\n",
            "16185/17600 (epoch 45), train_loss = 1.310, time/batch = 0.155\n",
            "16186/17600 (epoch 45), train_loss = 1.333, time/batch = 0.168\n",
            "16187/17600 (epoch 45), train_loss = 1.307, time/batch = 0.178\n",
            "16188/17600 (epoch 45), train_loss = 1.330, time/batch = 0.164\n",
            "16189/17600 (epoch 45), train_loss = 1.315, time/batch = 0.149\n",
            "16190/17600 (epoch 45), train_loss = 1.306, time/batch = 0.157\n",
            "16191/17600 (epoch 45), train_loss = 1.298, time/batch = 0.151\n",
            "16192/17600 (epoch 46), train_loss = 1.089, time/batch = 0.166\n",
            "16193/17600 (epoch 46), train_loss = 1.331, time/batch = 0.148\n",
            "16194/17600 (epoch 46), train_loss = 1.322, time/batch = 0.164\n",
            "16195/17600 (epoch 46), train_loss = 1.268, time/batch = 0.151\n",
            "16196/17600 (epoch 46), train_loss = 1.365, time/batch = 0.168\n",
            "16197/17600 (epoch 46), train_loss = 1.340, time/batch = 0.150\n",
            "16198/17600 (epoch 46), train_loss = 1.284, time/batch = 0.182\n",
            "16199/17600 (epoch 46), train_loss = 1.270, time/batch = 0.160\n",
            "16200/17600 (epoch 46), train_loss = 1.312, time/batch = 0.165\n",
            "16201/17600 (epoch 46), train_loss = 1.364, time/batch = 0.158\n",
            "16202/17600 (epoch 46), train_loss = 1.293, time/batch = 0.161\n",
            "16203/17600 (epoch 46), train_loss = 1.306, time/batch = 0.152\n",
            "16204/17600 (epoch 46), train_loss = 1.290, time/batch = 0.177\n",
            "16205/17600 (epoch 46), train_loss = 1.276, time/batch = 0.159\n",
            "16206/17600 (epoch 46), train_loss = 1.256, time/batch = 0.175\n",
            "16207/17600 (epoch 46), train_loss = 1.321, time/batch = 0.167\n",
            "16208/17600 (epoch 46), train_loss = 1.334, time/batch = 0.167\n",
            "16209/17600 (epoch 46), train_loss = 1.287, time/batch = 0.148\n",
            "16210/17600 (epoch 46), train_loss = 1.300, time/batch = 0.185\n",
            "16211/17600 (epoch 46), train_loss = 1.343, time/batch = 0.159\n",
            "16212/17600 (epoch 46), train_loss = 1.267, time/batch = 0.158\n",
            "16213/17600 (epoch 46), train_loss = 1.323, time/batch = 0.153\n",
            "16214/17600 (epoch 46), train_loss = 1.288, time/batch = 0.171\n",
            "16215/17600 (epoch 46), train_loss = 1.260, time/batch = 0.158\n",
            "16216/17600 (epoch 46), train_loss = 1.344, time/batch = 0.160\n",
            "16217/17600 (epoch 46), train_loss = 1.278, time/batch = 0.172\n",
            "16218/17600 (epoch 46), train_loss = 1.323, time/batch = 0.166\n",
            "16219/17600 (epoch 46), train_loss = 1.307, time/batch = 0.169\n",
            "16220/17600 (epoch 46), train_loss = 1.286, time/batch = 0.168\n",
            "16221/17600 (epoch 46), train_loss = 1.220, time/batch = 0.145\n",
            "16222/17600 (epoch 46), train_loss = 1.328, time/batch = 0.158\n",
            "16223/17600 (epoch 46), train_loss = 1.306, time/batch = 0.174\n",
            "16224/17600 (epoch 46), train_loss = 1.282, time/batch = 0.163\n",
            "16225/17600 (epoch 46), train_loss = 1.284, time/batch = 0.159\n",
            "16226/17600 (epoch 46), train_loss = 1.289, time/batch = 0.173\n",
            "16227/17600 (epoch 46), train_loss = 1.303, time/batch = 0.164\n",
            "16228/17600 (epoch 46), train_loss = 1.332, time/batch = 0.227\n",
            "16229/17600 (epoch 46), train_loss = 1.328, time/batch = 0.301\n",
            "16230/17600 (epoch 46), train_loss = 1.306, time/batch = 0.300\n",
            "16231/17600 (epoch 46), train_loss = 1.285, time/batch = 0.283\n",
            "16232/17600 (epoch 46), train_loss = 1.329, time/batch = 0.263\n",
            "16233/17600 (epoch 46), train_loss = 1.267, time/batch = 0.284\n",
            "16234/17600 (epoch 46), train_loss = 1.322, time/batch = 0.296\n",
            "16235/17600 (epoch 46), train_loss = 1.244, time/batch = 0.278\n",
            "16236/17600 (epoch 46), train_loss = 1.310, time/batch = 0.296\n",
            "16237/17600 (epoch 46), train_loss = 1.283, time/batch = 0.298\n",
            "16238/17600 (epoch 46), train_loss = 1.238, time/batch = 0.279\n",
            "16239/17600 (epoch 46), train_loss = 1.298, time/batch = 0.164\n",
            "16240/17600 (epoch 46), train_loss = 1.249, time/batch = 0.163\n",
            "16241/17600 (epoch 46), train_loss = 1.264, time/batch = 0.150\n",
            "16242/17600 (epoch 46), train_loss = 1.290, time/batch = 0.151\n",
            "16243/17600 (epoch 46), train_loss = 1.279, time/batch = 0.166\n",
            "16244/17600 (epoch 46), train_loss = 1.252, time/batch = 0.155\n",
            "16245/17600 (epoch 46), train_loss = 1.279, time/batch = 0.165\n",
            "16246/17600 (epoch 46), train_loss = 1.323, time/batch = 0.156\n",
            "16247/17600 (epoch 46), train_loss = 1.301, time/batch = 0.170\n",
            "16248/17600 (epoch 46), train_loss = 1.300, time/batch = 0.158\n",
            "16249/17600 (epoch 46), train_loss = 1.295, time/batch = 0.171\n",
            "16250/17600 (epoch 46), train_loss = 1.274, time/batch = 0.156\n",
            "16251/17600 (epoch 46), train_loss = 1.302, time/batch = 0.179\n",
            "16252/17600 (epoch 46), train_loss = 1.298, time/batch = 0.153\n",
            "16253/17600 (epoch 46), train_loss = 1.239, time/batch = 0.177\n",
            "16254/17600 (epoch 46), train_loss = 1.240, time/batch = 0.158\n",
            "16255/17600 (epoch 46), train_loss = 1.261, time/batch = 0.161\n",
            "16256/17600 (epoch 46), train_loss = 1.292, time/batch = 0.181\n",
            "16257/17600 (epoch 46), train_loss = 1.257, time/batch = 0.160\n",
            "16258/17600 (epoch 46), train_loss = 1.294, time/batch = 0.152\n",
            "16259/17600 (epoch 46), train_loss = 1.255, time/batch = 0.167\n",
            "16260/17600 (epoch 46), train_loss = 1.303, time/batch = 0.163\n",
            "16261/17600 (epoch 46), train_loss = 1.234, time/batch = 0.162\n",
            "16262/17600 (epoch 46), train_loss = 1.278, time/batch = 0.152\n",
            "16263/17600 (epoch 46), train_loss = 1.325, time/batch = 0.162\n",
            "16264/17600 (epoch 46), train_loss = 1.286, time/batch = 0.145\n",
            "16265/17600 (epoch 46), train_loss = 1.355, time/batch = 0.170\n",
            "16266/17600 (epoch 46), train_loss = 1.287, time/batch = 0.180\n",
            "16267/17600 (epoch 46), train_loss = 1.288, time/batch = 0.168\n",
            "16268/17600 (epoch 46), train_loss = 1.244, time/batch = 0.155\n",
            "16269/17600 (epoch 46), train_loss = 1.300, time/batch = 0.161\n",
            "16270/17600 (epoch 46), train_loss = 1.305, time/batch = 0.159\n",
            "16271/17600 (epoch 46), train_loss = 1.382, time/batch = 0.165\n",
            "16272/17600 (epoch 46), train_loss = 1.351, time/batch = 0.172\n",
            "16273/17600 (epoch 46), train_loss = 1.344, time/batch = 0.161\n",
            "16274/17600 (epoch 46), train_loss = 1.324, time/batch = 0.161\n",
            "16275/17600 (epoch 46), train_loss = 1.298, time/batch = 0.160\n",
            "16276/17600 (epoch 46), train_loss = 1.273, time/batch = 0.150\n",
            "16277/17600 (epoch 46), train_loss = 1.307, time/batch = 0.169\n",
            "16278/17600 (epoch 46), train_loss = 1.300, time/batch = 0.157\n",
            "16279/17600 (epoch 46), train_loss = 1.238, time/batch = 0.171\n",
            "16280/17600 (epoch 46), train_loss = 1.279, time/batch = 0.174\n",
            "16281/17600 (epoch 46), train_loss = 1.219, time/batch = 0.169\n",
            "16282/17600 (epoch 46), train_loss = 1.282, time/batch = 0.145\n",
            "16283/17600 (epoch 46), train_loss = 1.269, time/batch = 0.150\n",
            "16284/17600 (epoch 46), train_loss = 1.247, time/batch = 0.152\n",
            "16285/17600 (epoch 46), train_loss = 1.301, time/batch = 0.180\n",
            "16286/17600 (epoch 46), train_loss = 1.285, time/batch = 0.164\n",
            "16287/17600 (epoch 46), train_loss = 1.247, time/batch = 0.152\n",
            "16288/17600 (epoch 46), train_loss = 1.221, time/batch = 0.159\n",
            "16289/17600 (epoch 46), train_loss = 1.274, time/batch = 0.150\n",
            "16290/17600 (epoch 46), train_loss = 1.309, time/batch = 0.158\n",
            "16291/17600 (epoch 46), train_loss = 1.268, time/batch = 0.176\n",
            "16292/17600 (epoch 46), train_loss = 1.242, time/batch = 0.168\n",
            "16293/17600 (epoch 46), train_loss = 1.326, time/batch = 0.153\n",
            "16294/17600 (epoch 46), train_loss = 1.265, time/batch = 0.162\n",
            "16295/17600 (epoch 46), train_loss = 1.372, time/batch = 0.163\n",
            "16296/17600 (epoch 46), train_loss = 1.302, time/batch = 0.163\n",
            "16297/17600 (epoch 46), train_loss = 1.277, time/batch = 0.176\n",
            "16298/17600 (epoch 46), train_loss = 1.317, time/batch = 0.173\n",
            "16299/17600 (epoch 46), train_loss = 1.304, time/batch = 0.176\n",
            "16300/17600 (epoch 46), train_loss = 1.305, time/batch = 0.273\n",
            "16301/17600 (epoch 46), train_loss = 1.289, time/batch = 0.281\n",
            "16302/17600 (epoch 46), train_loss = 1.246, time/batch = 0.310\n",
            "16303/17600 (epoch 46), train_loss = 1.268, time/batch = 0.278\n",
            "16304/17600 (epoch 46), train_loss = 1.300, time/batch = 0.270\n",
            "16305/17600 (epoch 46), train_loss = 1.196, time/batch = 0.270\n",
            "16306/17600 (epoch 46), train_loss = 1.271, time/batch = 0.259\n",
            "16307/17600 (epoch 46), train_loss = 1.299, time/batch = 0.266\n",
            "16308/17600 (epoch 46), train_loss = 1.265, time/batch = 0.265\n",
            "16309/17600 (epoch 46), train_loss = 1.294, time/batch = 0.296\n",
            "16310/17600 (epoch 46), train_loss = 1.235, time/batch = 0.196\n",
            "16311/17600 (epoch 46), train_loss = 1.341, time/batch = 0.166\n",
            "16312/17600 (epoch 46), train_loss = 1.247, time/batch = 0.154\n",
            "16313/17600 (epoch 46), train_loss = 1.308, time/batch = 0.161\n",
            "16314/17600 (epoch 46), train_loss = 1.333, time/batch = 0.162\n",
            "16315/17600 (epoch 46), train_loss = 1.317, time/batch = 0.165\n",
            "16316/17600 (epoch 46), train_loss = 1.234, time/batch = 0.170\n",
            "16317/17600 (epoch 46), train_loss = 1.300, time/batch = 0.151\n",
            "16318/17600 (epoch 46), train_loss = 1.297, time/batch = 0.160\n",
            "16319/17600 (epoch 46), train_loss = 1.284, time/batch = 0.165\n",
            "16320/17600 (epoch 46), train_loss = 1.267, time/batch = 0.171\n",
            "16321/17600 (epoch 46), train_loss = 1.299, time/batch = 0.163\n",
            "16322/17600 (epoch 46), train_loss = 1.311, time/batch = 0.177\n",
            "16323/17600 (epoch 46), train_loss = 1.317, time/batch = 0.157\n",
            "16324/17600 (epoch 46), train_loss = 1.287, time/batch = 0.157\n",
            "16325/17600 (epoch 46), train_loss = 1.257, time/batch = 0.169\n",
            "16326/17600 (epoch 46), train_loss = 1.270, time/batch = 0.153\n",
            "16327/17600 (epoch 46), train_loss = 1.240, time/batch = 0.158\n",
            "16328/17600 (epoch 46), train_loss = 1.306, time/batch = 0.166\n",
            "16329/17600 (epoch 46), train_loss = 1.334, time/batch = 0.171\n",
            "16330/17600 (epoch 46), train_loss = 1.279, time/batch = 0.156\n",
            "16331/17600 (epoch 46), train_loss = 1.332, time/batch = 0.147\n",
            "16332/17600 (epoch 46), train_loss = 1.357, time/batch = 0.166\n",
            "16333/17600 (epoch 46), train_loss = 1.270, time/batch = 0.167\n",
            "16334/17600 (epoch 46), train_loss = 1.295, time/batch = 0.177\n",
            "16335/17600 (epoch 46), train_loss = 1.262, time/batch = 0.176\n",
            "16336/17600 (epoch 46), train_loss = 1.312, time/batch = 0.164\n",
            "16337/17600 (epoch 46), train_loss = 1.298, time/batch = 0.157\n",
            "16338/17600 (epoch 46), train_loss = 1.273, time/batch = 0.157\n",
            "16339/17600 (epoch 46), train_loss = 1.281, time/batch = 0.157\n",
            "16340/17600 (epoch 46), train_loss = 1.262, time/batch = 0.180\n",
            "16341/17600 (epoch 46), train_loss = 1.325, time/batch = 0.176\n",
            "16342/17600 (epoch 46), train_loss = 1.212, time/batch = 0.154\n",
            "16343/17600 (epoch 46), train_loss = 1.287, time/batch = 0.160\n",
            "16344/17600 (epoch 46), train_loss = 1.312, time/batch = 0.147\n",
            "16345/17600 (epoch 46), train_loss = 1.281, time/batch = 0.174\n",
            "16346/17600 (epoch 46), train_loss = 1.296, time/batch = 0.162\n",
            "16347/17600 (epoch 46), train_loss = 1.236, time/batch = 0.163\n",
            "16348/17600 (epoch 46), train_loss = 1.233, time/batch = 0.154\n",
            "16349/17600 (epoch 46), train_loss = 1.251, time/batch = 0.157\n",
            "16350/17600 (epoch 46), train_loss = 1.311, time/batch = 0.159\n",
            "16351/17600 (epoch 46), train_loss = 1.354, time/batch = 0.174\n",
            "16352/17600 (epoch 46), train_loss = 1.310, time/batch = 0.165\n",
            "16353/17600 (epoch 46), train_loss = 1.309, time/batch = 0.176\n",
            "16354/17600 (epoch 46), train_loss = 1.240, time/batch = 0.161\n",
            "16355/17600 (epoch 46), train_loss = 1.267, time/batch = 0.158\n",
            "16356/17600 (epoch 46), train_loss = 1.291, time/batch = 0.157\n",
            "16357/17600 (epoch 46), train_loss = 1.276, time/batch = 0.174\n",
            "16358/17600 (epoch 46), train_loss = 1.290, time/batch = 0.159\n",
            "16359/17600 (epoch 46), train_loss = 1.315, time/batch = 0.169\n",
            "16360/17600 (epoch 46), train_loss = 1.308, time/batch = 0.178\n",
            "16361/17600 (epoch 46), train_loss = 1.289, time/batch = 0.162\n",
            "16362/17600 (epoch 46), train_loss = 1.221, time/batch = 0.156\n",
            "16363/17600 (epoch 46), train_loss = 1.233, time/batch = 0.170\n",
            "16364/17600 (epoch 46), train_loss = 1.323, time/batch = 0.160\n",
