google_colab_fall_detection.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "google-colab-fall-detection",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7Xubqkl2b0WA",
        "colab_type": "text"
      },
      "source": [
        "## Install Data and Flair"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ClpwKxtxa6iA",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 170
        },
        "outputId": "2e51bb06-31eb-42a4-d5ce-4fc7c8c59846"
      },
      "source": [
        "!git clone https://github.com/nlp-pucrs/fall-detection -q\n",
        "!pip install flair -q"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\u001b[K     |████████████████████████████████| 143kB 9.4MB/s \n",
            "\u001b[K     |████████████████████████████████| 133kB 52.5MB/s \n",
            "\u001b[K     |████████████████████████████████| 655kB 53.8MB/s \n",
            "\u001b[K     |████████████████████████████████| 798kB 52.4MB/s \n",
            "\u001b[K     |████████████████████████████████| 1.0MB 28.6MB/s \n",
            "\u001b[?25h  Building wheel for sqlitedict (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Building wheel for regex (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Building wheel for mpld3 (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Building wheel for segtok (setup.py) ... \u001b[?25l\u001b[?25hdone\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "G9tQDaBTb6Og",
        "colab_type": "text"
      },
      "source": [
        "## Load Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l1XUe8EUbYVA",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 374
        },
        "outputId": "4d35248f-348b-446a-8170-1ad1a6880352"
      },
      "source": [
        "import pandas as pd\n",
        "\n",
        "quedas = pd.read_csv('fall-detection/training_data.csv')\n",
        "\n",
        "quedas.shape, quedas.columns"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((1078, 34), Index(['Unnamed: 0', 'Evolucao', 'Target',\n",
              "        'Condições do paciente antes da queda',\n",
              "        'Dano não classificado adequadamente', 'Dúvida para Janete',\n",
              "        'Fator de Risco - Comportamental', 'Fator de Risco - Extrínseco',\n",
              "        'Fator de Risco - Intrínseco',\n",
              "        'Foi necessário algum procedimento. Qual?', 'Grau do Dano - Grave',\n",
              "        'Grau do Dano - Leve', 'Grau do Dano - Moderado',\n",
              "        'Grau do Dano - Nenhum', 'Grau do Dano - Óbito',\n",
              "        'Necessitou de avaliação médica',\n",
              "        'Orientações para prevenção ANTES da queda?',\n",
              "        'Orientações para prevenção DEPOIS da queda?', 'Outro setor: Qual?',\n",
              "        'Paciente acompanhado', 'Paciente não acompanhado',\n",
              "        'Queda não classificada', 'Registro da MORSE',\n",
              "        'Se foi na unidade de internação. Qual local?',\n",
              "        'Tipo de Queda I - Acidental',\n",
              "        'Tipo de Queda I - Fisiológica Antecipada',\n",
              "        'Tipo de Queda I - Não Antecipada', 'Tipo de Queda II - Ao Solo',\n",
              "        'Tipo de Queda II - Nível inferior', 'Tipo de Queda III - Outro',\n",
              "        'Tipo de Queda III - Própria Altura', 'Tipo do Dano - Físico',\n",
              "        'Tipo do Dano - Psicológico', 'Tipo do Dano - Social'],\n",
              "       dtype='object'))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 2
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EouliGSfcgOg",
        "colab_type": "text"
      },
      "source": [
        "## Load Flair Sentences"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UeUkFN0Wbteq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from flair.data import Sentence\n",
        "\n",
        "sentences = []\n",
        "\n",
        "for i in quedas.index:\n",
        "  sentence = sentence = Sentence(quedas['Evolucao'][i], use_tokenizer=True)\n",
        "  sentence.add_label('queda', str(quedas['Target'][i]))\n",
        "  sentences.append(sentence)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ydDUxDBwdMHt",
        "colab_type": "text"
      },
      "source": [
        "## Load NILC Embeddgins"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BI-mC2HCcYA8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 156
        },
        "outputId": "9ab935ce-dae6-448f-e0b5-3e8086606c4a"
      },
      "source": [
        "from flair.embeddings import NILCEmbeddings\n",
        "\n",
        "nilc_embedding = NILCEmbeddings('fasttext')"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2019-07-23 19:45:52,946 http:https://143.107.183.175:22980/download.php?file=embeddings/fasttext/skip_s100.zip not found in cache, downloading to /tmp/tmpcgnlyp61\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 313991848/313991848 [19:22<00:00, 270055.52B/s]"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "2019-07-23 20:05:16,096 copying /tmp/tmpcgnlyp61 to cache at /root/.flair/embeddings/fasttext/skip_s100.zip\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "2019-07-23 20:05:16,720 removing temp file /tmp/tmpcgnlyp61\n",
            "2019-07-23 20:05:16,770 Reading embeddings from /root/.flair/embeddings/fasttext/skip_s100.zip\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/smart_open/smart_open_lib.py:398: UserWarning: This function is deprecated, use smart_open.open instead. See the migration notes for details: https://github.com/RaRe-Technologies/smart_open/blob/master/README.rst#migrating-to-the-new-open-function\n",