            "16365/17600 (epoch 46), train_loss = 1.292, time/batch = 0.167\n",
            "16366/17600 (epoch 46), train_loss = 1.282, time/batch = 0.160\n",
            "16367/17600 (epoch 46), train_loss = 1.285, time/batch = 0.160\n",
            "16368/17600 (epoch 46), train_loss = 1.199, time/batch = 0.142\n",
            "16369/17600 (epoch 46), train_loss = 1.261, time/batch = 0.156\n",
            "16370/17600 (epoch 46), train_loss = 1.261, time/batch = 0.149\n",
            "16371/17600 (epoch 46), train_loss = 1.242, time/batch = 0.214\n",
            "16372/17600 (epoch 46), train_loss = 1.240, time/batch = 0.300\n",
            "16373/17600 (epoch 46), train_loss = 1.294, time/batch = 0.295\n",
            "16374/17600 (epoch 46), train_loss = 1.351, time/batch = 0.287\n",
            "16375/17600 (epoch 46), train_loss = 1.266, time/batch = 0.265\n",
            "16376/17600 (epoch 46), train_loss = 1.267, time/batch = 0.291\n",
            "16377/17600 (epoch 46), train_loss = 1.231, time/batch = 0.295\n",
            "16378/17600 (epoch 46), train_loss = 1.304, time/batch = 0.280\n",
            "16379/17600 (epoch 46), train_loss = 1.315, time/batch = 0.302\n",
            "16380/17600 (epoch 46), train_loss = 1.235, time/batch = 0.301\n",
            "16381/17600 (epoch 46), train_loss = 1.328, time/batch = 0.207\n",
            "16382/17600 (epoch 46), train_loss = 1.297, time/batch = 0.166\n",
            "16383/17600 (epoch 46), train_loss = 1.276, time/batch = 0.156\n",
            "16384/17600 (epoch 46), train_loss = 1.284, time/batch = 0.179\n",
            "16385/17600 (epoch 46), train_loss = 1.303, time/batch = 0.151\n",
            "16386/17600 (epoch 46), train_loss = 1.303, time/batch = 0.167\n",
            "16387/17600 (epoch 46), train_loss = 1.312, time/batch = 0.164\n",
            "16388/17600 (epoch 46), train_loss = 1.267, time/batch = 0.170\n",
            "16389/17600 (epoch 46), train_loss = 1.320, time/batch = 0.162\n",
            "16390/17600 (epoch 46), train_loss = 1.267, time/batch = 0.160\n",
            "16391/17600 (epoch 46), train_loss = 1.302, time/batch = 0.170\n",
            "16392/17600 (epoch 46), train_loss = 1.304, time/batch = 0.160\n",
            "16393/17600 (epoch 46), train_loss = 1.271, time/batch = 0.160\n",
            "16394/17600 (epoch 46), train_loss = 1.261, time/batch = 0.158\n",
            "16395/17600 (epoch 46), train_loss = 1.289, time/batch = 0.157\n",
            "16396/17600 (epoch 46), train_loss = 1.323, time/batch = 0.148\n",
            "16397/17600 (epoch 46), train_loss = 1.370, time/batch = 0.167\n",
            "16398/17600 (epoch 46), train_loss = 1.286, time/batch = 0.144\n",
            "16399/17600 (epoch 46), train_loss = 1.286, time/batch = 0.160\n",
            "16400/17600 (epoch 46), train_loss = 1.290, time/batch = 0.167\n",
            "16401/17600 (epoch 46), train_loss = 1.268, time/batch = 0.156\n",
            "16402/17600 (epoch 46), train_loss = 1.313, time/batch = 0.154\n",
            "16403/17600 (epoch 46), train_loss = 1.305, time/batch = 0.173\n",
            "16404/17600 (epoch 46), train_loss = 1.274, time/batch = 0.185\n",
            "16405/17600 (epoch 46), train_loss = 1.348, time/batch = 0.166\n",
            "16406/17600 (epoch 46), train_loss = 1.386, time/batch = 0.159\n",
            "16407/17600 (epoch 46), train_loss = 1.275, time/batch = 0.168\n",
            "16408/17600 (epoch 46), train_loss = 1.293, time/batch = 0.146\n",
            "16409/17600 (epoch 46), train_loss = 1.252, time/batch = 0.168\n",
            "16410/17600 (epoch 46), train_loss = 1.284, time/batch = 0.177\n",
            "16411/17600 (epoch 46), train_loss = 1.266, time/batch = 0.167\n",
            "16412/17600 (epoch 46), train_loss = 1.261, time/batch = 0.172\n",
            "16413/17600 (epoch 46), train_loss = 1.284, time/batch = 0.161\n",
            "16414/17600 (epoch 46), train_loss = 1.354, time/batch = 0.164\n",
            "16415/17600 (epoch 46), train_loss = 1.327, time/batch = 0.168\n",
            "16416/17600 (epoch 46), train_loss = 1.244, time/batch = 0.169\n",
            "16417/17600 (epoch 46), train_loss = 1.290, time/batch = 0.175\n",
            "16418/17600 (epoch 46), train_loss = 1.296, time/batch = 0.168\n",
            "16419/17600 (epoch 46), train_loss = 1.274, time/batch = 0.165\n",
            "16420/17600 (epoch 46), train_loss = 1.306, time/batch = 0.161\n",
            "16421/17600 (epoch 46), train_loss = 1.271, time/batch = 0.160\n",
            "16422/17600 (epoch 46), train_loss = 1.260, time/batch = 0.172\n",
            "16423/17600 (epoch 46), train_loss = 1.324, time/batch = 0.166\n",
            "16424/17600 (epoch 46), train_loss = 1.322, time/batch = 0.180\n",
            "16425/17600 (epoch 46), train_loss = 1.336, time/batch = 0.162\n",
            "16426/17600 (epoch 46), train_loss = 1.253, time/batch = 0.145\n",
            "16427/17600 (epoch 46), train_loss = 1.283, time/batch = 0.152\n",
            "16428/17600 (epoch 46), train_loss = 1.274, time/batch = 0.166\n",
            "16429/17600 (epoch 46), train_loss = 1.343, time/batch = 0.167\n",
            "16430/17600 (epoch 46), train_loss = 1.340, time/batch = 0.150\n",
            "16431/17600 (epoch 46), train_loss = 1.234, time/batch = 0.167\n",
            "16432/17600 (epoch 46), train_loss = 1.326, time/batch = 0.152\n",
            "16433/17600 (epoch 46), train_loss = 1.274, time/batch = 0.170\n",
            "16434/17600 (epoch 46), train_loss = 1.311, time/batch = 0.165\n",
            "16435/17600 (epoch 46), train_loss = 1.303, time/batch = 0.186\n",
            "16436/17600 (epoch 46), train_loss = 1.270, time/batch = 0.163\n",
            "16437/17600 (epoch 46), train_loss = 1.310, time/batch = 0.162\n",
            "16438/17600 (epoch 46), train_loss = 1.303, time/batch = 0.161\n",
            "16439/17600 (epoch 46), train_loss = 1.275, time/batch = 0.168\n",
            "16440/17600 (epoch 46), train_loss = 1.254, time/batch = 0.146\n",
            "16441/17600 (epoch 46), train_loss = 1.232, time/batch = 0.181\n",
            "16442/17600 (epoch 46), train_loss = 1.365, time/batch = 0.250\n",
            "16443/17600 (epoch 46), train_loss = 1.303, time/batch = 0.257\n",
            "16444/17600 (epoch 46), train_loss = 1.296, time/batch = 0.274\n",
            "16445/17600 (epoch 46), train_loss = 1.304, time/batch = 0.297\n",
            "16446/17600 (epoch 46), train_loss = 1.322, time/batch = 0.276\n",
            "16447/17600 (epoch 46), train_loss = 1.330, time/batch = 0.276\n",
            "16448/17600 (epoch 46), train_loss = 1.375, time/batch = 0.266\n",
            "16449/17600 (epoch 46), train_loss = 1.346, time/batch = 0.294\n",
            "16450/17600 (epoch 46), train_loss = 1.298, time/batch = 0.279\n",
            "16451/17600 (epoch 46), train_loss = 1.300, time/batch = 0.291\n",
            "16452/17600 (epoch 46), train_loss = 1.275, time/batch = 0.224\n",
            "16453/17600 (epoch 46), train_loss = 1.285, time/batch = 0.188\n",
            "16454/17600 (epoch 46), train_loss = 1.264, time/batch = 0.164\n",
            "16455/17600 (epoch 46), train_loss = 1.222, time/batch = 0.159\n",
            "16456/17600 (epoch 46), train_loss = 1.273, time/batch = 0.149\n",
            "16457/17600 (epoch 46), train_loss = 1.280, time/batch = 0.162\n",
            "16458/17600 (epoch 46), train_loss = 1.234, time/batch = 0.149\n",
            "16459/17600 (epoch 46), train_loss = 1.237, time/batch = 0.183\n",
            "16460/17600 (epoch 46), train_loss = 1.268, time/batch = 0.149\n",
            "16461/17600 (epoch 46), train_loss = 1.302, time/batch = 0.161\n",
            "16462/17600 (epoch 46), train_loss = 1.225, time/batch = 0.155\n",
            "16463/17600 (epoch 46), train_loss = 1.295, time/batch = 0.165\n",
            "16464/17600 (epoch 46), train_loss = 1.262, time/batch = 0.145\n",
            "16465/17600 (epoch 46), train_loss = 1.290, time/batch = 0.162\n",
            "16466/17600 (epoch 46), train_loss = 1.272, time/batch = 0.169\n",
            "16467/17600 (epoch 46), train_loss = 1.326, time/batch = 0.168\n",
            "16468/17600 (epoch 46), train_loss = 1.291, time/batch = 0.155\n",
            "16469/17600 (epoch 46), train_loss = 1.316, time/batch = 0.165\n",
            "16470/17600 (epoch 46), train_loss = 1.281, time/batch = 0.157\n",
            "16471/17600 (epoch 46), train_loss = 1.288, time/batch = 0.174\n",
            "16472/17600 (epoch 46), train_loss = 1.311, time/batch = 0.175\n",
            "16473/17600 (epoch 46), train_loss = 1.359, time/batch = 0.165\n",
            "16474/17600 (epoch 46), train_loss = 1.331, time/batch = 0.158\n",
            "16475/17600 (epoch 46), train_loss = 1.329, time/batch = 0.160\n",
            "16476/17600 (epoch 46), train_loss = 1.345, time/batch = 0.149\n",
            "16477/17600 (epoch 46), train_loss = 1.358, time/batch = 0.170\n",
            "16478/17600 (epoch 46), train_loss = 1.305, time/batch = 0.171\n",
            "16479/17600 (epoch 46), train_loss = 1.352, time/batch = 0.163\n",
            "16480/17600 (epoch 46), train_loss = 1.289, time/batch = 0.151\n",
            "16481/17600 (epoch 46), train_loss = 1.284, time/batch = 0.150\n",
            "16482/17600 (epoch 46), train_loss = 1.349, time/batch = 0.152\n",
            "16483/17600 (epoch 46), train_loss = 1.254, time/batch = 0.172\n",
            "16484/17600 (epoch 46), train_loss = 1.250, time/batch = 0.153\n",
            "16485/17600 (epoch 46), train_loss = 1.325, time/batch = 0.188\n",
            "16486/17600 (epoch 46), train_loss = 1.272, time/batch = 0.156\n",
            "16487/17600 (epoch 46), train_loss = 1.332, time/batch = 0.171\n",
            "16488/17600 (epoch 46), train_loss = 1.278, time/batch = 0.154\n",
            "16489/17600 (epoch 46), train_loss = 1.305, time/batch = 0.162\n",
            "16490/17600 (epoch 46), train_loss = 1.292, time/batch = 0.165\n",
            "16491/17600 (epoch 46), train_loss = 1.304, time/batch = 0.185\n",
            "16492/17600 (epoch 46), train_loss = 1.334, time/batch = 0.147\n",
            "16493/17600 (epoch 46), train_loss = 1.294, time/batch = 0.168\n",
            "16494/17600 (epoch 46), train_loss = 1.362, time/batch = 0.158\n",
            "16495/17600 (epoch 46), train_loss = 1.353, time/batch = 0.160\n",
            "16496/17600 (epoch 46), train_loss = 1.364, time/batch = 0.164\n",
            "16497/17600 (epoch 46), train_loss = 1.280, time/batch = 0.174\n",
            "16498/17600 (epoch 46), train_loss = 1.313, time/batch = 0.150\n",
            "16499/17600 (epoch 46), train_loss = 1.321, time/batch = 0.164\n",
            "16500/17600 (epoch 46), train_loss = 1.317, time/batch = 0.146\n",
            "16501/17600 (epoch 46), train_loss = 1.320, time/batch = 0.154\n",
            "16502/17600 (epoch 46), train_loss = 1.357, time/batch = 0.165\n",
            "16503/17600 (epoch 46), train_loss = 1.308, time/batch = 0.177\n",
            "16504/17600 (epoch 46), train_loss = 1.341, time/batch = 0.162\n",
            "16505/17600 (epoch 46), train_loss = 1.276, time/batch = 0.166\n",
            "16506/17600 (epoch 46), train_loss = 1.290, time/batch = 0.158\n",
            "16507/17600 (epoch 46), train_loss = 1.261, time/batch = 0.166\n",
            "16508/17600 (epoch 46), train_loss = 1.355, time/batch = 0.152\n",
            "16509/17600 (epoch 46), train_loss = 1.343, time/batch = 0.172\n",
            "16510/17600 (epoch 46), train_loss = 1.341, time/batch = 0.167\n",
            "16511/17600 (epoch 46), train_loss = 1.324, time/batch = 0.156\n",
            "16512/17600 (epoch 46), train_loss = 1.363, time/batch = 0.161\n",
            "16513/17600 (epoch 46), train_loss = 1.328, time/batch = 0.185\n",
            "16514/17600 (epoch 46), train_loss = 1.386, time/batch = 0.272\n",
            "16515/17600 (epoch 46), train_loss = 1.355, time/batch = 0.276\n",
            "16516/17600 (epoch 46), train_loss = 1.342, time/batch = 0.293\n",
            "16517/17600 (epoch 46), train_loss = 1.290, time/batch = 0.283\n",
            "16518/17600 (epoch 46), train_loss = 1.368, time/batch = 0.271\n",
            "16519/17600 (epoch 46), train_loss = 1.344, time/batch = 0.296\n",
            "16520/17600 (epoch 46), train_loss = 1.382, time/batch = 0.270\n",
            "16521/17600 (epoch 46), train_loss = 1.351, time/batch = 0.253\n",
            "16522/17600 (epoch 46), train_loss = 1.270, time/batch = 0.287\n",
            "16523/17600 (epoch 46), train_loss = 1.300, time/batch = 0.294\n",
            "16524/17600 (epoch 46), train_loss = 1.346, time/batch = 0.178\n",
            "16525/17600 (epoch 46), train_loss = 1.359, time/batch = 0.170\n",
            "16526/17600 (epoch 46), train_loss = 1.315, time/batch = 0.157\n",
            "16527/17600 (epoch 46), train_loss = 1.308, time/batch = 0.156\n",
            "16528/17600 (epoch 46), train_loss = 1.313, time/batch = 0.164\n",
            "16529/17600 (epoch 46), train_loss = 1.354, time/batch = 0.151\n",
            "16530/17600 (epoch 46), train_loss = 1.300, time/batch = 0.153\n",
            "16531/17600 (epoch 46), train_loss = 1.328, time/batch = 0.165\n",
            "16532/17600 (epoch 46), train_loss = 1.297, time/batch = 0.157\n",
            "16533/17600 (epoch 46), train_loss = 1.341, time/batch = 0.163\n",
            "16534/17600 (epoch 46), train_loss = 1.347, time/batch = 0.162\n",
            "16535/17600 (epoch 46), train_loss = 1.310, time/batch = 0.169\n",
            "16536/17600 (epoch 46), train_loss = 1.288, time/batch = 0.155\n",
            "16537/17600 (epoch 46), train_loss = 1.308, time/batch = 0.166\n",
            "16538/17600 (epoch 46), train_loss = 1.331, time/batch = 0.153\n",
            "16539/17600 (epoch 46), train_loss = 1.305, time/batch = 0.153\n",
            "16540/17600 (epoch 46), train_loss = 1.328, time/batch = 0.162\n",
            "16541/17600 (epoch 46), train_loss = 1.314, time/batch = 0.178\n",
            "16542/17600 (epoch 46), train_loss = 1.304, time/batch = 0.151\n",
            "16543/17600 (epoch 46), train_loss = 1.296, time/batch = 0.167\n",
            "16544/17600 (epoch 47), train_loss = 1.087, time/batch = 0.160\n",
            "16545/17600 (epoch 47), train_loss = 1.327, time/batch = 0.176\n",
            "16546/17600 (epoch 47), train_loss = 1.320, time/batch = 0.160\n",
            "16547/17600 (epoch 47), train_loss = 1.267, time/batch = 0.155\n",
            "16548/17600 (epoch 47), train_loss = 1.364, time/batch = 0.165\n",
            "16549/17600 (epoch 47), train_loss = 1.339, time/batch = 0.160\n",
            "16550/17600 (epoch 47), train_loss = 1.282, time/batch = 0.164\n",
            "16551/17600 (epoch 47), train_loss = 1.268, time/batch = 0.151\n",
            "16552/17600 (epoch 47), train_loss = 1.310, time/batch = 0.182\n",
            "16553/17600 (epoch 47), train_loss = 1.362, time/batch = 0.150\n",
            "16554/17600 (epoch 47), train_loss = 1.291, time/batch = 0.167\n",
            "16555/17600 (epoch 47), train_loss = 1.304, time/batch = 0.160\n",
            "16556/17600 (epoch 47), train_loss = 1.288, time/batch = 0.166\n",
            "16557/17600 (epoch 47), train_loss = 1.275, time/batch = 0.157\n",
            "16558/17600 (epoch 47), train_loss = 1.255, time/batch = 0.174\n",
            "16559/17600 (epoch 47), train_loss = 1.320, time/batch = 0.151\n",
            "16560/17600 (epoch 47), train_loss = 1.332, time/batch = 0.168\n",
            "16561/17600 (epoch 47), train_loss = 1.285, time/batch = 0.156\n",
            "16562/17600 (epoch 47), train_loss = 1.298, time/batch = 0.170\n",
            "16563/17600 (epoch 47), train_loss = 1.341, time/batch = 0.160\n",
            "16564/17600 (epoch 47), train_loss = 1.266, time/batch = 0.184\n",
            "16565/17600 (epoch 47), train_loss = 1.321, time/batch = 0.168\n",
            "16566/17600 (epoch 47), train_loss = 1.287, time/batch = 0.159\n",
            "16567/17600 (epoch 47), train_loss = 1.258, time/batch = 0.168\n",
            "16568/17600 (epoch 47), train_loss = 1.342, time/batch = 0.163\n",
            "16569/17600 (epoch 47), train_loss = 1.276, time/batch = 0.161\n",
            "16570/17600 (epoch 47), train_loss = 1.322, time/batch = 0.168\n",
            "16571/17600 (epoch 47), train_loss = 1.305, time/batch = 0.207\n",
            "16572/17600 (epoch 47), train_loss = 1.284, time/batch = 0.168\n",
            "16573/17600 (epoch 47), train_loss = 1.219, time/batch = 0.181\n",
            "16574/17600 (epoch 47), train_loss = 1.327, time/batch = 0.161\n",
            "16575/17600 (epoch 47), train_loss = 1.304, time/batch = 0.160\n",
            "16576/17600 (epoch 47), train_loss = 1.280, time/batch = 0.172\n",
            "16577/17600 (epoch 47), train_loss = 1.282, time/batch = 0.159\n",
            "16578/17600 (epoch 47), train_loss = 1.288, time/batch = 0.161\n",
            "16579/17600 (epoch 47), train_loss = 1.302, time/batch = 0.155\n",
            "16580/17600 (epoch 47), train_loss = 1.330, time/batch = 0.170\n",
            "16581/17600 (epoch 47), train_loss = 1.326, time/batch = 0.166\n",
            "16582/17600 (epoch 47), train_loss = 1.304, time/batch = 0.154\n",
            "16583/17600 (epoch 47), train_loss = 1.283, time/batch = 0.243\n",
            "16584/17600 (epoch 47), train_loss = 1.327, time/batch = 0.256\n",
            "16585/17600 (epoch 47), train_loss = 1.265, time/batch = 0.263\n",
            "16586/17600 (epoch 47), train_loss = 1.321, time/batch = 0.302\n",
            "16587/17600 (epoch 47), train_loss = 1.242, time/batch = 0.271\n",
            "16588/17600 (epoch 47), train_loss = 1.308, time/batch = 0.272\n",
            "16589/17600 (epoch 47), train_loss = 1.282, time/batch = 0.264\n",
            "16590/17600 (epoch 47), train_loss = 1.236, time/batch = 0.258\n",
            "16591/17600 (epoch 47), train_loss = 1.296, time/batch = 0.283\n",
            "16592/17600 (epoch 47), train_loss = 1.247, time/batch = 0.258\n",
            "16593/17600 (epoch 47), train_loss = 1.263, time/batch = 0.302\n",
            "16594/17600 (epoch 47), train_loss = 1.288, time/batch = 0.169\n",
            "16595/17600 (epoch 47), train_loss = 1.278, time/batch = 0.175\n",
            "16596/17600 (epoch 47), train_loss = 1.251, time/batch = 0.151\n",
            "16597/17600 (epoch 47), train_loss = 1.277, time/batch = 0.170\n",
            "16598/17600 (epoch 47), train_loss = 1.321, time/batch = 0.155\n",
            "16599/17600 (epoch 47), train_loss = 1.299, time/batch = 0.168\n",
            "16600/17600 (epoch 47), train_loss = 1.298, time/batch = 0.160\n",
            "16601/17600 (epoch 47), train_loss = 1.293, time/batch = 0.170\n",
            "16602/17600 (epoch 47), train_loss = 1.272, time/batch = 0.165\n",
            "16603/17600 (epoch 47), train_loss = 1.300, time/batch = 0.158\n",
            "16604/17600 (epoch 47), train_loss = 1.297, time/batch = 0.162\n",
            "16605/17600 (epoch 47), train_loss = 1.237, time/batch = 0.164\n",
            "16606/17600 (epoch 47), train_loss = 1.239, time/batch = 0.162\n",
            "16607/17600 (epoch 47), train_loss = 1.259, time/batch = 0.170\n",
            "16608/17600 (epoch 47), train_loss = 1.290, time/batch = 0.161\n",
            "16609/17600 (epoch 47), train_loss = 1.256, time/batch = 0.147\n",
            "16610/17600 (epoch 47), train_loss = 1.292, time/batch = 0.170\n",
            "16611/17600 (epoch 47), train_loss = 1.253, time/batch = 0.148\n",
            "16612/17600 (epoch 47), train_loss = 1.301, time/batch = 0.165\n",
            "16613/17600 (epoch 47), train_loss = 1.232, time/batch = 0.152\n",
            "16614/17600 (epoch 47), train_loss = 1.276, time/batch = 0.167\n",
            "16615/17600 (epoch 47), train_loss = 1.323, time/batch = 0.167\n",
            "16616/17600 (epoch 47), train_loss = 1.285, time/batch = 0.177\n",
            "16617/17600 (epoch 47), train_loss = 1.353, time/batch = 0.156\n",
            "16618/17600 (epoch 47), train_loss = 1.285, time/batch = 0.160\n",
            "16619/17600 (epoch 47), train_loss = 1.286, time/batch = 0.152\n",
            "16620/17600 (epoch 47), train_loss = 1.243, time/batch = 0.173\n",
            "16621/17600 (epoch 47), train_loss = 1.299, time/batch = 0.160\n",
            "16622/17600 (epoch 47), train_loss = 1.303, time/batch = 0.150\n",
            "16623/17600 (epoch 47), train_loss = 1.380, time/batch = 0.152\n",
            "16624/17600 (epoch 47), train_loss = 1.349, time/batch = 0.173\n",
            "16625/17600 (epoch 47), train_loss = 1.342, time/batch = 0.167\n",
            "16626/17600 (epoch 47), train_loss = 1.322, time/batch = 0.170\n",
            "16627/17600 (epoch 47), train_loss = 1.296, time/batch = 0.167\n",
            "16628/17600 (epoch 47), train_loss = 1.271, time/batch = 0.173\n",
            "16629/17600 (epoch 47), train_loss = 1.305, time/batch = 0.154\n",
            "16630/17600 (epoch 47), train_loss = 1.298, time/batch = 0.179\n",
            "16631/17600 (epoch 47), train_loss = 1.237, time/batch = 0.160\n",
            "16632/17600 (epoch 47), train_loss = 1.277, time/batch = 0.177\n",
            "16633/17600 (epoch 47), train_loss = 1.217, time/batch = 0.162\n",
            "16634/17600 (epoch 47), train_loss = 1.280, time/batch = 0.160\n",
            "16635/17600 (epoch 47), train_loss = 1.267, time/batch = 0.147\n",
            "16636/17600 (epoch 47), train_loss = 1.245, time/batch = 0.154\n",
            "16637/17600 (epoch 47), train_loss = 1.300, time/batch = 0.153\n",
            "16638/17600 (epoch 47), train_loss = 1.284, time/batch = 0.152\n",
            "16639/17600 (epoch 47), train_loss = 1.245, time/batch = 0.163\n",
            "16640/17600 (epoch 47), train_loss = 1.219, time/batch = 0.157\n",
            "16641/17600 (epoch 47), train_loss = 1.273, time/batch = 0.168\n",
            "16642/17600 (epoch 47), train_loss = 1.307, time/batch = 0.149\n",
            "16643/17600 (epoch 47), train_loss = 1.266, time/batch = 0.162\n",
            "16644/17600 (epoch 47), train_loss = 1.241, time/batch = 0.162\n",
            "16645/17600 (epoch 47), train_loss = 1.324, time/batch = 0.172\n",
            "16646/17600 (epoch 47), train_loss = 1.263, time/batch = 0.167\n",
            "16647/17600 (epoch 47), train_loss = 1.370, time/batch = 0.142\n",
            "16648/17600 (epoch 47), train_loss = 1.301, time/batch = 0.157\n",
            "16649/17600 (epoch 47), train_loss = 1.276, time/batch = 0.158\n",
            "16650/17600 (epoch 47), train_loss = 1.316, time/batch = 0.155\n",
            "16651/17600 (epoch 47), train_loss = 1.303, time/batch = 0.155\n",
            "16652/17600 (epoch 47), train_loss = 1.303, time/batch = 0.176\n",
            "16653/17600 (epoch 47), train_loss = 1.288, time/batch = 0.153\n",
            "16654/17600 (epoch 47), train_loss = 1.244, time/batch = 0.171\n",
            "16655/17600 (epoch 47), train_loss = 1.266, time/batch = 0.204\n",
            "16656/17600 (epoch 47), train_loss = 1.298, time/batch = 0.266\n",
            "16657/17600 (epoch 47), train_loss = 1.195, time/batch = 0.280\n",
            "16658/17600 (epoch 47), train_loss = 1.269, time/batch = 0.287\n",
            "16659/17600 (epoch 47), train_loss = 1.298, time/batch = 0.269\n",
            "16660/17600 (epoch 47), train_loss = 1.263, time/batch = 0.279\n",
            "16661/17600 (epoch 47), train_loss = 1.292, time/batch = 0.277\n",
            "16662/17600 (epoch 47), train_loss = 1.233, time/batch = 0.272\n",
            "16663/17600 (epoch 47), train_loss = 1.339, time/batch = 0.282\n",
            "16664/17600 (epoch 47), train_loss = 1.245, time/batch = 0.308\n",
            "16665/17600 (epoch 47), train_loss = 1.307, time/batch = 0.263\n",
            "16666/17600 (epoch 47), train_loss = 1.331, time/batch = 0.155\n",
            "16667/17600 (epoch 47), train_loss = 1.315, time/batch = 0.169\n",
            "16668/17600 (epoch 47), train_loss = 1.232, time/batch = 0.159\n",
            "16669/17600 (epoch 47), train_loss = 1.299, time/batch = 0.171\n",
            "16670/17600 (epoch 47), train_loss = 1.296, time/batch = 0.168\n",
            "16671/17600 (epoch 47), train_loss = 1.283, time/batch = 0.172\n",
            "16672/17600 (epoch 47), train_loss = 1.265, time/batch = 0.151\n",
            "16673/17600 (epoch 47), train_loss = 1.297, time/batch = 0.167\n",
            "16674/17600 (epoch 47), train_loss = 1.309, time/batch = 0.150\n",
            "16675/17600 (epoch 47), train_loss = 1.315, time/batch = 0.176\n",
            "16676/17600 (epoch 47), train_loss = 1.286, time/batch = 0.171\n",
            "16677/17600 (epoch 47), train_loss = 1.255, time/batch = 0.159\n",
            "16678/17600 (epoch 47), train_loss = 1.268, time/batch = 0.174\n",
            "16679/17600 (epoch 47), train_loss = 1.238, time/batch = 0.168\n",
            "16680/17600 (epoch 47), train_loss = 1.304, time/batch = 0.166\n",
            "16681/17600 (epoch 47), train_loss = 1.332, time/batch = 0.166\n",
            "16682/17600 (epoch 47), train_loss = 1.277, time/batch = 0.153\n",
            "16683/17600 (epoch 47), train_loss = 1.331, time/batch = 0.176\n",
            "16684/17600 (epoch 47), train_loss = 1.355, time/batch = 0.144\n",
            "16685/17600 (epoch 47), train_loss = 1.269, time/batch = 0.151\n",
            "16686/17600 (epoch 47), train_loss = 1.294, time/batch = 0.152\n",
            "16687/17600 (epoch 47), train_loss = 1.261, time/batch = 0.176\n",
            "16688/17600 (epoch 47), train_loss = 1.310, time/batch = 0.160\n",
            "16689/17600 (epoch 47), train_loss = 1.296, time/batch = 0.177\n",
            "16690/17600 (epoch 47), train_loss = 1.271, time/batch = 0.153\n",
            "16691/17600 (epoch 47), train_loss = 1.279, time/batch = 0.158\n",
            "16692/17600 (epoch 47), train_loss = 1.260, time/batch = 0.149\n",
            "16693/17600 (epoch 47), train_loss = 1.323, time/batch = 0.167\n",
            "16694/17600 (epoch 47), train_loss = 1.210, time/batch = 0.160\n",
            "16695/17600 (epoch 47), train_loss = 1.286, time/batch = 0.169\n",
            "16696/17600 (epoch 47), train_loss = 1.310, time/batch = 0.151\n",
            "16697/17600 (epoch 47), train_loss = 1.280, time/batch = 0.156\n",
            "16698/17600 (epoch 47), train_loss = 1.295, time/batch = 0.157\n",
            "16699/17600 (epoch 47), train_loss = 1.235, time/batch = 0.166\n",
            "16700/17600 (epoch 47), train_loss = 1.231, time/batch = 0.163\n",
            "16701/17600 (epoch 47), train_loss = 1.250, time/batch = 0.162\n",
            "16702/17600 (epoch 47), train_loss = 1.310, time/batch = 0.165\n",
            "16703/17600 (epoch 47), train_loss = 1.352, time/batch = 0.168\n",
            "16704/17600 (epoch 47), train_loss = 1.309, time/batch = 0.148\n",
            "16705/17600 (epoch 47), train_loss = 1.307, time/batch = 0.170\n",
            "16706/17600 (epoch 47), train_loss = 1.239, time/batch = 0.158\n",
            "16707/17600 (epoch 47), train_loss = 1.265, time/batch = 0.169\n",
            "16708/17600 (epoch 47), train_loss = 1.290, time/batch = 0.173\n",
            "16709/17600 (epoch 47), train_loss = 1.274, time/batch = 0.159\n",
            "16710/17600 (epoch 47), train_loss = 1.288, time/batch = 0.172\n",
            "16711/17600 (epoch 47), train_loss = 1.314, time/batch = 0.167\n",
            "16712/17600 (epoch 47), train_loss = 1.307, time/batch = 0.161\n",
            "16713/17600 (epoch 47), train_loss = 1.288, time/batch = 0.164\n",
            "16714/17600 (epoch 47), train_loss = 1.220, time/batch = 0.169\n",
            "16715/17600 (epoch 47), train_loss = 1.231, time/batch = 0.162\n",
            "16716/17600 (epoch 47), train_loss = 1.322, time/batch = 0.149\n",
            "16717/17600 (epoch 47), train_loss = 1.290, time/batch = 0.173\n",
            "16718/17600 (epoch 47), train_loss = 1.280, time/batch = 0.151\n",
            "16719/17600 (epoch 47), train_loss = 1.284, time/batch = 0.164\n",
            "16720/17600 (epoch 47), train_loss = 1.197, time/batch = 0.154\n",
            "16721/17600 (epoch 47), train_loss = 1.259, time/batch = 0.185\n",
            "16722/17600 (epoch 47), train_loss = 1.259, time/batch = 0.156\n",
            "16723/17600 (epoch 47), train_loss = 1.240, time/batch = 0.157\n",
            "16724/17600 (epoch 47), train_loss = 1.238, time/batch = 0.160\n",
            "16725/17600 (epoch 47), train_loss = 1.293, time/batch = 0.161\n",