            "  'See the migration notes for details: %s' % _MIGRATION_NOTES_URL\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YaQ1Q0z6dTqZ",
        "colab_type": "text"
      },
      "source": [
        "## Train the Model\n",
        "based on https://github.com/zalandoresearch/flair/blob/master/resources/docs/TUTORIAL_7_TRAINING_A_MODEL.md"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HogsPhH0dzYb",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "1079b064-d3ef-4c8c-aab4-4bd3a9ecb395"
      },
      "source": [
        "from flair.data import Corpus\n",
        "from flair.embeddings import DocumentRNNEmbeddings\n",
        "from flair.models import TextClassifier\n",
        "from flair.trainers import ModelTrainer\n",
        "\n",
        "# 1. get the corpus\n",
        "corpus = Corpus(sentences[:1000], sentences[1000:], sentences[1000:]) \n",
        "\n",
        "# 2. create the label dictionary\n",
        "label_dict = corpus.make_label_dictionary()\n",
        "\n",
        "# 3. make a list of word embeddings\n",
        "word_embeddings = [nilc_embedding]\n",
        "\n",
        "# 4. initialize document embedding by passing list of word embeddings\n",
        "document_embeddings: DocumentRNNEmbeddings = DocumentRNNEmbeddings(word_embeddings,\n",
        "                                                                     hidden_size=512,\n",
        "                                                                     reproject_words=True,\n",
        "                                                                     reproject_words_dimension=256,\n",
        "                                                                     )\n",
        "\n",
        "# 5. create the text classifier\n",
        "classifier = TextClassifier(document_embeddings, label_dictionary=label_dict)\n",
        "\n",
        "# 6. initialize the text classifier trainer\n",
        "trainer = ModelTrainer(classifier, corpus)\n",
        "\n",
        "# 7. start the training\n",
        "trainer.train('fall-train/',\n",
        "              learning_rate=0.1,\n",
        "              mini_batch_size=32,\n",
        "              anneal_factor=0.5,\n",
        "              patience=5,\n",
        "              max_epochs=150)"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "2019-07-23 20:06:53,537 {'1', '0'}\n",
            "2019-07-23 20:07:00,368 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:07:00,369 Evaluation method: MICRO_F1_SCORE\n",
            "2019-07-23 20:07:00,722 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:07:01,661 epoch 1 - iter 0/32 - loss 0.71004659\n",
            "2019-07-23 20:07:02,625 epoch 1 - iter 3/32 - loss 0.67612775\n",
            "2019-07-23 20:07:04,262 epoch 1 - iter 6/32 - loss 0.63835382\n",
            "2019-07-23 20:07:05,054 epoch 1 - iter 9/32 - loss 0.64605298\n",
            "2019-07-23 20:07:05,910 epoch 1 - iter 12/32 - loss 0.63523429\n",
            "2019-07-23 20:07:06,692 epoch 1 - iter 15/32 - loss 0.64211859\n",
            "2019-07-23 20:07:07,400 epoch 1 - iter 18/32 - loss 0.64011177\n",
            "2019-07-23 20:07:08,790 epoch 1 - iter 21/32 - loss 0.63915680\n",
            "2019-07-23 20:07:09,650 epoch 1 - iter 24/32 - loss 0.63274335\n",
            "2019-07-23 20:07:10,544 epoch 1 - iter 27/32 - loss 0.63265841\n",
            "2019-07-23 20:07:11,361 epoch 1 - iter 30/32 - loss 0.63256536\n",
            "2019-07-23 20:07:11,653 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:07:11,656 EPOCH 1 done: loss 0.6307 - lr 0.1000 - bad epochs 0\n",
            "2019-07-23 20:07:12,762 DEV : loss 0.6338959336280823 - score 0.6795\n",
            "2019-07-23 20:07:14,752 TEST : loss 0.6338959336280823 - score 0.6795\n",
            "2019-07-23 20:07:23,843 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:07:24,816 epoch 2 - iter 0/32 - loss 0.69986916\n",
            "2019-07-23 20:07:25,833 epoch 2 - iter 3/32 - loss 0.67607388\n",
            "2019-07-23 20:07:26,653 epoch 2 - iter 6/32 - loss 0.66171655\n",
            "2019-07-23 20:07:27,612 epoch 2 - iter 9/32 - loss 0.64544104\n",
            "2019-07-23 20:07:29,475 epoch 2 - iter 12/32 - loss 0.62086360\n",
            "2019-07-23 20:07:30,326 epoch 2 - iter 15/32 - loss 0.62573059\n",
            "2019-07-23 20:07:31,189 epoch 2 - iter 18/32 - loss 0.62928702\n",
            "2019-07-23 20:07:32,101 epoch 2 - iter 21/32 - loss 0.62904267\n",
            "2019-07-23 20:07:32,813 epoch 2 - iter 24/32 - loss 0.62871893\n",
            "2019-07-23 20:07:34,245 epoch 2 - iter 27/32 - loss 0.62172181\n",
            "2019-07-23 20:07:35,002 epoch 2 - iter 30/32 - loss 0.62657407\n",
            "2019-07-23 20:07:35,335 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:07:35,338 EPOCH 2 done: loss 0.6268 - lr 0.1000 - bad epochs 0\n",
            "2019-07-23 20:07:36,529 DEV : loss 0.6206819415092468 - score 0.6795\n",
            "2019-07-23 20:07:37,787 TEST : loss 0.6206819415092468 - score 0.6795\n",
            "2019-07-23 20:07:46,673 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:07:47,625 epoch 3 - iter 0/32 - loss 0.52789903\n",
            "2019-07-23 20:07:49,552 epoch 3 - iter 3/32 - loss 0.64780125\n",
            "2019-07-23 20:07:50,384 epoch 3 - iter 6/32 - loss 0.64204093\n",
            "2019-07-23 20:07:51,170 epoch 3 - iter 9/32 - loss 0.62954110\n",
            "2019-07-23 20:07:52,014 epoch 3 - iter 12/32 - loss 0.61830426\n",
            "2019-07-23 20:07:52,897 epoch 3 - iter 15/32 - loss 0.62174308\n",
            "2019-07-23 20:07:53,700 epoch 3 - iter 18/32 - loss 0.61946702\n",
            "2019-07-23 20:07:55,322 epoch 3 - iter 21/32 - loss 0.61254731\n",
            "2019-07-23 20:07:56,274 epoch 3 - iter 24/32 - loss 0.61227872\n",
            "2019-07-23 20:07:56,966 epoch 3 - iter 27/32 - loss 0.61453323\n",
            "2019-07-23 20:07:57,778 epoch 3 - iter 30/32 - loss 0.61620996\n",
            "2019-07-23 20:07:58,128 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:07:58,130 EPOCH 3 done: loss 0.6114 - lr 0.1000 - bad epochs 1\n",
            "2019-07-23 20:07:59,306 DEV : loss 0.6689352989196777 - score 0.6795\n",
            "2019-07-23 20:08:01,364 TEST : loss 0.6689352989196777 - score 0.6795\n",
            "2019-07-23 20:08:10,233 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:08:11,423 epoch 4 - iter 0/32 - loss 0.74675107\n",
            "2019-07-23 20:08:12,550 epoch 4 - iter 3/32 - loss 0.56249788\n",
            "2019-07-23 20:08:13,302 epoch 4 - iter 6/32 - loss 0.56832833\n",
            "2019-07-23 20:08:14,295 epoch 4 - iter 9/32 - loss 0.59157046\n",
            "2019-07-23 20:08:15,881 epoch 4 - iter 12/32 - loss 0.59245978\n",
            "2019-07-23 20:08:16,835 epoch 4 - iter 15/32 - loss 0.60527283\n",
            "2019-07-23 20:08:17,817 epoch 4 - iter 18/32 - loss 0.59478760\n",
            "2019-07-23 20:08:18,552 epoch 4 - iter 21/32 - loss 0.60547816\n",
            "2019-07-23 20:08:19,425 epoch 4 - iter 24/32 - loss 0.60352654\n",
            "2019-07-23 20:08:20,749 epoch 4 - iter 27/32 - loss 0.60720855\n",
            "2019-07-23 20:08:21,516 epoch 4 - iter 30/32 - loss 0.60700629\n",
            "2019-07-23 20:08:21,882 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:08:21,883 EPOCH 4 done: loss 0.6068 - lr 0.1000 - bad epochs 2\n",
            "2019-07-23 20:08:23,093 DEV : loss 0.5677545070648193 - score 0.7436\n",
            "2019-07-23 20:08:24,341 TEST : loss 0.5677545070648193 - score 0.7436\n",
            "2019-07-23 20:08:33,118 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:08:34,184 epoch 5 - iter 0/32 - loss 0.64574754\n",
            "2019-07-23 20:08:35,053 epoch 5 - iter 3/32 - loss 0.64047869\n",
            "2019-07-23 20:08:36,646 epoch 5 - iter 6/32 - loss 0.60521707\n",
            "2019-07-23 20:08:37,539 epoch 5 - iter 9/32 - loss 0.59019954\n",
            "2019-07-23 20:08:38,450 epoch 5 - iter 12/32 - loss 0.56802624\n",
            "2019-07-23 20:08:39,259 epoch 5 - iter 15/32 - loss 0.57199643\n",
            "2019-07-23 20:08:40,071 epoch 5 - iter 18/32 - loss 0.58619550\n",
            "2019-07-23 20:08:41,628 epoch 5 - iter 21/32 - loss 0.59172706\n",
            "2019-07-23 20:08:42,537 epoch 5 - iter 24/32 - loss 0.59139432\n",
            "2019-07-23 20:08:43,280 epoch 5 - iter 27/32 - loss 0.58259192\n",
            "2019-07-23 20:08:44,203 epoch 5 - iter 30/32 - loss 0.57991635\n",
            "2019-07-23 20:08:44,515 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:08:44,517 EPOCH 5 done: loss 0.5916 - lr 0.1000 - bad epochs 0\n",
            "2019-07-23 20:08:45,698 DEV : loss 0.6289049983024597 - score 0.7949\n",
            "2019-07-23 20:08:47,753 TEST : loss 0.6289049983024597 - score 0.7949\n",
            "2019-07-23 20:08:56,649 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:08:57,576 epoch 6 - iter 0/32 - loss 0.64872706\n",
            "2019-07-23 20:08:58,553 epoch 6 - iter 3/32 - loss 0.54973806\n",
            "2019-07-23 20:08:59,608 epoch 6 - iter 6/32 - loss 0.60032598\n",
            "2019-07-23 20:09:00,665 epoch 6 - iter 9/32 - loss 0.59764549\n",
            "2019-07-23 20:09:01,549 epoch 6 - iter 12/32 - loss 0.58652295\n",
            "2019-07-23 20:09:03,015 epoch 6 - iter 15/32 - loss 0.58111814\n",
            "2019-07-23 20:09:04,017 epoch 6 - iter 18/32 - loss 0.58156097\n",
            "2019-07-23 20:09:04,799 epoch 6 - iter 21/32 - loss 0.59223349\n",
            "2019-07-23 20:09:05,586 epoch 6 - iter 24/32 - loss 0.59481783\n",
            "2019-07-23 20:09:06,341 epoch 6 - iter 27/32 - loss 0.59991587\n",
            "2019-07-23 20:09:07,819 epoch 6 - iter 30/32 - loss 0.59642676\n",
            "2019-07-23 20:09:08,149 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:09:08,150 EPOCH 6 done: loss 0.5912 - lr 0.1000 - bad epochs 0\n",
            "2019-07-23 20:09:09,365 DEV : loss 0.5476975440979004 - score 0.6795\n",
            "2019-07-23 20:09:10,633 TEST : loss 0.5476975440979004 - score 0.6795\n",
            "2019-07-23 20:09:10,635 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:09:11,613 epoch 7 - iter 0/32 - loss 0.47415152\n",
            "2019-07-23 20:09:12,774 epoch 7 - iter 3/32 - loss 0.54030324\n",
            "2019-07-23 20:09:14,360 epoch 7 - iter 6/32 - loss 0.58167473\n",
            "2019-07-23 20:09:15,331 epoch 7 - iter 9/32 - loss 0.57599611\n",
            "2019-07-23 20:09:16,125 epoch 7 - iter 12/32 - loss 0.57468819\n",
            "2019-07-23 20:09:16,987 epoch 7 - iter 15/32 - loss 0.56920500\n",
            "2019-07-23 20:09:17,860 epoch 7 - iter 18/32 - loss 0.55620252\n",
            "2019-07-23 20:09:19,348 epoch 7 - iter 21/32 - loss 0.55688237\n",