            "16726/17600 (epoch 47), train_loss = 1.349, time/batch = 0.155\n",
            "16727/17600 (epoch 47), train_loss = 1.265, time/batch = 0.281\n",
            "16728/17600 (epoch 47), train_loss = 1.265, time/batch = 0.258\n",
            "16729/17600 (epoch 47), train_loss = 1.229, time/batch = 0.287\n",
            "16730/17600 (epoch 47), train_loss = 1.303, time/batch = 0.270\n",
            "16731/17600 (epoch 47), train_loss = 1.313, time/batch = 0.279\n",
            "16732/17600 (epoch 47), train_loss = 1.234, time/batch = 0.253\n",
            "16733/17600 (epoch 47), train_loss = 1.326, time/batch = 0.250\n",
            "16734/17600 (epoch 47), train_loss = 1.296, time/batch = 0.259\n",
            "16735/17600 (epoch 47), train_loss = 1.275, time/batch = 0.279\n",
            "16736/17600 (epoch 47), train_loss = 1.283, time/batch = 0.286\n",
            "16737/17600 (epoch 47), train_loss = 1.302, time/batch = 0.274\n",
            "16738/17600 (epoch 47), train_loss = 1.301, time/batch = 0.158\n",
            "16739/17600 (epoch 47), train_loss = 1.311, time/batch = 0.173\n",
            "16740/17600 (epoch 47), train_loss = 1.266, time/batch = 0.151\n",
            "16741/17600 (epoch 47), train_loss = 1.318, time/batch = 0.160\n",
            "16742/17600 (epoch 47), train_loss = 1.266, time/batch = 0.166\n",
            "16743/17600 (epoch 47), train_loss = 1.300, time/batch = 0.166\n",
            "16744/17600 (epoch 47), train_loss = 1.302, time/batch = 0.148\n",
            "16745/17600 (epoch 47), train_loss = 1.269, time/batch = 0.177\n",
            "16746/17600 (epoch 47), train_loss = 1.259, time/batch = 0.159\n",
            "16747/17600 (epoch 47), train_loss = 1.288, time/batch = 0.159\n",
            "16748/17600 (epoch 47), train_loss = 1.321, time/batch = 0.159\n",
            "16749/17600 (epoch 47), train_loss = 1.368, time/batch = 0.164\n",
            "16750/17600 (epoch 47), train_loss = 1.284, time/batch = 0.155\n",
            "16751/17600 (epoch 47), train_loss = 1.285, time/batch = 0.153\n",
            "16752/17600 (epoch 47), train_loss = 1.289, time/batch = 0.167\n",
            "16753/17600 (epoch 47), train_loss = 1.267, time/batch = 0.153\n",
            "16754/17600 (epoch 47), train_loss = 1.312, time/batch = 0.183\n",
            "16755/17600 (epoch 47), train_loss = 1.304, time/batch = 0.155\n",
            "16756/17600 (epoch 47), train_loss = 1.273, time/batch = 0.164\n",
            "16757/17600 (epoch 47), train_loss = 1.346, time/batch = 0.155\n",
            "16758/17600 (epoch 47), train_loss = 1.384, time/batch = 0.180\n",
            "16759/17600 (epoch 47), train_loss = 1.273, time/batch = 0.168\n",
            "16760/17600 (epoch 47), train_loss = 1.291, time/batch = 0.173\n",
            "16761/17600 (epoch 47), train_loss = 1.251, time/batch = 0.157\n",
            "16762/17600 (epoch 47), train_loss = 1.282, time/batch = 0.165\n",
            "16763/17600 (epoch 47), train_loss = 1.264, time/batch = 0.156\n",
            "16764/17600 (epoch 47), train_loss = 1.260, time/batch = 0.185\n",
            "16765/17600 (epoch 47), train_loss = 1.282, time/batch = 0.155\n",
            "16766/17600 (epoch 47), train_loss = 1.352, time/batch = 0.171\n",
            "16767/17600 (epoch 47), train_loss = 1.326, time/batch = 0.163\n",
            "16768/17600 (epoch 47), train_loss = 1.243, time/batch = 0.172\n",
            "16769/17600 (epoch 47), train_loss = 1.288, time/batch = 0.167\n",
            "16770/17600 (epoch 47), train_loss = 1.293, time/batch = 0.174\n",
            "16771/17600 (epoch 47), train_loss = 1.272, time/batch = 0.170\n",
            "16772/17600 (epoch 47), train_loss = 1.304, time/batch = 0.160\n",
            "16773/17600 (epoch 47), train_loss = 1.269, time/batch = 0.174\n",
            "16774/17600 (epoch 47), train_loss = 1.259, time/batch = 0.158\n",
            "16775/17600 (epoch 47), train_loss = 1.321, time/batch = 0.165\n",
            "16776/17600 (epoch 47), train_loss = 1.321, time/batch = 0.169\n",
            "16777/17600 (epoch 47), train_loss = 1.334, time/batch = 0.164\n",
            "16778/17600 (epoch 47), train_loss = 1.251, time/batch = 0.167\n",
            "16779/17600 (epoch 47), train_loss = 1.281, time/batch = 0.156\n",
            "16780/17600 (epoch 47), train_loss = 1.273, time/batch = 0.158\n",
            "16781/17600 (epoch 47), train_loss = 1.341, time/batch = 0.148\n",
            "16782/17600 (epoch 47), train_loss = 1.339, time/batch = 0.166\n",
            "16783/17600 (epoch 47), train_loss = 1.232, time/batch = 0.178\n",
            "16784/17600 (epoch 47), train_loss = 1.324, time/batch = 0.147\n",
            "16785/17600 (epoch 47), train_loss = 1.272, time/batch = 0.168\n",
            "16786/17600 (epoch 47), train_loss = 1.309, time/batch = 0.150\n",
            "16787/17600 (epoch 47), train_loss = 1.301, time/batch = 0.160\n",
            "16788/17600 (epoch 47), train_loss = 1.269, time/batch = 0.157\n",
            "16789/17600 (epoch 47), train_loss = 1.309, time/batch = 0.181\n",
            "16790/17600 (epoch 47), train_loss = 1.301, time/batch = 0.160\n",
            "16791/17600 (epoch 47), train_loss = 1.273, time/batch = 0.166\n",
            "16792/17600 (epoch 47), train_loss = 1.252, time/batch = 0.151\n",
            "16793/17600 (epoch 47), train_loss = 1.230, time/batch = 0.157\n",
            "16794/17600 (epoch 47), train_loss = 1.363, time/batch = 0.150\n",
            "16795/17600 (epoch 47), train_loss = 1.301, time/batch = 0.186\n",
            "16796/17600 (epoch 47), train_loss = 1.294, time/batch = 0.157\n",
            "16797/17600 (epoch 47), train_loss = 1.302, time/batch = 0.173\n",
            "16798/17600 (epoch 47), train_loss = 1.321, time/batch = 0.193\n",
            "16799/17600 (epoch 47), train_loss = 1.328, time/batch = 0.279\n",
            "16800/17600 (epoch 47), train_loss = 1.373, time/batch = 0.277\n",
            "16801/17600 (epoch 47), train_loss = 1.344, time/batch = 0.292\n",
            "16802/17600 (epoch 47), train_loss = 1.296, time/batch = 0.281\n",
            "16803/17600 (epoch 47), train_loss = 1.299, time/batch = 0.267\n",
            "16804/17600 (epoch 47), train_loss = 1.274, time/batch = 0.300\n",
            "16805/17600 (epoch 47), train_loss = 1.284, time/batch = 0.276\n",
            "16806/17600 (epoch 47), train_loss = 1.262, time/batch = 0.282\n",
            "16807/17600 (epoch 47), train_loss = 1.221, time/batch = 0.297\n",
            "16808/17600 (epoch 47), train_loss = 1.272, time/batch = 0.210\n",
            "16809/17600 (epoch 47), train_loss = 1.278, time/batch = 0.150\n",
            "16810/17600 (epoch 47), train_loss = 1.233, time/batch = 0.157\n",
            "16811/17600 (epoch 47), train_loss = 1.236, time/batch = 0.158\n",
            "16812/17600 (epoch 47), train_loss = 1.266, time/batch = 0.150\n",
            "16813/17600 (epoch 47), train_loss = 1.301, time/batch = 0.162\n",
            "16814/17600 (epoch 47), train_loss = 1.223, time/batch = 0.170\n",
            "16815/17600 (epoch 47), train_loss = 1.294, time/batch = 0.163\n",
            "16816/17600 (epoch 47), train_loss = 1.260, time/batch = 0.165\n",
            "16817/17600 (epoch 47), train_loss = 1.289, time/batch = 0.163\n",
            "16818/17600 (epoch 47), train_loss = 1.271, time/batch = 0.173\n",
            "16819/17600 (epoch 47), train_loss = 1.325, time/batch = 0.154\n",
            "16820/17600 (epoch 47), train_loss = 1.289, time/batch = 0.181\n",
            "16821/17600 (epoch 47), train_loss = 1.314, time/batch = 0.148\n",
            "16822/17600 (epoch 47), train_loss = 1.280, time/batch = 0.173\n",
            "16823/17600 (epoch 47), train_loss = 1.286, time/batch = 0.146\n",
            "16824/17600 (epoch 47), train_loss = 1.309, time/batch = 0.168\n",
            "16825/17600 (epoch 47), train_loss = 1.358, time/batch = 0.152\n",
            "16826/17600 (epoch 47), train_loss = 1.330, time/batch = 0.161\n",
            "16827/17600 (epoch 47), train_loss = 1.327, time/batch = 0.168\n",
            "16828/17600 (epoch 47), train_loss = 1.343, time/batch = 0.163\n",
            "16829/17600 (epoch 47), train_loss = 1.356, time/batch = 0.147\n",
            "16830/17600 (epoch 47), train_loss = 1.304, time/batch = 0.155\n",
            "16831/17600 (epoch 47), train_loss = 1.351, time/batch = 0.167\n",
            "16832/17600 (epoch 47), train_loss = 1.287, time/batch = 0.178\n",
            "16833/17600 (epoch 47), train_loss = 1.282, time/batch = 0.177\n",
            "16834/17600 (epoch 47), train_loss = 1.348, time/batch = 0.167\n",
            "16835/17600 (epoch 47), train_loss = 1.252, time/batch = 0.152\n",
            "16836/17600 (epoch 47), train_loss = 1.249, time/batch = 0.166\n",
            "16837/17600 (epoch 47), train_loss = 1.323, time/batch = 0.166\n",
            "16838/17600 (epoch 47), train_loss = 1.271, time/batch = 0.153\n",
            "16839/17600 (epoch 47), train_loss = 1.330, time/batch = 0.188\n",
            "16840/17600 (epoch 47), train_loss = 1.277, time/batch = 0.148\n",
            "16841/17600 (epoch 47), train_loss = 1.303, time/batch = 0.163\n",
            "16842/17600 (epoch 47), train_loss = 1.291, time/batch = 0.157\n",
            "16843/17600 (epoch 47), train_loss = 1.302, time/batch = 0.162\n",
            "16844/17600 (epoch 47), train_loss = 1.332, time/batch = 0.158\n",
            "16845/17600 (epoch 47), train_loss = 1.292, time/batch = 0.185\n",
            "16846/17600 (epoch 47), train_loss = 1.360, time/batch = 0.157\n",
            "16847/17600 (epoch 47), train_loss = 1.351, time/batch = 0.166\n",
            "16848/17600 (epoch 47), train_loss = 1.362, time/batch = 0.153\n",
            "16849/17600 (epoch 47), train_loss = 1.278, time/batch = 0.162\n",
            "16850/17600 (epoch 47), train_loss = 1.311, time/batch = 0.158\n",
            "16851/17600 (epoch 47), train_loss = 1.320, time/batch = 0.184\n",
            "16852/17600 (epoch 47), train_loss = 1.315, time/batch = 0.167\n",
            "16853/17600 (epoch 47), train_loss = 1.318, time/batch = 0.158\n",
            "16854/17600 (epoch 47), train_loss = 1.355, time/batch = 0.160\n",
            "16855/17600 (epoch 47), train_loss = 1.306, time/batch = 0.169\n",
            "16856/17600 (epoch 47), train_loss = 1.339, time/batch = 0.159\n",
            "16857/17600 (epoch 47), train_loss = 1.274, time/batch = 0.158\n",
            "16858/17600 (epoch 47), train_loss = 1.289, time/batch = 0.169\n",
            "16859/17600 (epoch 47), train_loss = 1.260, time/batch = 0.162\n",
            "16860/17600 (epoch 47), train_loss = 1.353, time/batch = 0.158\n",
            "16861/17600 (epoch 47), train_loss = 1.342, time/batch = 0.168\n",
            "16862/17600 (epoch 47), train_loss = 1.339, time/batch = 0.153\n",
            "16863/17600 (epoch 47), train_loss = 1.323, time/batch = 0.176\n",
            "16864/17600 (epoch 47), train_loss = 1.361, time/batch = 0.169\n",
            "16865/17600 (epoch 47), train_loss = 1.326, time/batch = 0.155\n",
            "16866/17600 (epoch 47), train_loss = 1.384, time/batch = 0.145\n",
            "16867/17600 (epoch 47), train_loss = 1.353, time/batch = 0.158\n",
            "16868/17600 (epoch 47), train_loss = 1.341, time/batch = 0.148\n",
            "16869/17600 (epoch 47), train_loss = 1.288, time/batch = 0.198\n",
            "16870/17600 (epoch 47), train_loss = 1.367, time/batch = 0.283\n",
            "16871/17600 (epoch 47), train_loss = 1.342, time/batch = 0.258\n",
            "16872/17600 (epoch 47), train_loss = 1.380, time/batch = 0.302\n",
            "16873/17600 (epoch 47), train_loss = 1.350, time/batch = 0.281\n",
            "16874/17600 (epoch 47), train_loss = 1.269, time/batch = 0.285\n",
            "16875/17600 (epoch 47), train_loss = 1.299, time/batch = 0.261\n",
            "16876/17600 (epoch 47), train_loss = 1.344, time/batch = 0.251\n",
            "16877/17600 (epoch 47), train_loss = 1.357, time/batch = 0.299\n",
            "16878/17600 (epoch 47), train_loss = 1.313, time/batch = 0.298\n",
            "16879/17600 (epoch 47), train_loss = 1.307, time/batch = 0.299\n",
            "16880/17600 (epoch 47), train_loss = 1.311, time/batch = 0.155\n",
            "16881/17600 (epoch 47), train_loss = 1.353, time/batch = 0.162\n",
            "16882/17600 (epoch 47), train_loss = 1.299, time/batch = 0.152\n",
            "16883/17600 (epoch 47), train_loss = 1.326, time/batch = 0.171\n",
            "16884/17600 (epoch 47), train_loss = 1.296, time/batch = 0.162\n",
            "16885/17600 (epoch 47), train_loss = 1.339, time/batch = 0.168\n",
            "16886/17600 (epoch 47), train_loss = 1.345, time/batch = 0.157\n",
            "16887/17600 (epoch 47), train_loss = 1.308, time/batch = 0.167\n",
            "16888/17600 (epoch 47), train_loss = 1.286, time/batch = 0.152\n",
            "16889/17600 (epoch 47), train_loss = 1.306, time/batch = 0.178\n",
            "16890/17600 (epoch 47), train_loss = 1.330, time/batch = 0.149\n",
            "16891/17600 (epoch 47), train_loss = 1.303, time/batch = 0.166\n",
            "16892/17600 (epoch 47), train_loss = 1.326, time/batch = 0.155\n",
            "16893/17600 (epoch 47), train_loss = 1.312, time/batch = 0.164\n",
            "16894/17600 (epoch 47), train_loss = 1.303, time/batch = 0.154\n",
            "16895/17600 (epoch 47), train_loss = 1.295, time/batch = 0.177\n",
            "16896/17600 (epoch 48), train_loss = 1.080, time/batch = 0.160\n",
            "16897/17600 (epoch 48), train_loss = 1.328, time/batch = 0.148\n",
            "16898/17600 (epoch 48), train_loss = 1.318, time/batch = 0.158\n",
            "16899/17600 (epoch 48), train_loss = 1.265, time/batch = 0.186\n",
            "16900/17600 (epoch 48), train_loss = 1.362, time/batch = 0.186\n",
            "16901/17600 (epoch 48), train_loss = 1.337, time/batch = 0.165\n",
            "16902/17600 (epoch 48), train_loss = 1.280, time/batch = 0.167\n",
            "16903/17600 (epoch 48), train_loss = 1.267, time/batch = 0.164\n",
            "16904/17600 (epoch 48), train_loss = 1.308, time/batch = 0.163\n",
            "16905/17600 (epoch 48), train_loss = 1.360, time/batch = 0.150\n",