            "2019-07-23 20:09:20,147 epoch 7 - iter 24/32 - loss 0.55740715\n",
            "2019-07-23 20:09:21,042 epoch 7 - iter 27/32 - loss 0.55964127\n",
            "2019-07-23 20:09:21,888 epoch 7 - iter 30/32 - loss 0.56305387\n",
            "2019-07-23 20:09:22,220 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:09:22,226 EPOCH 7 done: loss 0.5646 - lr 0.1000 - bad epochs 1\n",
            "2019-07-23 20:09:23,443 DEV : loss 0.7135956287384033 - score 0.6795\n",
            "2019-07-23 20:09:24,693 TEST : loss 0.7135956287384033 - score 0.6795\n",
            "2019-07-23 20:09:24,696 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:09:26,572 epoch 8 - iter 0/32 - loss 0.79271895\n",
            "2019-07-23 20:09:27,459 epoch 8 - iter 3/32 - loss 0.60562849\n",
            "2019-07-23 20:09:28,451 epoch 8 - iter 6/32 - loss 0.58365655\n",
            "2019-07-23 20:09:29,366 epoch 8 - iter 9/32 - loss 0.59688060\n",
            "2019-07-23 20:09:30,849 epoch 8 - iter 12/32 - loss 0.58574643\n",
            "2019-07-23 20:09:31,709 epoch 8 - iter 15/32 - loss 0.57454839\n",
            "2019-07-23 20:09:32,800 epoch 8 - iter 18/32 - loss 0.57120573\n",
            "2019-07-23 20:09:33,577 epoch 8 - iter 21/32 - loss 0.55043093\n",
            "2019-07-23 20:09:34,459 epoch 8 - iter 24/32 - loss 0.55841713\n",
            "2019-07-23 20:09:35,323 epoch 8 - iter 27/32 - loss 0.57086847\n",
            "2019-07-23 20:09:36,056 epoch 8 - iter 30/32 - loss 0.56644372\n",
            "2019-07-23 20:09:36,446 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:09:36,448 EPOCH 8 done: loss 0.5597 - lr 0.1000 - bad epochs 2\n",
            "2019-07-23 20:09:38,680 DEV : loss 0.5339107513427734 - score 0.8333\n",
            "2019-07-23 20:09:39,962 TEST : loss 0.5339107513427734 - score 0.8333\n",
            "2019-07-23 20:09:48,778 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:09:49,896 epoch 9 - iter 0/32 - loss 0.49634349\n",
            "2019-07-23 20:09:50,697 epoch 9 - iter 3/32 - loss 0.53307544\n",
            "2019-07-23 20:09:51,699 epoch 9 - iter 6/32 - loss 0.50302202\n",
            "2019-07-23 20:09:53,274 epoch 9 - iter 9/32 - loss 0.50529843\n",
            "2019-07-23 20:09:54,179 epoch 9 - iter 12/32 - loss 0.53346646\n",
            "2019-07-23 20:09:55,073 epoch 9 - iter 15/32 - loss 0.52662603\n",
            "2019-07-23 20:09:55,899 epoch 9 - iter 18/32 - loss 0.53635887\n",
            "2019-07-23 20:09:56,766 epoch 9 - iter 21/32 - loss 0.53751146\n",
            "2019-07-23 20:09:58,163 epoch 9 - iter 24/32 - loss 0.53681062\n",
            "2019-07-23 20:09:58,936 epoch 9 - iter 27/32 - loss 0.54281240\n",
            "2019-07-23 20:09:59,784 epoch 9 - iter 30/32 - loss 0.54542807\n",
            "2019-07-23 20:10:00,109 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:10:00,111 EPOCH 9 done: loss 0.5494 - lr 0.1000 - bad epochs 0\n",
            "2019-07-23 20:10:01,308 DEV : loss 0.5685648918151855 - score 0.6795\n",
            "2019-07-23 20:10:02,563 TEST : loss 0.5685648918151855 - score 0.6795\n",
            "2019-07-23 20:10:02,566 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:10:03,454 epoch 10 - iter 0/32 - loss 0.58036983\n",
            "2019-07-23 20:10:05,147 epoch 10 - iter 3/32 - loss 0.57742582\n",
            "2019-07-23 20:10:06,102 epoch 10 - iter 6/32 - loss 0.56798823\n",
            "2019-07-23 20:10:06,964 epoch 10 - iter 9/32 - loss 0.59131569\n",
            "2019-07-23 20:10:07,748 epoch 10 - iter 12/32 - loss 0.55884322\n",
            "2019-07-23 20:10:08,669 epoch 10 - iter 15/32 - loss 0.53842559\n",
            "2019-07-23 20:10:10,013 epoch 10 - iter 18/32 - loss 0.54368465\n",
            "2019-07-23 20:10:10,817 epoch 10 - iter 21/32 - loss 0.54889446\n",
            "2019-07-23 20:10:11,805 epoch 10 - iter 24/32 - loss 0.55401478\n",
            "2019-07-23 20:10:12,611 epoch 10 - iter 27/32 - loss 0.55816574\n",
            "2019-07-23 20:10:13,436 epoch 10 - iter 30/32 - loss 0.56201489\n",
            "2019-07-23 20:10:13,761 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:10:13,763 EPOCH 10 done: loss 0.5638 - lr 0.1000 - bad epochs 1\n",
            "2019-07-23 20:10:15,813 DEV : loss 0.6335679888725281 - score 0.6795\n",
            "2019-07-23 20:10:17,076 TEST : loss 0.6335679888725281 - score 0.6795\n",
            "2019-07-23 20:10:17,078 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:10:18,049 epoch 11 - iter 0/32 - loss 0.64477175\n",
            "2019-07-23 20:10:19,042 epoch 11 - iter 3/32 - loss 0.54383688\n",
            "2019-07-23 20:10:19,867 epoch 11 - iter 6/32 - loss 0.57440974\n",
            "2019-07-23 20:10:21,529 epoch 11 - iter 9/32 - loss 0.58662651\n",
            "2019-07-23 20:10:22,271 epoch 11 - iter 12/32 - loss 0.57240277\n",
            "2019-07-23 20:10:23,289 epoch 11 - iter 15/32 - loss 0.57141232\n",
            "2019-07-23 20:10:24,086 epoch 11 - iter 18/32 - loss 0.57044202\n",
            "2019-07-23 20:10:25,012 epoch 11 - iter 21/32 - loss 0.55636582\n",
            "2019-07-23 20:10:26,470 epoch 11 - iter 24/32 - loss 0.56734114\n",
            "2019-07-23 20:10:27,237 epoch 11 - iter 27/32 - loss 0.56797000\n",
            "2019-07-23 20:10:27,871 epoch 11 - iter 30/32 - loss 0.56799287\n",
            "2019-07-23 20:10:28,206 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:10:28,208 EPOCH 11 done: loss 0.5635 - lr 0.1000 - bad epochs 2\n",
            "2019-07-23 20:10:29,436 DEV : loss 0.5892531275749207 - score 0.6795\n",
            "2019-07-23 20:10:30,710 TEST : loss 0.5892531275749207 - score 0.6795\n",
            "2019-07-23 20:10:30,712 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:10:31,576 epoch 12 - iter 0/32 - loss 0.51080376\n",
            "2019-07-23 20:10:33,516 epoch 12 - iter 3/32 - loss 0.59423386\n",
            "2019-07-23 20:10:34,556 epoch 12 - iter 6/32 - loss 0.55835158\n",
            "2019-07-23 20:10:35,519 epoch 12 - iter 9/32 - loss 0.54534706\n",
            "2019-07-23 20:10:36,245 epoch 12 - iter 12/32 - loss 0.53448205\n",
            "2019-07-23 20:10:37,180 epoch 12 - iter 15/32 - loss 0.52635790\n",
            "2019-07-23 20:10:37,984 epoch 12 - iter 18/32 - loss 0.52234287\n",
            "2019-07-23 20:10:39,477 epoch 12 - iter 21/32 - loss 0.52272148\n",
            "2019-07-23 20:10:40,321 epoch 12 - iter 24/32 - loss 0.53050391\n",
            "2019-07-23 20:10:41,134 epoch 12 - iter 27/32 - loss 0.53794200\n",
            "2019-07-23 20:10:42,046 epoch 12 - iter 30/32 - loss 0.53550404\n",
            "2019-07-23 20:10:42,363 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:10:42,365 EPOCH 12 done: loss 0.5443 - lr 0.1000 - bad epochs 3\n",
            "2019-07-23 20:10:43,576 DEV : loss 0.5735622644424438 - score 0.6795\n",
            "2019-07-23 20:10:45,647 TEST : loss 0.5735622644424438 - score 0.6795\n",
            "2019-07-23 20:10:45,648 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:10:46,726 epoch 13 - iter 0/32 - loss 0.60388803\n",
            "2019-07-23 20:10:47,732 epoch 13 - iter 3/32 - loss 0.61491420\n",
            "2019-07-23 20:10:48,440 epoch 13 - iter 6/32 - loss 0.57641701\n",
            "2019-07-23 20:10:49,411 epoch 13 - iter 9/32 - loss 0.56470824\n",
            "2019-07-23 20:10:50,989 epoch 13 - iter 12/32 - loss 0.57381948\n",
            "2019-07-23 20:10:51,772 epoch 13 - iter 15/32 - loss 0.55132900\n",
            "2019-07-23 20:10:52,815 epoch 13 - iter 18/32 - loss 0.55199315\n",
            "2019-07-23 20:10:53,533 epoch 13 - iter 21/32 - loss 0.55528490\n",
            "2019-07-23 20:10:54,539 epoch 13 - iter 24/32 - loss 0.55515980\n",
            "2019-07-23 20:10:56,025 epoch 13 - iter 27/32 - loss 0.54794996\n",
            "2019-07-23 20:10:56,679 epoch 13 - iter 30/32 - loss 0.54613523\n",
            "2019-07-23 20:10:57,040 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:10:57,042 EPOCH 13 done: loss 0.5496 - lr 0.1000 - bad epochs 4\n",
            "2019-07-23 20:10:58,241 DEV : loss 0.5254485607147217 - score 0.8333\n",
            "2019-07-23 20:10:59,465 TEST : loss 0.5254485607147217 - score 0.8333\n",
            "2019-07-23 20:11:08,442 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:11:09,471 epoch 14 - iter 0/32 - loss 0.55614018\n",
            "2019-07-23 20:11:10,373 epoch 14 - iter 3/32 - loss 0.46837613\n",
            "2019-07-23 20:11:12,112 epoch 14 - iter 6/32 - loss 0.52527947\n",
            "2019-07-23 20:11:12,936 epoch 14 - iter 9/32 - loss 0.53075786\n",
            "2019-07-23 20:11:13,823 epoch 14 - iter 12/32 - loss 0.53582021\n",
            "2019-07-23 20:11:14,796 epoch 14 - iter 15/32 - loss 0.52552495\n",
            "2019-07-23 20:11:16,261 epoch 14 - iter 18/32 - loss 0.53398511\n",
            "2019-07-23 20:11:17,010 epoch 14 - iter 21/32 - loss 0.53643331\n",
            "2019-07-23 20:11:17,862 epoch 14 - iter 24/32 - loss 0.54080065\n",
            "2019-07-23 20:11:18,607 epoch 14 - iter 27/32 - loss 0.54424649\n",
            "2019-07-23 20:11:19,276 epoch 14 - iter 30/32 - loss 0.54404707\n",
            "2019-07-23 20:11:19,601 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:11:19,604 EPOCH 14 done: loss 0.5447 - lr 0.1000 - bad epochs 5\n",
            "2019-07-23 20:11:20,807 DEV : loss 0.546250581741333 - score 0.8333\n",
            "2019-07-23 20:11:22,932 TEST : loss 0.546250581741333 - score 0.8333\n",
            "Epoch    13: reducing learning rate of group 0 to 5.0000e-02.\n",
            "2019-07-23 20:11:31,785 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:11:32,796 epoch 15 - iter 0/32 - loss 0.63517576\n",
            "2019-07-23 20:11:33,715 epoch 15 - iter 3/32 - loss 0.52473752\n",
            "2019-07-23 20:11:34,581 epoch 15 - iter 6/32 - loss 0.54176153\n",
            "2019-07-23 20:11:35,587 epoch 15 - iter 9/32 - loss 0.54310924\n",
            "2019-07-23 20:11:36,451 epoch 15 - iter 12/32 - loss 0.52822273\n",
            "2019-07-23 20:11:38,112 epoch 15 - iter 15/32 - loss 0.53676596\n",
            "2019-07-23 20:11:38,996 epoch 15 - iter 18/32 - loss 0.53165095\n",
            "2019-07-23 20:11:39,925 epoch 15 - iter 21/32 - loss 0.53806276\n",
            "2019-07-23 20:11:40,724 epoch 15 - iter 24/32 - loss 0.54065084\n",
            "2019-07-23 20:11:41,565 epoch 15 - iter 27/32 - loss 0.53433631\n",
            "2019-07-23 20:11:43,040 epoch 15 - iter 30/32 - loss 0.53804193\n",
            "2019-07-23 20:11:43,371 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:11:43,375 EPOCH 15 done: loss 0.5309 - lr 0.0500 - bad epochs 0\n",
            "2019-07-23 20:11:44,600 DEV : loss 0.5204155445098877 - score 0.8333\n",
            "2019-07-23 20:11:45,873 TEST : loss 0.5204155445098877 - score 0.8333\n",
            "2019-07-23 20:11:54,904 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:11:55,931 epoch 16 - iter 0/32 - loss 0.61583143\n",