            "16906/17600 (epoch 48), train_loss = 1.290, time/batch = 0.179\n",
            "16907/17600 (epoch 48), train_loss = 1.302, time/batch = 0.154\n",
            "16908/17600 (epoch 48), train_loss = 1.286, time/batch = 0.168\n",
            "16909/17600 (epoch 48), train_loss = 1.273, time/batch = 0.160\n",
            "16910/17600 (epoch 48), train_loss = 1.252, time/batch = 0.171\n",
            "16911/17600 (epoch 48), train_loss = 1.318, time/batch = 0.152\n",
            "16912/17600 (epoch 48), train_loss = 1.331, time/batch = 0.173\n",
            "16913/17600 (epoch 48), train_loss = 1.283, time/batch = 0.166\n",
            "16914/17600 (epoch 48), train_loss = 1.296, time/batch = 0.159\n",
            "16915/17600 (epoch 48), train_loss = 1.339, time/batch = 0.170\n",
            "16916/17600 (epoch 48), train_loss = 1.264, time/batch = 0.154\n",
            "16917/17600 (epoch 48), train_loss = 1.319, time/batch = 0.157\n",
            "16918/17600 (epoch 48), train_loss = 1.286, time/batch = 0.158\n",
            "16919/17600 (epoch 48), train_loss = 1.256, time/batch = 0.182\n",
            "16920/17600 (epoch 48), train_loss = 1.340, time/batch = 0.144\n",
            "16921/17600 (epoch 48), train_loss = 1.275, time/batch = 0.162\n",
            "16922/17600 (epoch 48), train_loss = 1.320, time/batch = 0.164\n",
            "16923/17600 (epoch 48), train_loss = 1.304, time/batch = 0.162\n",
            "16924/17600 (epoch 48), train_loss = 1.282, time/batch = 0.153\n",
            "16925/17600 (epoch 48), train_loss = 1.218, time/batch = 0.184\n",
            "16926/17600 (epoch 48), train_loss = 1.325, time/batch = 0.165\n",
            "16927/17600 (epoch 48), train_loss = 1.302, time/batch = 0.157\n",
            "16928/17600 (epoch 48), train_loss = 1.278, time/batch = 0.162\n",
            "16929/17600 (epoch 48), train_loss = 1.280, time/batch = 0.170\n",
            "16930/17600 (epoch 48), train_loss = 1.286, time/batch = 0.145\n",
            "16931/17600 (epoch 48), train_loss = 1.300, time/batch = 0.183\n",
            "16932/17600 (epoch 48), train_loss = 1.329, time/batch = 0.161\n",
            "16933/17600 (epoch 48), train_loss = 1.324, time/batch = 0.168\n",
            "16934/17600 (epoch 48), train_loss = 1.302, time/batch = 0.166\n",
            "16935/17600 (epoch 48), train_loss = 1.281, time/batch = 0.158\n",
            "16936/17600 (epoch 48), train_loss = 1.325, time/batch = 0.151\n",
            "16937/17600 (epoch 48), train_loss = 1.264, time/batch = 0.173\n",
            "16938/17600 (epoch 48), train_loss = 1.319, time/batch = 0.168\n",
            "16939/17600 (epoch 48), train_loss = 1.241, time/batch = 0.220\n",
            "16940/17600 (epoch 48), train_loss = 1.306, time/batch = 0.269\n",
            "16941/17600 (epoch 48), train_loss = 1.280, time/batch = 0.272\n",
            "16942/17600 (epoch 48), train_loss = 1.234, time/batch = 0.296\n",
            "16943/17600 (epoch 48), train_loss = 1.294, time/batch = 0.267\n",
            "16944/17600 (epoch 48), train_loss = 1.246, time/batch = 0.267\n",
            "16945/17600 (epoch 48), train_loss = 1.260, time/batch = 0.263\n",
            "16946/17600 (epoch 48), train_loss = 1.287, time/batch = 0.286\n",
            "16947/17600 (epoch 48), train_loss = 1.276, time/batch = 0.237\n",
            "16948/17600 (epoch 48), train_loss = 1.250, time/batch = 0.295\n",
            "16949/17600 (epoch 48), train_loss = 1.276, time/batch = 0.294\n",
            "16950/17600 (epoch 48), train_loss = 1.319, time/batch = 0.173\n",
            "16951/17600 (epoch 48), train_loss = 1.298, time/batch = 0.158\n",
            "16952/17600 (epoch 48), train_loss = 1.297, time/batch = 0.152\n",
            "16953/17600 (epoch 48), train_loss = 1.291, time/batch = 0.168\n",
            "16954/17600 (epoch 48), train_loss = 1.271, time/batch = 0.158\n",
            "16955/17600 (epoch 48), train_loss = 1.298, time/batch = 0.164\n",
            "16956/17600 (epoch 48), train_loss = 1.295, time/batch = 0.166\n",
            "16957/17600 (epoch 48), train_loss = 1.235, time/batch = 0.180\n",
            "16958/17600 (epoch 48), train_loss = 1.237, time/batch = 0.163\n",
            "16959/17600 (epoch 48), train_loss = 1.257, time/batch = 0.160\n",
            "16960/17600 (epoch 48), train_loss = 1.288, time/batch = 0.156\n",
            "16961/17600 (epoch 48), train_loss = 1.254, time/batch = 0.163\n",
            "16962/17600 (epoch 48), train_loss = 1.290, time/batch = 0.174\n",
            "16963/17600 (epoch 48), train_loss = 1.252, time/batch = 0.168\n",
            "16964/17600 (epoch 48), train_loss = 1.300, time/batch = 0.158\n",
            "16965/17600 (epoch 48), train_loss = 1.231, time/batch = 0.165\n",
            "16966/17600 (epoch 48), train_loss = 1.274, time/batch = 0.161\n",
            "16967/17600 (epoch 48), train_loss = 1.322, time/batch = 0.170\n",
            "16968/17600 (epoch 48), train_loss = 1.283, time/batch = 0.162\n",
            "16969/17600 (epoch 48), train_loss = 1.351, time/batch = 0.163\n",
            "16970/17600 (epoch 48), train_loss = 1.284, time/batch = 0.157\n",
            "16971/17600 (epoch 48), train_loss = 1.285, time/batch = 0.174\n",
            "16972/17600 (epoch 48), train_loss = 1.241, time/batch = 0.149\n",
            "16973/17600 (epoch 48), train_loss = 1.297, time/batch = 0.169\n",
            "16974/17600 (epoch 48), train_loss = 1.302, time/batch = 0.154\n",
            "16975/17600 (epoch 48), train_loss = 1.379, time/batch = 0.177\n",
            "16976/17600 (epoch 48), train_loss = 1.348, time/batch = 0.161\n",
            "16977/17600 (epoch 48), train_loss = 1.341, time/batch = 0.172\n",
            "16978/17600 (epoch 48), train_loss = 1.321, time/batch = 0.150\n",
            "16979/17600 (epoch 48), train_loss = 1.295, time/batch = 0.158\n",
            "16980/17600 (epoch 48), train_loss = 1.270, time/batch = 0.152\n",
            "16981/17600 (epoch 48), train_loss = 1.304, time/batch = 0.172\n",
            "16982/17600 (epoch 48), train_loss = 1.297, time/batch = 0.157\n",
            "16983/17600 (epoch 48), train_loss = 1.235, time/batch = 0.180\n",
            "16984/17600 (epoch 48), train_loss = 1.276, time/batch = 0.159\n",
            "16985/17600 (epoch 48), train_loss = 1.216, time/batch = 0.173\n",
            "16986/17600 (epoch 48), train_loss = 1.279, time/batch = 0.160\n",
            "16987/17600 (epoch 48), train_loss = 1.265, time/batch = 0.184\n",
            "16988/17600 (epoch 48), train_loss = 1.244, time/batch = 0.160\n",
            "16989/17600 (epoch 48), train_loss = 1.298, time/batch = 0.173\n",
            "16990/17600 (epoch 48), train_loss = 1.282, time/batch = 0.149\n",
            "16991/17600 (epoch 48), train_loss = 1.243, time/batch = 0.163\n",
            "16992/17600 (epoch 48), train_loss = 1.217, time/batch = 0.160\n",
            "16993/17600 (epoch 48), train_loss = 1.271, time/batch = 0.174\n",
            "16994/17600 (epoch 48), train_loss = 1.306, time/batch = 0.157\n",
            "16995/17600 (epoch 48), train_loss = 1.265, time/batch = 0.161\n",
            "16996/17600 (epoch 48), train_loss = 1.240, time/batch = 0.152\n",
            "16997/17600 (epoch 48), train_loss = 1.323, time/batch = 0.173\n",
            "16998/17600 (epoch 48), train_loss = 1.262, time/batch = 0.150\n",
            "16999/17600 (epoch 48), train_loss = 1.368, time/batch = 0.170\n",
            "17000/17600 (epoch 48), train_loss = 1.299, time/batch = 0.165\n",
            "Model saved to checkpoints/dracula/dracula!\n",
            "17001/17600 (epoch 48), train_loss = 1.275, time/batch = 0.160\n",
            "17002/17600 (epoch 48), train_loss = 1.314, time/batch = 0.164\n",
            "17003/17600 (epoch 48), train_loss = 1.302, time/batch = 0.157\n",
            "17004/17600 (epoch 48), train_loss = 1.302, time/batch = 0.192\n",
            "17005/17600 (epoch 48), train_loss = 1.286, time/batch = 0.157\n",
            "17006/17600 (epoch 48), train_loss = 1.243, time/batch = 0.165\n",
            "17007/17600 (epoch 48), train_loss = 1.264, time/batch = 0.157\n",
            "17008/17600 (epoch 48), train_loss = 1.297, time/batch = 0.174\n",
            "17009/17600 (epoch 48), train_loss = 1.194, time/batch = 0.161\n",
            "17010/17600 (epoch 48), train_loss = 1.268, time/batch = 0.183\n",
            "17011/17600 (epoch 48), train_loss = 1.296, time/batch = 0.177\n",
            "17012/17600 (epoch 48), train_loss = 1.261, time/batch = 0.179\n",
            "17013/17600 (epoch 48), train_loss = 1.291, time/batch = 0.163\n",
            "17014/17600 (epoch 48), train_loss = 1.232, time/batch = 0.169\n",
            "17015/17600 (epoch 48), train_loss = 1.337, time/batch = 0.168\n",
            "17016/17600 (epoch 48), train_loss = 1.244, time/batch = 0.203\n",
            "17017/17600 (epoch 48), train_loss = 1.305, time/batch = 0.154\n",
            "17018/17600 (epoch 48), train_loss = 1.330, time/batch = 0.170\n",
            "17019/17600 (epoch 48), train_loss = 1.313, time/batch = 0.156\n",
            "17020/17600 (epoch 48), train_loss = 1.231, time/batch = 0.171\n",
            "17021/17600 (epoch 48), train_loss = 1.297, time/batch = 0.162\n",
            "17022/17600 (epoch 48), train_loss = 1.294, time/batch = 0.177\n",
            "17023/17600 (epoch 48), train_loss = 1.281, time/batch = 0.167\n",
            "17024/17600 (epoch 48), train_loss = 1.264, time/batch = 0.165\n",
            "17025/17600 (epoch 48), train_loss = 1.295, time/batch = 0.156\n",
            "17026/17600 (epoch 48), train_loss = 1.308, time/batch = 0.166\n",
            "17027/17600 (epoch 48), train_loss = 1.313, time/batch = 0.167\n",
            "17028/17600 (epoch 48), train_loss = 1.284, time/batch = 0.181\n",
            "17029/17600 (epoch 48), train_loss = 1.254, time/batch = 0.154\n",
            "17030/17600 (epoch 48), train_loss = 1.267, time/batch = 0.163\n",
            "17031/17600 (epoch 48), train_loss = 1.237, time/batch = 0.152\n",
            "17032/17600 (epoch 48), train_loss = 1.303, time/batch = 0.166\n",
            "17033/17600 (epoch 48), train_loss = 1.330, time/batch = 0.158\n",
            "17034/17600 (epoch 48), train_loss = 1.275, time/batch = 0.163\n",
            "17035/17600 (epoch 48), train_loss = 1.330, time/batch = 0.167\n",
            "17036/17600 (epoch 48), train_loss = 1.353, time/batch = 0.162\n",
            "17037/17600 (epoch 48), train_loss = 1.267, time/batch = 0.154\n",
            "17038/17600 (epoch 48), train_loss = 1.292, time/batch = 0.161\n",
            "17039/17600 (epoch 48), train_loss = 1.259, time/batch = 0.166\n",
            "17040/17600 (epoch 48), train_loss = 1.308, time/batch = 0.184\n",
            "17041/17600 (epoch 48), train_loss = 1.295, time/batch = 0.169\n",
            "17042/17600 (epoch 48), train_loss = 1.270, time/batch = 0.162\n",
            "17043/17600 (epoch 48), train_loss = 1.278, time/batch = 0.163\n",
            "17044/17600 (epoch 48), train_loss = 1.259, time/batch = 0.161\n",
            "17045/17600 (epoch 48), train_loss = 1.321, time/batch = 0.160\n",
            "17046/17600 (epoch 48), train_loss = 1.208, time/batch = 0.167\n",
            "17047/17600 (epoch 48), train_loss = 1.285, time/batch = 0.184\n",
            "17048/17600 (epoch 48), train_loss = 1.309, time/batch = 0.158\n",
            "17049/17600 (epoch 48), train_loss = 1.278, time/batch = 0.159\n",
            "17050/17600 (epoch 48), train_loss = 1.293, time/batch = 0.173\n",
            "17051/17600 (epoch 48), train_loss = 1.233, time/batch = 0.150\n",
            "17052/17600 (epoch 48), train_loss = 1.230, time/batch = 0.179\n",
            "17053/17600 (epoch 48), train_loss = 1.249, time/batch = 0.278\n",
            "17054/17600 (epoch 48), train_loss = 1.308, time/batch = 0.261\n",
            "17055/17600 (epoch 48), train_loss = 1.351, time/batch = 0.266\n",
            "17056/17600 (epoch 48), train_loss = 1.308, time/batch = 0.255\n",
            "17057/17600 (epoch 48), train_loss = 1.305, time/batch = 0.288\n",
            "17058/17600 (epoch 48), train_loss = 1.237, time/batch = 0.271\n",
            "17059/17600 (epoch 48), train_loss = 1.263, time/batch = 0.281\n",
            "17060/17600 (epoch 48), train_loss = 1.288, time/batch = 0.261\n",
            "17061/17600 (epoch 48), train_loss = 1.273, time/batch = 0.283\n",
            "17062/17600 (epoch 48), train_loss = 1.286, time/batch = 0.271\n",
            "17063/17600 (epoch 48), train_loss = 1.312, time/batch = 0.272\n",
            "17064/17600 (epoch 48), train_loss = 1.305, time/batch = 0.154\n",
            "17065/17600 (epoch 48), train_loss = 1.286, time/batch = 0.177\n",
            "17066/17600 (epoch 48), train_loss = 1.218, time/batch = 0.155\n",
            "17067/17600 (epoch 48), train_loss = 1.229, time/batch = 0.159\n",
            "17068/17600 (epoch 48), train_loss = 1.320, time/batch = 0.143\n",
            "17069/17600 (epoch 48), train_loss = 1.288, time/batch = 0.156\n",
            "17070/17600 (epoch 48), train_loss = 1.279, time/batch = 0.141\n",
            "17071/17600 (epoch 48), train_loss = 1.282, time/batch = 0.173\n",
            "17072/17600 (epoch 48), train_loss = 1.195, time/batch = 0.179\n",
            "17073/17600 (epoch 48), train_loss = 1.257, time/batch = 0.155\n",
            "17074/17600 (epoch 48), train_loss = 1.257, time/batch = 0.171\n",
            "17075/17600 (epoch 48), train_loss = 1.238, time/batch = 0.155\n",
            "17076/17600 (epoch 48), train_loss = 1.237, time/batch = 0.168\n",
            "17077/17600 (epoch 48), train_loss = 1.291, time/batch = 0.150\n",
            "17078/17600 (epoch 48), train_loss = 1.348, time/batch = 0.181\n",
            "17079/17600 (epoch 48), train_loss = 1.263, time/batch = 0.165\n",
            "17080/17600 (epoch 48), train_loss = 1.263, time/batch = 0.155\n",
            "17081/17600 (epoch 48), train_loss = 1.228, time/batch = 0.166\n",
            "17082/17600 (epoch 48), train_loss = 1.301, time/batch = 0.156\n",
            "17083/17600 (epoch 48), train_loss = 1.312, time/batch = 0.157\n",
            "17084/17600 (epoch 48), train_loss = 1.233, time/batch = 0.178\n",
            "17085/17600 (epoch 48), train_loss = 1.324, time/batch = 0.178\n",
            "17086/17600 (epoch 48), train_loss = 1.295, time/batch = 0.165\n",
            "17087/17600 (epoch 48), train_loss = 1.273, time/batch = 0.160\n",
            "17088/17600 (epoch 48), train_loss = 1.281, time/batch = 0.165\n",
            "17089/17600 (epoch 48), train_loss = 1.300, time/batch = 0.163\n",
            "17090/17600 (epoch 48), train_loss = 1.300, time/batch = 0.168\n",
            "17091/17600 (epoch 48), train_loss = 1.309, time/batch = 0.166\n",
            "17092/17600 (epoch 48), train_loss = 1.264, time/batch = 0.168\n",
            "17093/17600 (epoch 48), train_loss = 1.316, time/batch = 0.161\n",
            "17094/17600 (epoch 48), train_loss = 1.264, time/batch = 0.173\n",
            "17095/17600 (epoch 48), train_loss = 1.299, time/batch = 0.154\n",
            "17096/17600 (epoch 48), train_loss = 1.300, time/batch = 0.160\n",
            "17097/17600 (epoch 48), train_loss = 1.268, time/batch = 0.185\n",
            "17098/17600 (epoch 48), train_loss = 1.257, time/batch = 0.166\n",
            "17099/17600 (epoch 48), train_loss = 1.286, time/batch = 0.169\n",
            "17100/17600 (epoch 48), train_loss = 1.319, time/batch = 0.160\n",
            "17101/17600 (epoch 48), train_loss = 1.366, time/batch = 0.167\n",
            "17102/17600 (epoch 48), train_loss = 1.283, time/batch = 0.163\n",
            "17103/17600 (epoch 48), train_loss = 1.283, time/batch = 0.177\n",
            "17104/17600 (epoch 48), train_loss = 1.288, time/batch = 0.150\n",
            "17105/17600 (epoch 48), train_loss = 1.265, time/batch = 0.165\n",
            "17106/17600 (epoch 48), train_loss = 1.310, time/batch = 0.151\n",
            "17107/17600 (epoch 48), train_loss = 1.303, time/batch = 0.183\n",
            "17108/17600 (epoch 48), train_loss = 1.272, time/batch = 0.157\n",
            "17109/17600 (epoch 48), train_loss = 1.345, time/batch = 0.173\n",
            "17110/17600 (epoch 48), train_loss = 1.382, time/batch = 0.155\n",
            "17111/17600 (epoch 48), train_loss = 1.271, time/batch = 0.163\n",
            "17112/17600 (epoch 48), train_loss = 1.289, time/batch = 0.150\n",
            "17113/17600 (epoch 48), train_loss = 1.249, time/batch = 0.175\n",
            "17114/17600 (epoch 48), train_loss = 1.280, time/batch = 0.166\n",
            "17115/17600 (epoch 48), train_loss = 1.262, time/batch = 0.175\n",
            "17116/17600 (epoch 48), train_loss = 1.258, time/batch = 0.154\n",
            "17117/17600 (epoch 48), train_loss = 1.280, time/batch = 0.171\n",
            "17118/17600 (epoch 48), train_loss = 1.350, time/batch = 0.153\n",
            "17119/17600 (epoch 48), train_loss = 1.324, time/batch = 0.173\n",
            "17120/17600 (epoch 48), train_loss = 1.242, time/batch = 0.157\n",
            "17121/17600 (epoch 48), train_loss = 1.285, time/batch = 0.152\n",
            "17122/17600 (epoch 48), train_loss = 1.291, time/batch = 0.171\n",
            "17123/17600 (epoch 48), train_loss = 1.270, time/batch = 0.184\n",
            "17124/17600 (epoch 48), train_loss = 1.303, time/batch = 0.229\n",
            "17125/17600 (epoch 48), train_loss = 1.267, time/batch = 0.280\n",
            "17126/17600 (epoch 48), train_loss = 1.257, time/batch = 0.293\n",
            "17127/17600 (epoch 48), train_loss = 1.319, time/batch = 0.293\n",
            "17128/17600 (epoch 48), train_loss = 1.319, time/batch = 0.284\n",
            "17129/17600 (epoch 48), train_loss = 1.332, time/batch = 0.274\n",
            "17130/17600 (epoch 48), train_loss = 1.249, time/batch = 0.287\n",
            "17131/17600 (epoch 48), train_loss = 1.279, time/batch = 0.281\n",
            "17132/17600 (epoch 48), train_loss = 1.271, time/batch = 0.315\n",
            "17133/17600 (epoch 48), train_loss = 1.339, time/batch = 0.280\n",
            "17134/17600 (epoch 48), train_loss = 1.337, time/batch = 0.251\n",
            "17135/17600 (epoch 48), train_loss = 1.231, time/batch = 0.155\n",
            "17136/17600 (epoch 48), train_loss = 1.323, time/batch = 0.166\n",
            "17137/17600 (epoch 48), train_loss = 1.270, time/batch = 0.167\n",
            "17138/17600 (epoch 48), train_loss = 1.308, time/batch = 0.168\n",
            "17139/17600 (epoch 48), train_loss = 1.300, time/batch = 0.162\n",
            "17140/17600 (epoch 48), train_loss = 1.267, time/batch = 0.171\n",
            "17141/17600 (epoch 48), train_loss = 1.308, time/batch = 0.151\n",
            "17142/17600 (epoch 48), train_loss = 1.300, time/batch = 0.169\n",
            "17143/17600 (epoch 48), train_loss = 1.272, time/batch = 0.151\n",
            "17144/17600 (epoch 48), train_loss = 1.251, time/batch = 0.161\n",
            "17145/17600 (epoch 48), train_loss = 1.229, time/batch = 0.159\n",
            "17146/17600 (epoch 48), train_loss = 1.361, time/batch = 0.188\n",
            "17147/17600 (epoch 48), train_loss = 1.299, time/batch = 0.159\n",
            "17148/17600 (epoch 48), train_loss = 1.292, time/batch = 0.161\n",
            "17149/17600 (epoch 48), train_loss = 1.301, time/batch = 0.156\n",
            "17150/17600 (epoch 48), train_loss = 1.319, time/batch = 0.160\n",
            "17151/17600 (epoch 48), train_loss = 1.326, time/batch = 0.155\n",
            "17152/17600 (epoch 48), train_loss = 1.371, time/batch = 0.174\n",
            "17153/17600 (epoch 48), train_loss = 1.343, time/batch = 0.147\n",
            "17154/17600 (epoch 48), train_loss = 1.295, time/batch = 0.173\n",
            "17155/17600 (epoch 48), train_loss = 1.297, time/batch = 0.151\n",
            "17156/17600 (epoch 48), train_loss = 1.272, time/batch = 0.167\n",
            "17157/17600 (epoch 48), train_loss = 1.282, time/batch = 0.164\n",
            "17158/17600 (epoch 48), train_loss = 1.260, time/batch = 0.178\n",
            "17159/17600 (epoch 48), train_loss = 1.219, time/batch = 0.155\n",
            "17160/17600 (epoch 48), train_loss = 1.270, time/batch = 0.167\n",
            "17161/17600 (epoch 48), train_loss = 1.277, time/batch = 0.153\n",
            "17162/17600 (epoch 48), train_loss = 1.231, time/batch = 0.154\n",
            "17163/17600 (epoch 48), train_loss = 1.234, time/batch = 0.154\n",
            "17164/17600 (epoch 48), train_loss = 1.264, time/batch = 0.159\n",
            "17165/17600 (epoch 48), train_loss = 1.299, time/batch = 0.164\n",
            "17166/17600 (epoch 48), train_loss = 1.221, time/batch = 0.160\n",
            "17167/17600 (epoch 48), train_loss = 1.292, time/batch = 0.150\n",
            "17168/17600 (epoch 48), train_loss = 1.259, time/batch = 0.173\n",
            "17169/17600 (epoch 48), train_loss = 1.288, time/batch = 0.154\n",
            "17170/17600 (epoch 48), train_loss = 1.270, time/batch = 0.163\n",
            "17171/17600 (epoch 48), train_loss = 1.323, time/batch = 0.166\n",
            "17172/17600 (epoch 48), train_loss = 1.288, time/batch = 0.171\n",
            "17173/17600 (epoch 48), train_loss = 1.312, time/batch = 0.144\n",
            "17174/17600 (epoch 48), train_loss = 1.278, time/batch = 0.171\n",
            "17175/17600 (epoch 48), train_loss = 1.284, time/batch = 0.157\n",
            "17176/17600 (epoch 48), train_loss = 1.307, time/batch = 0.170\n",
            "17177/17600 (epoch 48), train_loss = 1.356, time/batch = 0.160\n",
            "17178/17600 (epoch 48), train_loss = 1.329, time/batch = 0.185\n",
            "17179/17600 (epoch 48), train_loss = 1.326, time/batch = 0.150\n",
            "17180/17600 (epoch 48), train_loss = 1.341, time/batch = 0.154\n",
            "17181/17600 (epoch 48), train_loss = 1.354, time/batch = 0.152\n",
            "17182/17600 (epoch 48), train_loss = 1.302, time/batch = 0.170\n",
            "17183/17600 (epoch 48), train_loss = 1.350, time/batch = 0.150\n",
            "17184/17600 (epoch 48), train_loss = 1.286, time/batch = 0.183\n",
            "17185/17600 (epoch 48), train_loss = 1.280, time/batch = 0.163\n",
            "17186/17600 (epoch 48), train_loss = 1.346, time/batch = 0.168\n",
            "17187/17600 (epoch 48), train_loss = 1.250, time/batch = 0.153\n",
            "17188/17600 (epoch 48), train_loss = 1.247, time/batch = 0.172\n",
            "17189/17600 (epoch 48), train_loss = 1.322, time/batch = 0.151\n",
            "17190/17600 (epoch 48), train_loss = 1.269, time/batch = 0.194\n",
            "17191/17600 (epoch 48), train_loss = 1.328, time/batch = 0.158\n",
            "17192/17600 (epoch 48), train_loss = 1.275, time/batch = 0.161\n",
            "17193/17600 (epoch 48), train_loss = 1.301, time/batch = 0.152\n",
            "17194/17600 (epoch 48), train_loss = 1.290, time/batch = 0.179\n",
            "17195/17600 (epoch 48), train_loss = 1.300, time/batch = 0.168\n",
            "17196/17600 (epoch 48), train_loss = 1.330, time/batch = 0.272\n",
            "17197/17600 (epoch 48), train_loss = 1.291, time/batch = 0.285\n",
            "17198/17600 (epoch 48), train_loss = 1.359, time/batch = 0.288\n",
            "17199/17600 (epoch 48), train_loss = 1.349, time/batch = 0.269\n",
            "17200/17600 (epoch 48), train_loss = 1.360, time/batch = 0.283\n",
            "17201/17600 (epoch 48), train_loss = 1.277, time/batch = 0.271\n",
            "17202/17600 (epoch 48), train_loss = 1.309, time/batch = 0.301\n",
            "17203/17600 (epoch 48), train_loss = 1.319, time/batch = 0.326\n",
            "17204/17600 (epoch 48), train_loss = 1.314, time/batch = 0.276\n",
            "17205/17600 (epoch 48), train_loss = 1.317, time/batch = 0.294\n",
            "17206/17600 (epoch 48), train_loss = 1.353, time/batch = 0.172\n",
            "17207/17600 (epoch 48), train_loss = 1.304, time/batch = 0.166\n",
            "17208/17600 (epoch 48), train_loss = 1.338, time/batch = 0.170\n",
            "17209/17600 (epoch 48), train_loss = 1.273, time/batch = 0.172\n",
            "17210/17600 (epoch 48), train_loss = 1.287, time/batch = 0.153\n",
            "17211/17600 (epoch 48), train_loss = 1.259, time/batch = 0.170\n",
            "17212/17600 (epoch 48), train_loss = 1.352, time/batch = 0.145\n",
            "17213/17600 (epoch 48), train_loss = 1.341, time/batch = 0.173\n",
            "17214/17600 (epoch 48), train_loss = 1.337, time/batch = 0.162\n",
            "17215/17600 (epoch 48), train_loss = 1.321, time/batch = 0.172\n",
            "17216/17600 (epoch 48), train_loss = 1.359, time/batch = 0.169\n",
            "17217/17600 (epoch 48), train_loss = 1.325, time/batch = 0.145\n",
            "17218/17600 (epoch 48), train_loss = 1.382, time/batch = 0.159\n",
            "17219/17600 (epoch 48), train_loss = 1.352, time/batch = 0.162\n",
            "17220/17600 (epoch 48), train_loss = 1.339, time/batch = 0.166\n",
            "17221/17600 (epoch 48), train_loss = 1.287, time/batch = 0.169\n",
            "17222/17600 (epoch 48), train_loss = 1.365, time/batch = 0.174\n",
            "17223/17600 (epoch 48), train_loss = 1.341, time/batch = 0.166\n",
            "17224/17600 (epoch 48), train_loss = 1.379, time/batch = 0.160\n",
            "17225/17600 (epoch 48), train_loss = 1.348, time/batch = 0.158\n",
            "17226/17600 (epoch 48), train_loss = 1.267, time/batch = 0.172\n",
            "17227/17600 (epoch 48), train_loss = 1.298, time/batch = 0.173\n",
            "17228/17600 (epoch 48), train_loss = 1.342, time/batch = 0.167\n",
            "17229/17600 (epoch 48), train_loss = 1.356, time/batch = 0.151\n",
            "17230/17600 (epoch 48), train_loss = 1.312, time/batch = 0.166\n",
            "17231/17600 (epoch 48), train_loss = 1.306, time/batch = 0.166\n",
            "17232/17600 (epoch 48), train_loss = 1.308, time/batch = 0.162\n",
            "17233/17600 (epoch 48), train_loss = 1.351, time/batch = 0.170\n",
            "17234/17600 (epoch 48), train_loss = 1.297, time/batch = 0.175\n",
            "17235/17600 (epoch 48), train_loss = 1.324, time/batch = 0.151\n",
            "17236/17600 (epoch 48), train_loss = 1.294, time/batch = 0.168\n",
            "17237/17600 (epoch 48), train_loss = 1.338, time/batch = 0.155\n",
            "17238/17600 (epoch 48), train_loss = 1.343, time/batch = 0.169\n",
            "17239/17600 (epoch 48), train_loss = 1.307, time/batch = 0.186\n",
            "17240/17600 (epoch 48), train_loss = 1.284, time/batch = 0.164\n",
            "17241/17600 (epoch 48), train_loss = 1.305, time/batch = 0.159\n",
            "17242/17600 (epoch 48), train_loss = 1.328, time/batch = 0.168\n",
            "17243/17600 (epoch 48), train_loss = 1.302, time/batch = 0.160\n",
            "17244/17600 (epoch 48), train_loss = 1.324, time/batch = 0.158\n",
            "17245/17600 (epoch 48), train_loss = 1.311, time/batch = 0.175\n",
            "17246/17600 (epoch 48), train_loss = 1.302, time/batch = 0.148\n",
            "17247/17600 (epoch 48), train_loss = 1.293, time/batch = 0.159\n",
            "17248/17600 (epoch 49), train_loss = 1.077, time/batch = 0.169\n",
            "17249/17600 (epoch 49), train_loss = 1.326, time/batch = 0.150\n",
            "17250/17600 (epoch 49), train_loss = 1.316, time/batch = 0.170\n",
            "17251/17600 (epoch 49), train_loss = 1.264, time/batch = 0.146\n",
            "17252/17600 (epoch 49), train_loss = 1.361, time/batch = 0.152\n",
            "17253/17600 (epoch 49), train_loss = 1.335, time/batch = 0.145\n",
            "17254/17600 (epoch 49), train_loss = 1.278, time/batch = 0.168\n",
            "17255/17600 (epoch 49), train_loss = 1.265, time/batch = 0.161\n",
            "17256/17600 (epoch 49), train_loss = 1.306, time/batch = 0.161\n",
            "17257/17600 (epoch 49), train_loss = 1.358, time/batch = 0.174\n",
            "17258/17600 (epoch 49), train_loss = 1.288, time/batch = 0.170\n",
            "17259/17600 (epoch 49), train_loss = 1.300, time/batch = 0.158\n",
            "17260/17600 (epoch 49), train_loss = 1.285, time/batch = 0.171\n",
            "17261/17600 (epoch 49), train_loss = 1.272, time/batch = 0.166\n",
            "17262/17600 (epoch 49), train_loss = 1.251, time/batch = 0.167\n",
            "17263/17600 (epoch 49), train_loss = 1.318, time/batch = 0.183\n",
            "17264/17600 (epoch 49), train_loss = 1.329, time/batch = 0.158\n",
            "17265/17600 (epoch 49), train_loss = 1.282, time/batch = 0.215\n",