            "2019-07-23 20:11:56,839 epoch 16 - iter 3/32 - loss 0.55397029\n",
            "2019-07-23 20:11:57,814 epoch 16 - iter 6/32 - loss 0.54178643\n",
            "2019-07-23 20:11:59,687 epoch 16 - iter 9/32 - loss 0.54118463\n",
            "2019-07-23 20:12:00,597 epoch 16 - iter 12/32 - loss 0.54602877\n",
            "2019-07-23 20:12:01,462 epoch 16 - iter 15/32 - loss 0.55135366\n",
            "2019-07-23 20:12:02,386 epoch 16 - iter 18/32 - loss 0.54593627\n",
            "2019-07-23 20:12:03,265 epoch 16 - iter 21/32 - loss 0.54117991\n",
            "2019-07-23 20:12:04,839 epoch 16 - iter 24/32 - loss 0.53175862\n",
            "2019-07-23 20:12:05,465 epoch 16 - iter 27/32 - loss 0.53389468\n",
            "2019-07-23 20:12:06,249 epoch 16 - iter 30/32 - loss 0.53458461\n",
            "2019-07-23 20:12:06,577 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:12:06,579 EPOCH 16 done: loss 0.5433 - lr 0.0500 - bad epochs 1\n",
            "2019-07-23 20:12:07,844 DEV : loss 0.5240322351455688 - score 0.8333\n",
            "2019-07-23 20:12:09,938 TEST : loss 0.5240322351455688 - score 0.8333\n",
            "2019-07-23 20:12:18,603 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:12:19,821 epoch 17 - iter 0/32 - loss 0.50203288\n",
            "2019-07-23 20:12:20,783 epoch 17 - iter 3/32 - loss 0.56525922\n",
            "2019-07-23 20:12:21,601 epoch 17 - iter 6/32 - loss 0.56661007\n",
            "2019-07-23 20:12:22,703 epoch 17 - iter 9/32 - loss 0.52684913\n",
            "2019-07-23 20:12:23,456 epoch 17 - iter 12/32 - loss 0.52147623\n",
            "2019-07-23 20:12:25,025 epoch 17 - iter 15/32 - loss 0.54118754\n",
            "2019-07-23 20:12:25,932 epoch 17 - iter 18/32 - loss 0.56331918\n",
            "2019-07-23 20:12:26,718 epoch 17 - iter 21/32 - loss 0.56169129\n",
            "2019-07-23 20:12:27,655 epoch 17 - iter 24/32 - loss 0.56286809\n",
            "2019-07-23 20:12:28,333 epoch 17 - iter 27/32 - loss 0.54597220\n",
            "2019-07-23 20:12:29,740 epoch 17 - iter 30/32 - loss 0.54427193\n",
            "2019-07-23 20:12:30,110 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:12:30,114 EPOCH 17 done: loss 0.5352 - lr 0.0500 - bad epochs 2\n",
            "2019-07-23 20:12:31,343 DEV : loss 0.5568338632583618 - score 0.8205\n",
            "2019-07-23 20:12:32,623 TEST : loss 0.5568338632583618 - score 0.8205\n",
            "2019-07-23 20:12:32,628 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:12:33,718 epoch 18 - iter 0/32 - loss 0.57983541\n",
            "2019-07-23 20:12:34,501 epoch 18 - iter 3/32 - loss 0.52343847\n",
            "2019-07-23 20:12:35,449 epoch 18 - iter 6/32 - loss 0.54165599\n",
            "2019-07-23 20:12:37,076 epoch 18 - iter 9/32 - loss 0.54514270\n",
            "2019-07-23 20:12:38,064 epoch 18 - iter 12/32 - loss 0.54422991\n",
            "2019-07-23 20:12:38,924 epoch 18 - iter 15/32 - loss 0.54525654\n",
            "2019-07-23 20:12:39,714 epoch 18 - iter 18/32 - loss 0.53500984\n",
            "2019-07-23 20:12:40,557 epoch 18 - iter 21/32 - loss 0.53304693\n",
            "2019-07-23 20:12:42,065 epoch 18 - iter 24/32 - loss 0.53833676\n",
            "2019-07-23 20:12:42,888 epoch 18 - iter 27/32 - loss 0.53151395\n",
            "2019-07-23 20:12:43,550 epoch 18 - iter 30/32 - loss 0.53409613\n",
            "2019-07-23 20:12:43,944 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:12:43,946 EPOCH 18 done: loss 0.5373 - lr 0.0500 - bad epochs 3\n",
            "2019-07-23 20:12:45,142 DEV : loss 0.5204470753669739 - score 0.8333\n",
            "2019-07-23 20:12:46,392 TEST : loss 0.5204470753669739 - score 0.8333\n",
            "2019-07-23 20:12:55,268 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:12:57,171 epoch 19 - iter 0/32 - loss 0.38441426\n",
            "2019-07-23 20:12:57,983 epoch 19 - iter 3/32 - loss 0.55734299\n",
            "2019-07-23 20:12:58,906 epoch 19 - iter 6/32 - loss 0.54657183\n",
            "2019-07-23 20:12:59,924 epoch 19 - iter 9/32 - loss 0.53682533\n",
            "2019-07-23 20:13:00,740 epoch 19 - iter 12/32 - loss 0.54546343\n",
            "2019-07-23 20:13:02,384 epoch 19 - iter 15/32 - loss 0.54334626\n",
            "2019-07-23 20:13:03,169 epoch 19 - iter 18/32 - loss 0.55831753\n",
            "2019-07-23 20:13:03,798 epoch 19 - iter 21/32 - loss 0.56092769\n",
            "2019-07-23 20:13:04,864 epoch 19 - iter 24/32 - loss 0.54093950\n",
            "2019-07-23 20:13:05,594 epoch 19 - iter 27/32 - loss 0.54728229\n",
            "2019-07-23 20:13:07,069 epoch 19 - iter 30/32 - loss 0.54454735\n",
            "2019-07-23 20:13:07,426 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:13:07,427 EPOCH 19 done: loss 0.5432 - lr 0.0500 - bad epochs 4\n",
            "2019-07-23 20:13:08,753 DEV : loss 0.5428981781005859 - score 0.8333\n",
            "2019-07-23 20:13:10,033 TEST : loss 0.5428981781005859 - score 0.8333\n",
            "2019-07-23 20:13:18,904 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:13:20,134 epoch 20 - iter 0/32 - loss 0.51437581\n",
            "2019-07-23 20:13:20,998 epoch 20 - iter 3/32 - loss 0.58694817\n",
            "2019-07-23 20:13:21,877 epoch 20 - iter 6/32 - loss 0.53250454\n",
            "2019-07-23 20:13:23,686 epoch 20 - iter 9/32 - loss 0.52607419\n",
            "2019-07-23 20:13:24,470 epoch 20 - iter 12/32 - loss 0.53754180\n",
            "2019-07-23 20:13:25,283 epoch 20 - iter 15/32 - loss 0.53174434\n",
            "2019-07-23 20:13:26,059 epoch 20 - iter 18/32 - loss 0.53832160\n",
            "2019-07-23 20:13:26,975 epoch 20 - iter 21/32 - loss 0.53716025\n",
            "2019-07-23 20:13:28,582 epoch 20 - iter 24/32 - loss 0.53670436\n",
            "2019-07-23 20:13:29,322 epoch 20 - iter 27/32 - loss 0.53496716\n",
            "2019-07-23 20:13:30,314 epoch 20 - iter 30/32 - loss 0.53067225\n",
            "2019-07-23 20:13:30,675 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:13:30,677 EPOCH 20 done: loss 0.5354 - lr 0.0500 - bad epochs 5\n",
            "2019-07-23 20:13:31,888 DEV : loss 0.5209543704986572 - score 0.8333\n",
            "2019-07-23 20:13:33,165 TEST : loss 0.5209543704986572 - score 0.8333\n",
            "Epoch    19: reducing learning rate of group 0 to 2.5000e-02.\n",
            "2019-07-23 20:13:42,095 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:13:44,031 epoch 21 - iter 0/32 - loss 0.64446384\n",
            "2019-07-23 20:13:44,976 epoch 21 - iter 3/32 - loss 0.59304915\n",
            "2019-07-23 20:13:45,661 epoch 21 - iter 6/32 - loss 0.60407822\n",
            "2019-07-23 20:13:46,619 epoch 21 - iter 9/32 - loss 0.57012665\n",
            "2019-07-23 20:13:48,239 epoch 21 - iter 12/32 - loss 0.56011892\n",
            "2019-07-23 20:13:49,031 epoch 21 - iter 15/32 - loss 0.55970827\n",
            "2019-07-23 20:13:49,763 epoch 21 - iter 18/32 - loss 0.55655001\n",
            "2019-07-23 20:13:50,486 epoch 21 - iter 21/32 - loss 0.54406277\n",
            "2019-07-23 20:13:51,447 epoch 21 - iter 24/32 - loss 0.54117355\n",
            "2019-07-23 20:13:52,279 epoch 21 - iter 27/32 - loss 0.53404142\n",
            "2019-07-23 20:13:53,237 epoch 21 - iter 30/32 - loss 0.53398833\n",
            "2019-07-23 20:13:53,581 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:13:53,583 EPOCH 21 done: loss 0.5294 - lr 0.0250 - bad epochs 0\n",
            "2019-07-23 20:13:55,881 DEV : loss 0.5265529751777649 - score 0.8333\n",
            "2019-07-23 20:13:57,163 TEST : loss 0.5265529751777649 - score 0.8333\n",
            "2019-07-23 20:14:06,021 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:14:07,148 epoch 22 - iter 0/32 - loss 0.62491661\n",
            "2019-07-23 20:14:08,091 epoch 22 - iter 3/32 - loss 0.62265113\n",
            "2019-07-23 20:14:08,975 epoch 22 - iter 6/32 - loss 0.57348868\n",
            "2019-07-23 20:14:09,824 epoch 22 - iter 9/32 - loss 0.55570208\n",
            "2019-07-23 20:14:11,505 epoch 22 - iter 12/32 - loss 0.55762352\n",
            "2019-07-23 20:14:12,421 epoch 22 - iter 15/32 - loss 0.54646276\n",
            "2019-07-23 20:14:13,338 epoch 22 - iter 18/32 - loss 0.55511646\n",
            "2019-07-23 20:14:14,209 epoch 22 - iter 21/32 - loss 0.53969918\n",
            "2019-07-23 20:14:15,729 epoch 22 - iter 24/32 - loss 0.52768374\n",
            "2019-07-23 20:14:16,531 epoch 22 - iter 27/32 - loss 0.52921739\n",
            "2019-07-23 20:14:17,148 epoch 22 - iter 30/32 - loss 0.52271130\n",
            "2019-07-23 20:14:17,458 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:14:17,460 EPOCH 22 done: loss 0.5310 - lr 0.0250 - bad epochs 1\n",
            "2019-07-23 20:14:18,694 DEV : loss 0.5421192646026611 - score 0.8333\n",
            "2019-07-23 20:14:19,993 TEST : loss 0.5421192646026611 - score 0.8333\n",
            "2019-07-23 20:14:28,970 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:14:29,972 epoch 23 - iter 0/32 - loss 0.48819000\n",
            "2019-07-23 20:14:30,909 epoch 23 - iter 3/32 - loss 0.48011810\n",
            "2019-07-23 20:14:32,824 epoch 23 - iter 6/32 - loss 0.47675591\n",
            "2019-07-23 20:14:33,649 epoch 23 - iter 9/32 - loss 0.47253437\n",
            "2019-07-23 20:14:34,496 epoch 23 - iter 12/32 - loss 0.50184904\n",
            "2019-07-23 20:14:35,359 epoch 23 - iter 15/32 - loss 0.51932632\n",
            "2019-07-23 20:14:36,757 epoch 23 - iter 18/32 - loss 0.51495630\n",
            "2019-07-23 20:14:37,575 epoch 23 - iter 21/32 - loss 0.51928840\n",
            "2019-07-23 20:14:38,375 epoch 23 - iter 24/32 - loss 0.52292972\n",
            "2019-07-23 20:14:39,380 epoch 23 - iter 27/32 - loss 0.53369310\n",
            "2019-07-23 20:14:40,263 epoch 23 - iter 30/32 - loss 0.53307774\n",
            "2019-07-23 20:14:40,611 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:14:40,614 EPOCH 23 done: loss 0.5319 - lr 0.0250 - bad epochs 2\n",
            "2019-07-23 20:14:42,785 DEV : loss 0.5209918022155762 - score 0.8333\n",
            "2019-07-23 20:14:44,093 TEST : loss 0.5209918022155762 - score 0.8333\n",
            "2019-07-23 20:14:53,071 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:14:54,059 epoch 24 - iter 0/32 - loss 0.55218798\n",
            "2019-07-23 20:14:55,057 epoch 24 - iter 3/32 - loss 0.50542461\n",
            "2019-07-23 20:14:56,117 epoch 24 - iter 6/32 - loss 0.52209456\n",
            "2019-07-23 20:14:56,914 epoch 24 - iter 9/32 - loss 0.53447726\n",
            "2019-07-23 20:14:58,548 epoch 24 - iter 12/32 - loss 0.52729119\n",
            "2019-07-23 20:14:59,473 epoch 24 - iter 15/32 - loss 0.54176675\n",
            "2019-07-23 20:15:00,314 epoch 24 - iter 18/32 - loss 0.52732886\n",
            "2019-07-23 20:15:01,130 epoch 24 - iter 21/32 - loss 0.54059812\n",
            "2019-07-23 20:15:02,546 epoch 24 - iter 24/32 - loss 0.53781746\n",
            "2019-07-23 20:15:03,349 epoch 24 - iter 27/32 - loss 0.53097760\n",