            "17266/17600 (epoch 49), train_loss = 1.295, time/batch = 0.289\n",
            "17267/17600 (epoch 49), train_loss = 1.337, time/batch = 0.280\n",
            "17268/17600 (epoch 49), train_loss = 1.262, time/batch = 0.247\n",
            "17269/17600 (epoch 49), train_loss = 1.318, time/batch = 0.270\n",
            "17270/17600 (epoch 49), train_loss = 1.285, time/batch = 0.260\n",
            "17271/17600 (epoch 49), train_loss = 1.255, time/batch = 0.288\n",
            "17272/17600 (epoch 49), train_loss = 1.338, time/batch = 0.285\n",
            "17273/17600 (epoch 49), train_loss = 1.274, time/batch = 0.286\n",
            "17274/17600 (epoch 49), train_loss = 1.318, time/batch = 0.299\n",
            "17275/17600 (epoch 49), train_loss = 1.302, time/batch = 0.298\n",
            "17276/17600 (epoch 49), train_loss = 1.281, time/batch = 0.163\n",
            "17277/17600 (epoch 49), train_loss = 1.216, time/batch = 0.185\n",
            "17278/17600 (epoch 49), train_loss = 1.324, time/batch = 0.162\n",
            "17279/17600 (epoch 49), train_loss = 1.301, time/batch = 0.165\n",
            "17280/17600 (epoch 49), train_loss = 1.277, time/batch = 0.158\n",
            "17281/17600 (epoch 49), train_loss = 1.279, time/batch = 0.205\n",
            "17282/17600 (epoch 49), train_loss = 1.284, time/batch = 0.170\n",
            "17283/17600 (epoch 49), train_loss = 1.299, time/batch = 0.161\n",
            "17284/17600 (epoch 49), train_loss = 1.328, time/batch = 0.150\n",
            "17285/17600 (epoch 49), train_loss = 1.323, time/batch = 0.166\n",
            "17286/17600 (epoch 49), train_loss = 1.300, time/batch = 0.156\n",
            "17287/17600 (epoch 49), train_loss = 1.280, time/batch = 0.178\n",
            "17288/17600 (epoch 49), train_loss = 1.323, time/batch = 0.151\n",
            "17289/17600 (epoch 49), train_loss = 1.262, time/batch = 0.174\n",
            "17290/17600 (epoch 49), train_loss = 1.317, time/batch = 0.158\n",
            "17291/17600 (epoch 49), train_loss = 1.239, time/batch = 0.159\n",
            "17292/17600 (epoch 49), train_loss = 1.304, time/batch = 0.152\n",
            "17293/17600 (epoch 49), train_loss = 1.279, time/batch = 0.178\n",
            "17294/17600 (epoch 49), train_loss = 1.232, time/batch = 0.159\n",
            "17295/17600 (epoch 49), train_loss = 1.293, time/batch = 0.161\n",
            "17296/17600 (epoch 49), train_loss = 1.244, time/batch = 0.154\n",
            "17297/17600 (epoch 49), train_loss = 1.259, time/batch = 0.161\n",
            "17298/17600 (epoch 49), train_loss = 1.285, time/batch = 0.158\n",
            "17299/17600 (epoch 49), train_loss = 1.274, time/batch = 0.160\n",
            "17300/17600 (epoch 49), train_loss = 1.249, time/batch = 0.165\n",
            "17301/17600 (epoch 49), train_loss = 1.275, time/batch = 0.174\n",
            "17302/17600 (epoch 49), train_loss = 1.317, time/batch = 0.158\n",
            "17303/17600 (epoch 49), train_loss = 1.296, time/batch = 0.158\n",
            "17304/17600 (epoch 49), train_loss = 1.295, time/batch = 0.154\n",
            "17305/17600 (epoch 49), train_loss = 1.290, time/batch = 0.166\n",
            "17306/17600 (epoch 49), train_loss = 1.269, time/batch = 0.172\n",
            "17307/17600 (epoch 49), train_loss = 1.297, time/batch = 0.160\n",
            "17308/17600 (epoch 49), train_loss = 1.293, time/batch = 0.171\n",
            "17309/17600 (epoch 49), train_loss = 1.233, time/batch = 0.161\n",
            "17310/17600 (epoch 49), train_loss = 1.235, time/batch = 0.157\n",
            "17311/17600 (epoch 49), train_loss = 1.256, time/batch = 0.163\n",
            "17312/17600 (epoch 49), train_loss = 1.287, time/batch = 0.182\n",
            "17313/17600 (epoch 49), train_loss = 1.252, time/batch = 0.174\n",
            "17314/17600 (epoch 49), train_loss = 1.288, time/batch = 0.156\n",
            "17315/17600 (epoch 49), train_loss = 1.250, time/batch = 0.153\n",
            "17316/17600 (epoch 49), train_loss = 1.299, time/batch = 0.151\n",
            "17317/17600 (epoch 49), train_loss = 1.230, time/batch = 0.155\n",
            "17318/17600 (epoch 49), train_loss = 1.273, time/batch = 0.155\n",
            "17319/17600 (epoch 49), train_loss = 1.321, time/batch = 0.185\n",
            "17320/17600 (epoch 49), train_loss = 1.282, time/batch = 0.167\n",
            "17321/17600 (epoch 49), train_loss = 1.350, time/batch = 0.155\n",
            "17322/17600 (epoch 49), train_loss = 1.283, time/batch = 0.165\n",
            "17323/17600 (epoch 49), train_loss = 1.283, time/batch = 0.154\n",
            "17324/17600 (epoch 49), train_loss = 1.239, time/batch = 0.167\n",
            "17325/17600 (epoch 49), train_loss = 1.296, time/batch = 0.177\n",
            "17326/17600 (epoch 49), train_loss = 1.300, time/batch = 0.154\n",
            "17327/17600 (epoch 49), train_loss = 1.377, time/batch = 0.163\n",
            "17328/17600 (epoch 49), train_loss = 1.347, time/batch = 0.147\n",
            "17329/17600 (epoch 49), train_loss = 1.339, time/batch = 0.164\n",
            "17330/17600 (epoch 49), train_loss = 1.319, time/batch = 0.150\n",
            "17331/17600 (epoch 49), train_loss = 1.294, time/batch = 0.169\n",
            "17332/17600 (epoch 49), train_loss = 1.268, time/batch = 0.159\n",
            "17333/17600 (epoch 49), train_loss = 1.302, time/batch = 0.165\n",
            "17334/17600 (epoch 49), train_loss = 1.295, time/batch = 0.159\n",
            "17335/17600 (epoch 49), train_loss = 1.234, time/batch = 0.174\n",
            "17336/17600 (epoch 49), train_loss = 1.274, time/batch = 0.192\n",
            "17337/17600 (epoch 49), train_loss = 1.215, time/batch = 0.259\n",
            "17338/17600 (epoch 49), train_loss = 1.277, time/batch = 0.296\n",
            "17339/17600 (epoch 49), train_loss = 1.264, time/batch = 0.302\n",
            "17340/17600 (epoch 49), train_loss = 1.242, time/batch = 0.282\n",
            "17341/17600 (epoch 49), train_loss = 1.296, time/batch = 0.289\n",
            "17342/17600 (epoch 49), train_loss = 1.281, time/batch = 0.270\n",
            "17343/17600 (epoch 49), train_loss = 1.241, time/batch = 0.307\n",
            "17344/17600 (epoch 49), train_loss = 1.216, time/batch = 0.277\n",
            "17345/17600 (epoch 49), train_loss = 1.269, time/batch = 0.300\n",
            "17346/17600 (epoch 49), train_loss = 1.305, time/batch = 0.277\n",
            "17347/17600 (epoch 49), train_loss = 1.263, time/batch = 0.152\n",
            "17348/17600 (epoch 49), train_loss = 1.239, time/batch = 0.165\n",
            "17349/17600 (epoch 49), train_loss = 1.321, time/batch = 0.168\n",
            "17350/17600 (epoch 49), train_loss = 1.260, time/batch = 0.174\n",
            "17351/17600 (epoch 49), train_loss = 1.366, time/batch = 0.147\n",
            "17352/17600 (epoch 49), train_loss = 1.298, time/batch = 0.159\n",
            "17353/17600 (epoch 49), train_loss = 1.273, time/batch = 0.159\n",
            "17354/17600 (epoch 49), train_loss = 1.313, time/batch = 0.171\n",
            "17355/17600 (epoch 49), train_loss = 1.301, time/batch = 0.158\n",
            "17356/17600 (epoch 49), train_loss = 1.300, time/batch = 0.195\n",
            "17357/17600 (epoch 49), train_loss = 1.285, time/batch = 0.153\n",
            "17358/17600 (epoch 49), train_loss = 1.241, time/batch = 0.151\n",
            "17359/17600 (epoch 49), train_loss = 1.263, time/batch = 0.157\n",
            "17360/17600 (epoch 49), train_loss = 1.295, time/batch = 0.154\n",
            "17361/17600 (epoch 49), train_loss = 1.192, time/batch = 0.164\n",
            "17362/17600 (epoch 49), train_loss = 1.266, time/batch = 0.178\n",
            "17363/17600 (epoch 49), train_loss = 1.295, time/batch = 0.154\n",
            "17364/17600 (epoch 49), train_loss = 1.259, time/batch = 0.158\n",
            "17365/17600 (epoch 49), train_loss = 1.289, time/batch = 0.167\n",
            "17366/17600 (epoch 49), train_loss = 1.230, time/batch = 0.171\n",
            "17367/17600 (epoch 49), train_loss = 1.336, time/batch = 0.147\n",
            "17368/17600 (epoch 49), train_loss = 1.242, time/batch = 0.170\n",
            "17369/17600 (epoch 49), train_loss = 1.304, time/batch = 0.165\n",
            "17370/17600 (epoch 49), train_loss = 1.329, time/batch = 0.155\n",
            "17371/17600 (epoch 49), train_loss = 1.312, time/batch = 0.150\n",
            "17372/17600 (epoch 49), train_loss = 1.229, time/batch = 0.162\n",
            "17373/17600 (epoch 49), train_loss = 1.296, time/batch = 0.151\n",
            "17374/17600 (epoch 49), train_loss = 1.293, time/batch = 0.168\n",
            "17375/17600 (epoch 49), train_loss = 1.280, time/batch = 0.168\n",
            "17376/17600 (epoch 49), train_loss = 1.262, time/batch = 0.177\n",
            "17377/17600 (epoch 49), train_loss = 1.294, time/batch = 0.151\n",
            "17378/17600 (epoch 49), train_loss = 1.306, time/batch = 0.165\n",
            "17379/17600 (epoch 49), train_loss = 1.312, time/batch = 0.153\n",
            "17380/17600 (epoch 49), train_loss = 1.283, time/batch = 0.173\n",
            "17381/17600 (epoch 49), train_loss = 1.252, time/batch = 0.166\n",
            "17382/17600 (epoch 49), train_loss = 1.266, time/batch = 0.162\n",
            "17383/17600 (epoch 49), train_loss = 1.235, time/batch = 0.157\n",
            "17384/17600 (epoch 49), train_loss = 1.301, time/batch = 0.159\n",
            "17385/17600 (epoch 49), train_loss = 1.329, time/batch = 0.172\n",
            "17386/17600 (epoch 49), train_loss = 1.274, time/batch = 0.154\n",
            "17387/17600 (epoch 49), train_loss = 1.329, time/batch = 0.155\n",
            "17388/17600 (epoch 49), train_loss = 1.352, time/batch = 0.175\n",
            "17389/17600 (epoch 49), train_loss = 1.266, time/batch = 0.175\n",
            "17390/17600 (epoch 49), train_loss = 1.291, time/batch = 0.163\n",
            "17391/17600 (epoch 49), train_loss = 1.258, time/batch = 0.164\n",
            "17392/17600 (epoch 49), train_loss = 1.307, time/batch = 0.153\n",
            "17393/17600 (epoch 49), train_loss = 1.294, time/batch = 0.168\n",
            "17394/17600 (epoch 49), train_loss = 1.269, time/batch = 0.171\n",
            "17395/17600 (epoch 49), train_loss = 1.276, time/batch = 0.164\n",
            "17396/17600 (epoch 49), train_loss = 1.257, time/batch = 0.163\n",
            "17397/17600 (epoch 49), train_loss = 1.320, time/batch = 0.160\n",
            "17398/17600 (epoch 49), train_loss = 1.206, time/batch = 0.157\n",
            "17399/17600 (epoch 49), train_loss = 1.283, time/batch = 0.160\n",
            "17400/17600 (epoch 49), train_loss = 1.307, time/batch = 0.172\n",
            "17401/17600 (epoch 49), train_loss = 1.277, time/batch = 0.163\n",
            "17402/17600 (epoch 49), train_loss = 1.291, time/batch = 0.155\n",
            "17403/17600 (epoch 49), train_loss = 1.232, time/batch = 0.167\n",
            "17404/17600 (epoch 49), train_loss = 1.228, time/batch = 0.158\n",
            "17405/17600 (epoch 49), train_loss = 1.248, time/batch = 0.165\n",
            "17406/17600 (epoch 49), train_loss = 1.307, time/batch = 0.166\n",
            "17407/17600 (epoch 49), train_loss = 1.349, time/batch = 0.163\n",
            "17408/17600 (epoch 49), train_loss = 1.307, time/batch = 0.261\n",
            "17409/17600 (epoch 49), train_loss = 1.304, time/batch = 0.299\n",
            "17410/17600 (epoch 49), train_loss = 1.236, time/batch = 0.281\n",
            "17411/17600 (epoch 49), train_loss = 1.262, time/batch = 0.280\n",
            "17412/17600 (epoch 49), train_loss = 1.287, time/batch = 0.268\n",
            "17413/17600 (epoch 49), train_loss = 1.271, time/batch = 0.284\n",
            "17414/17600 (epoch 49), train_loss = 1.285, time/batch = 0.299\n",
            "17415/17600 (epoch 49), train_loss = 1.310, time/batch = 0.303\n",
            "17416/17600 (epoch 49), train_loss = 1.304, time/batch = 0.274\n",
            "17417/17600 (epoch 49), train_loss = 1.284, time/batch = 0.280\n",
            "17418/17600 (epoch 49), train_loss = 1.216, time/batch = 0.198\n",
            "17419/17600 (epoch 49), train_loss = 1.228, time/batch = 0.167\n",
            "17420/17600 (epoch 49), train_loss = 1.319, time/batch = 0.159\n",
            "17421/17600 (epoch 49), train_loss = 1.287, time/batch = 0.167\n",
            "17422/17600 (epoch 49), train_loss = 1.278, time/batch = 0.160\n",
            "17423/17600 (epoch 49), train_loss = 1.281, time/batch = 0.170\n",
            "17424/17600 (epoch 49), train_loss = 1.194, time/batch = 0.173\n",
            "17425/17600 (epoch 49), train_loss = 1.256, time/batch = 0.164\n",
            "17426/17600 (epoch 49), train_loss = 1.256, time/batch = 0.152\n",
            "17427/17600 (epoch 49), train_loss = 1.237, time/batch = 0.163\n",
            "17428/17600 (epoch 49), train_loss = 1.235, time/batch = 0.148\n",
            "17429/17600 (epoch 49), train_loss = 1.290, time/batch = 0.161\n",
            "17430/17600 (epoch 49), train_loss = 1.347, time/batch = 0.153\n",
            "17431/17600 (epoch 49), train_loss = 1.262, time/batch = 0.174\n",
            "17432/17600 (epoch 49), train_loss = 1.262, time/batch = 0.154\n",
            "17433/17600 (epoch 49), train_loss = 1.226, time/batch = 0.168\n",
            "17434/17600 (epoch 49), train_loss = 1.299, time/batch = 0.154\n",
            "17435/17600 (epoch 49), train_loss = 1.310, time/batch = 0.168\n",
            "17436/17600 (epoch 49), train_loss = 1.231, time/batch = 0.157\n",
            "17437/17600 (epoch 49), train_loss = 1.322, time/batch = 0.172\n",
            "17438/17600 (epoch 49), train_loss = 1.294, time/batch = 0.157\n",
            "17439/17600 (epoch 49), train_loss = 1.272, time/batch = 0.166\n",
            "17440/17600 (epoch 49), train_loss = 1.280, time/batch = 0.153\n",
            "17441/17600 (epoch 49), train_loss = 1.299, time/batch = 0.171\n",
            "17442/17600 (epoch 49), train_loss = 1.298, time/batch = 0.150\n",
            "17443/17600 (epoch 49), train_loss = 1.308, time/batch = 0.175\n",
            "17444/17600 (epoch 49), train_loss = 1.262, time/batch = 0.159\n",
            "17445/17600 (epoch 49), train_loss = 1.315, time/batch = 0.161\n",
            "17446/17600 (epoch 49), train_loss = 1.262, time/batch = 0.156\n",
            "17447/17600 (epoch 49), train_loss = 1.297, time/batch = 0.162\n",
            "17448/17600 (epoch 49), train_loss = 1.299, time/batch = 0.155\n",
            "17449/17600 (epoch 49), train_loss = 1.266, time/batch = 0.177\n",
            "17450/17600 (epoch 49), train_loss = 1.256, time/batch = 0.180\n",
            "17451/17600 (epoch 49), train_loss = 1.284, time/batch = 0.171\n",
            "17452/17600 (epoch 49), train_loss = 1.317, time/batch = 0.162\n",
            "17453/17600 (epoch 49), train_loss = 1.365, time/batch = 0.162\n",
            "17454/17600 (epoch 49), train_loss = 1.281, time/batch = 0.157\n",
            "17455/17600 (epoch 49), train_loss = 1.282, time/batch = 0.170\n",
            "17456/17600 (epoch 49), train_loss = 1.286, time/batch = 0.172\n",
            "17457/17600 (epoch 49), train_loss = 1.264, time/batch = 0.164\n",
            "17458/17600 (epoch 49), train_loss = 1.309, time/batch = 0.160\n",
            "17459/17600 (epoch 49), train_loss = 1.301, time/batch = 0.164\n",
            "17460/17600 (epoch 49), train_loss = 1.271, time/batch = 0.147\n",
            "17461/17600 (epoch 49), train_loss = 1.343, time/batch = 0.166\n",
            "17462/17600 (epoch 49), train_loss = 1.381, time/batch = 0.172\n",
            "17463/17600 (epoch 49), train_loss = 1.269, time/batch = 0.164\n",
            "17464/17600 (epoch 49), train_loss = 1.288, time/batch = 0.177\n",
            "17465/17600 (epoch 49), train_loss = 1.248, time/batch = 0.151\n",
            "17466/17600 (epoch 49), train_loss = 1.278, time/batch = 0.164\n",
            "17467/17600 (epoch 49), train_loss = 1.260, time/batch = 0.155\n",
            "17468/17600 (epoch 49), train_loss = 1.256, time/batch = 0.184\n",
            "17469/17600 (epoch 49), train_loss = 1.278, time/batch = 0.151\n",
            "17470/17600 (epoch 49), train_loss = 1.349, time/batch = 0.153\n",
            "17471/17600 (epoch 49), train_loss = 1.323, time/batch = 0.150\n",
            "17472/17600 (epoch 49), train_loss = 1.240, time/batch = 0.164\n",
            "17473/17600 (epoch 49), train_loss = 1.283, time/batch = 0.141\n",
            "17474/17600 (epoch 49), train_loss = 1.289, time/batch = 0.156\n",
            "17475/17600 (epoch 49), train_loss = 1.268, time/batch = 0.154\n",
            "17476/17600 (epoch 49), train_loss = 1.301, time/batch = 0.157\n",
            "17477/17600 (epoch 49), train_loss = 1.265, time/batch = 0.156\n",
            "17478/17600 (epoch 49), train_loss = 1.255, time/batch = 0.165\n",
            "17479/17600 (epoch 49), train_loss = 1.318, time/batch = 0.208\n",
            "17480/17600 (epoch 49), train_loss = 1.317, time/batch = 0.279\n",
            "17481/17600 (epoch 49), train_loss = 1.330, time/batch = 0.250\n",
            "17482/17600 (epoch 49), train_loss = 1.248, time/batch = 0.287\n",
            "17483/17600 (epoch 49), train_loss = 1.278, time/batch = 0.275\n",
            "17484/17600 (epoch 49), train_loss = 1.269, time/batch = 0.277\n",
            "17485/17600 (epoch 49), train_loss = 1.338, time/batch = 0.314\n",
            "17486/17600 (epoch 49), train_loss = 1.336, time/batch = 0.254\n",
            "17487/17600 (epoch 49), train_loss = 1.229, time/batch = 0.271\n",
            "17488/17600 (epoch 49), train_loss = 1.321, time/batch = 0.309\n",
            "17489/17600 (epoch 49), train_loss = 1.269, time/batch = 0.286\n",
            "17490/17600 (epoch 49), train_loss = 1.306, time/batch = 0.185\n",
            "17491/17600 (epoch 49), train_loss = 1.298, time/batch = 0.167\n",
            "17492/17600 (epoch 49), train_loss = 1.265, time/batch = 0.172\n",
            "17493/17600 (epoch 49), train_loss = 1.307, time/batch = 0.169\n",
            "17494/17600 (epoch 49), train_loss = 1.299, time/batch = 0.181\n",
            "17495/17600 (epoch 49), train_loss = 1.271, time/batch = 0.181\n",
            "17496/17600 (epoch 49), train_loss = 1.250, time/batch = 0.167\n",
            "17497/17600 (epoch 49), train_loss = 1.227, time/batch = 0.172\n",
            "17498/17600 (epoch 49), train_loss = 1.359, time/batch = 0.166\n",
            "17499/17600 (epoch 49), train_loss = 1.298, time/batch = 0.188\n",
            "17500/17600 (epoch 49), train_loss = 1.290, time/batch = 0.174\n",
            "17501/17600 (epoch 49), train_loss = 1.299, time/batch = 0.169\n",
            "17502/17600 (epoch 49), train_loss = 1.318, time/batch = 0.158\n",
            "17503/17600 (epoch 49), train_loss = 1.325, time/batch = 0.169\n",
            "17504/17600 (epoch 49), train_loss = 1.369, time/batch = 0.173\n",
            "17505/17600 (epoch 49), train_loss = 1.341, time/batch = 0.179\n",
            "17506/17600 (epoch 49), train_loss = 1.293, time/batch = 0.160\n",
            "17507/17600 (epoch 49), train_loss = 1.296, time/batch = 0.162\n",
            "17508/17600 (epoch 49), train_loss = 1.270, time/batch = 0.161\n",
            "17509/17600 (epoch 49), train_loss = 1.281, time/batch = 0.173\n",
            "17510/17600 (epoch 49), train_loss = 1.259, time/batch = 0.164\n",
            "17511/17600 (epoch 49), train_loss = 1.218, time/batch = 0.174\n",
            "17512/17600 (epoch 49), train_loss = 1.269, time/batch = 0.171\n",
            "17513/17600 (epoch 49), train_loss = 1.275, time/batch = 0.154\n",
            "17514/17600 (epoch 49), train_loss = 1.230, time/batch = 0.172\n",
            "17515/17600 (epoch 49), train_loss = 1.233, time/batch = 0.161\n",
            "17516/17600 (epoch 49), train_loss = 1.262, time/batch = 0.178\n",
            "17517/17600 (epoch 49), train_loss = 1.298, time/batch = 0.170\n",
            "17518/17600 (epoch 49), train_loss = 1.219, time/batch = 0.172\n",
            "17519/17600 (epoch 49), train_loss = 1.291, time/batch = 0.147\n",
            "17520/17600 (epoch 49), train_loss = 1.257, time/batch = 0.163\n",
            "17521/17600 (epoch 49), train_loss = 1.287, time/batch = 0.164\n",
            "17522/17600 (epoch 49), train_loss = 1.268, time/batch = 0.168\n",
            "17523/17600 (epoch 49), train_loss = 1.322, time/batch = 0.159\n",
            "17524/17600 (epoch 49), train_loss = 1.287, time/batch = 0.195\n",
            "17525/17600 (epoch 49), train_loss = 1.311, time/batch = 0.150\n",
            "17526/17600 (epoch 49), train_loss = 1.276, time/batch = 0.168\n",
            "17527/17600 (epoch 49), train_loss = 1.282, time/batch = 0.165\n",
            "17528/17600 (epoch 49), train_loss = 1.305, time/batch = 0.156\n",
            "17529/17600 (epoch 49), train_loss = 1.355, time/batch = 0.145\n",
            "17530/17600 (epoch 49), train_loss = 1.328, time/batch = 0.174\n",
            "17531/17600 (epoch 49), train_loss = 1.324, time/batch = 0.158\n",
            "17532/17600 (epoch 49), train_loss = 1.339, time/batch = 0.176\n",
            "17533/17600 (epoch 49), train_loss = 1.352, time/batch = 0.188\n",
            "17534/17600 (epoch 49), train_loss = 1.301, time/batch = 0.183\n",
            "17535/17600 (epoch 49), train_loss = 1.349, time/batch = 0.167\n",
            "17536/17600 (epoch 49), train_loss = 1.284, time/batch = 0.182\n",
            "17537/17600 (epoch 49), train_loss = 1.279, time/batch = 0.158\n",
            "17538/17600 (epoch 49), train_loss = 1.345, time/batch = 0.171\n",
            "17539/17600 (epoch 49), train_loss = 1.249, time/batch = 0.153\n",
            "17540/17600 (epoch 49), train_loss = 1.246, time/batch = 0.161\n",
            "17541/17600 (epoch 49), train_loss = 1.320, time/batch = 0.156\n",
            "17542/17600 (epoch 49), train_loss = 1.267, time/batch = 0.190\n",
            "17543/17600 (epoch 49), train_loss = 1.327, time/batch = 0.154\n",
            "17544/17600 (epoch 49), train_loss = 1.274, time/batch = 0.169\n",
            "17545/17600 (epoch 49), train_loss = 1.300, time/batch = 0.170\n",
            "17546/17600 (epoch 49), train_loss = 1.289, time/batch = 0.167\n",
            "17547/17600 (epoch 49), train_loss = 1.299, time/batch = 0.145\n",
            "17548/17600 (epoch 49), train_loss = 1.329, time/batch = 0.172\n",
            "17549/17600 (epoch 49), train_loss = 1.289, time/batch = 0.192\n",
            "17550/17600 (epoch 49), train_loss = 1.357, time/batch = 0.280\n",
            "17551/17600 (epoch 49), train_loss = 1.347, time/batch = 0.289\n",
            "17552/17600 (epoch 49), train_loss = 1.359, time/batch = 0.287\n",
            "17553/17600 (epoch 49), train_loss = 1.275, time/batch = 0.237\n",
            "17554/17600 (epoch 49), train_loss = 1.307, time/batch = 0.268\n",
            "17555/17600 (epoch 49), train_loss = 1.318, time/batch = 0.272\n",
            "17556/17600 (epoch 49), train_loss = 1.312, time/batch = 0.296\n",
            "17557/17600 (epoch 49), train_loss = 1.315, time/batch = 0.284\n",
            "17558/17600 (epoch 49), train_loss = 1.351, time/batch = 0.287\n",
            "17559/17600 (epoch 49), train_loss = 1.303, time/batch = 0.269\n",
            "17560/17600 (epoch 49), train_loss = 1.336, time/batch = 0.239\n",
            "17561/17600 (epoch 49), train_loss = 1.271, time/batch = 0.157\n",
            "17562/17600 (epoch 49), train_loss = 1.286, time/batch = 0.159\n",
            "17563/17600 (epoch 49), train_loss = 1.258, time/batch = 0.159\n",
            "17564/17600 (epoch 49), train_loss = 1.350, time/batch = 0.151\n",
            "17565/17600 (epoch 49), train_loss = 1.339, time/batch = 0.179\n",
            "17566/17600 (epoch 49), train_loss = 1.335, time/batch = 0.181\n",
            "17567/17600 (epoch 49), train_loss = 1.320, time/batch = 0.172\n",
            "17568/17600 (epoch 49), train_loss = 1.357, time/batch = 0.161\n",
            "17569/17600 (epoch 49), train_loss = 1.323, time/batch = 0.161\n",
            "17570/17600 (epoch 49), train_loss = 1.381, time/batch = 0.165\n",
            "17571/17600 (epoch 49), train_loss = 1.350, time/batch = 0.163\n",
            "17572/17600 (epoch 49), train_loss = 1.338, time/batch = 0.163\n",
            "17573/17600 (epoch 49), train_loss = 1.285, time/batch = 0.159\n",
            "17574/17600 (epoch 49), train_loss = 1.364, time/batch = 0.170\n",
            "17575/17600 (epoch 49), train_loss = 1.339, time/batch = 0.150\n",
            "17576/17600 (epoch 49), train_loss = 1.377, time/batch = 0.173\n",
            "17577/17600 (epoch 49), train_loss = 1.347, time/batch = 0.156\n",
            "17578/17600 (epoch 49), train_loss = 1.265, time/batch = 0.164\n",
            "17579/17600 (epoch 49), train_loss = 1.296, time/batch = 0.170\n",
            "17580/17600 (epoch 49), train_loss = 1.340, time/batch = 0.153\n",
            "17581/17600 (epoch 49), train_loss = 1.354, time/batch = 0.159\n",
            "17582/17600 (epoch 49), train_loss = 1.311, time/batch = 0.165\n",
            "17583/17600 (epoch 49), train_loss = 1.305, time/batch = 0.152\n",
            "17584/17600 (epoch 49), train_loss = 1.306, time/batch = 0.167\n",
            "17585/17600 (epoch 49), train_loss = 1.350, time/batch = 0.167\n",
            "17586/17600 (epoch 49), train_loss = 1.296, time/batch = 0.171\n",
            "17587/17600 (epoch 49), train_loss = 1.322, time/batch = 0.166\n",
            "17588/17600 (epoch 49), train_loss = 1.293, time/batch = 0.166\n",
            "17589/17600 (epoch 49), train_loss = 1.337, time/batch = 0.162\n",
            "17590/17600 (epoch 49), train_loss = 1.341, time/batch = 0.167\n",
            "17591/17600 (epoch 49), train_loss = 1.305, time/batch = 0.170\n",
            "17592/17600 (epoch 49), train_loss = 1.282, time/batch = 0.173\n",
            "17593/17600 (epoch 49), train_loss = 1.303, time/batch = 0.150\n",
            "17594/17600 (epoch 49), train_loss = 1.326, time/batch = 0.166\n",
            "17595/17600 (epoch 49), train_loss = 1.300, time/batch = 0.159\n",
            "17596/17600 (epoch 49), train_loss = 1.322, time/batch = 0.155\n",
            "17597/17600 (epoch 49), train_loss = 1.309, time/batch = 0.152\n",
            "17598/17600 (epoch 49), train_loss = 1.300, time/batch = 0.173\n",
            "17599/17600 (epoch 49), train_loss = 1.292, time/batch = 0.148\n",
            "Model saved to checkpoints/dracula/dracula!\n",
            "Converting model to ml5js: dracula dracula-17599\n",
            "Done! The output model is in models\n",
            "Check https://ml5js.org/docs/training-lstm for more information.\n"
          ]
        }
      ],
      "source": [
        "# entrenar con archivo quijote.txt\n",
        "!export PYENV_ROOT=\"$HOME/.pyenv\" && export PATH=\"$PYENV_ROOT/bin:$PATH\" && eval \"$(pyenv init --path)\" && python train.py --data_path ./dracula.txt"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "oVJ1_7E1sfVY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8ededff6-646f-4dea-8007-bdb15e074083"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "  adding: models/dracula/ (stored 0%)\n",
            "  adding: models/dracula/embedding (deflated 7%)\n",
            "  adding: models/dracula/rnnlm_multi_rnn_cell_cell_0_basic_lstm_cell_bias (deflated 6%)\n",
            "  adding: models/dracula/Variable (stored 0%)\n",
            "  adding: models/dracula/rnnlm_multi_rnn_cell_cell_1_basic_lstm_cell_bias (deflated 5%)\n",
            "  adding: models/dracula/rnnlm_multi_rnn_cell_cell_1_basic_lstm_cell_kernel (deflated 7%)\n",
            "  adding: models/dracula/rnnlm_multi_rnn_cell_cell_0_basic_lstm_cell_kernel (deflated 7%)\n",
            "  adding: models/dracula/vocab.json (deflated 63%)\n",
            "  adding: models/dracula/rnnlm_softmax_w (deflated 7%)\n",
            "  adding: models/dracula/rnnlm_softmax_b (stored 0%)\n",
            "  adding: models/dracula/manifest.json (deflated 80%)\n"
          ]
        }
      ],
      "source": [
        "# comprimir modelo en .zip\n",
        "!zip -r ./modelo.zip ./models/dracula\n",
        "# refrescar disco duro, click derecho en modelo.zip y descargar\n",
        "# después recomiendo borrar el archivo, por si repetimos este paso con otro corpus"
      ]
    }
  ],
  "metadata": {
    "colab": {
      "provenance": [],
      "authorship_tag": "ABX9TyPje8lsrdbLfZfCo+9YExB+",
      "include_colab_link": true
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}