            "2019-07-23 20:15:04,281 epoch 24 - iter 30/32 - loss 0.52868445\n",
            "2019-07-23 20:15:04,636 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:15:04,638 EPOCH 24 done: loss 0.5331 - lr 0.0250 - bad epochs 3\n",
            "2019-07-23 20:15:05,849 DEV : loss 0.5615201592445374 - score 0.8077\n",
            "2019-07-23 20:15:07,124 TEST : loss 0.5615201592445374 - score 0.8077\n",
            "2019-07-23 20:15:07,127 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:15:08,098 epoch 25 - iter 0/32 - loss 0.61897862\n",
            "2019-07-23 20:15:09,974 epoch 25 - iter 3/32 - loss 0.48322889\n",
            "2019-07-23 20:15:10,581 epoch 25 - iter 6/32 - loss 0.47761829\n",
            "2019-07-23 20:15:11,314 epoch 25 - iter 9/32 - loss 0.51393539\n",
            "2019-07-23 20:15:12,419 epoch 25 - iter 12/32 - loss 0.53096772\n",
            "2019-07-23 20:15:13,231 epoch 25 - iter 15/32 - loss 0.52860980\n",
            "2019-07-23 20:15:14,903 epoch 25 - iter 18/32 - loss 0.53009826\n",
            "2019-07-23 20:15:15,650 epoch 25 - iter 21/32 - loss 0.52137930\n",
            "2019-07-23 20:15:16,414 epoch 25 - iter 24/32 - loss 0.52658755\n",
            "2019-07-23 20:15:17,315 epoch 25 - iter 27/32 - loss 0.52339613\n",
            "2019-07-23 20:15:18,100 epoch 25 - iter 30/32 - loss 0.52626149\n",
            "2019-07-23 20:15:18,465 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:15:18,466 EPOCH 25 done: loss 0.5221 - lr 0.0250 - bad epochs 4\n",
            "2019-07-23 20:15:19,685 DEV : loss 0.5354334115982056 - score 0.8333\n",
            "2019-07-23 20:15:21,875 TEST : loss 0.5354334115982056 - score 0.8333\n",
            "2019-07-23 20:15:30,773 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:15:31,714 epoch 26 - iter 0/32 - loss 0.58606303\n",
            "2019-07-23 20:15:32,893 epoch 26 - iter 3/32 - loss 0.55806533\n",
            "2019-07-23 20:15:33,767 epoch 26 - iter 6/32 - loss 0.57161842\n",
            "2019-07-23 20:15:34,682 epoch 26 - iter 9/32 - loss 0.57652068\n",
            "2019-07-23 20:15:36,474 epoch 26 - iter 12/32 - loss 0.58205713\n",
            "2019-07-23 20:15:37,239 epoch 26 - iter 15/32 - loss 0.57348328\n",
            "2019-07-23 20:15:37,968 epoch 26 - iter 18/32 - loss 0.54824179\n",
            "2019-07-23 20:15:38,769 epoch 26 - iter 21/32 - loss 0.53714343\n",
            "2019-07-23 20:15:39,620 epoch 26 - iter 24/32 - loss 0.53307437\n",
            "2019-07-23 20:15:40,519 epoch 26 - iter 27/32 - loss 0.52858280\n",
            "2019-07-23 20:15:41,951 epoch 26 - iter 30/32 - loss 0.53444390\n",
            "2019-07-23 20:15:42,323 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:15:42,325 EPOCH 26 done: loss 0.5382 - lr 0.0250 - bad epochs 5\n",
            "2019-07-23 20:15:43,596 DEV : loss 0.5283479690551758 - score 0.8333\n",
            "2019-07-23 20:15:44,873 TEST : loss 0.5283479690551758 - score 0.8333\n",
            "Epoch    25: reducing learning rate of group 0 to 1.2500e-02.\n",
            "2019-07-23 20:15:53,875 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:15:54,901 epoch 27 - iter 0/32 - loss 0.55133426\n",
            "2019-07-23 20:15:56,023 epoch 27 - iter 3/32 - loss 0.45928377\n",
            "2019-07-23 20:15:57,631 epoch 27 - iter 6/32 - loss 0.51463358\n",
            "2019-07-23 20:15:58,603 epoch 27 - iter 9/32 - loss 0.53976870\n",
            "2019-07-23 20:15:59,357 epoch 27 - iter 12/32 - loss 0.53324531\n",
            "2019-07-23 20:16:00,312 epoch 27 - iter 15/32 - loss 0.53783174\n",
            "2019-07-23 20:16:01,278 epoch 27 - iter 18/32 - loss 0.52391903\n",
            "2019-07-23 20:16:02,736 epoch 27 - iter 21/32 - loss 0.52751386\n",
            "2019-07-23 20:16:03,578 epoch 27 - iter 24/32 - loss 0.52807472\n",
            "2019-07-23 20:16:04,451 epoch 27 - iter 27/32 - loss 0.52831492\n",
            "2019-07-23 20:16:05,381 epoch 27 - iter 30/32 - loss 0.53445431\n",
            "2019-07-23 20:16:05,716 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:16:05,718 EPOCH 27 done: loss 0.5307 - lr 0.0125 - bad epochs 0\n",
            "2019-07-23 20:16:06,976 DEV : loss 0.5247582197189331 - score 0.8333\n",
            "2019-07-23 20:16:08,291 TEST : loss 0.5247582197189331 - score 0.8333\n",
            "2019-07-23 20:16:17,253 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:16:19,079 epoch 28 - iter 0/32 - loss 0.53270537\n",
            "2019-07-23 20:16:19,993 epoch 28 - iter 3/32 - loss 0.54573773\n",
            "2019-07-23 20:16:20,933 epoch 28 - iter 6/32 - loss 0.49703228\n",
            "2019-07-23 20:16:21,822 epoch 28 - iter 9/32 - loss 0.47841509\n",
            "2019-07-23 20:16:22,749 epoch 28 - iter 12/32 - loss 0.49650440\n",
            "2019-07-23 20:16:24,257 epoch 28 - iter 15/32 - loss 0.48908434\n",
            "2019-07-23 20:16:25,235 epoch 28 - iter 18/32 - loss 0.50780344\n",
            "2019-07-23 20:16:26,139 epoch 28 - iter 21/32 - loss 0.51962529\n",
            "2019-07-23 20:16:26,977 epoch 28 - iter 24/32 - loss 0.52528398\n",
            "2019-07-23 20:16:27,957 epoch 28 - iter 27/32 - loss 0.52898790\n",
            "2019-07-23 20:16:29,265 epoch 28 - iter 30/32 - loss 0.52366847\n",
            "2019-07-23 20:16:29,588 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:16:29,590 EPOCH 28 done: loss 0.5229 - lr 0.0125 - bad epochs 1\n",
            "2019-07-23 20:16:30,899 DEV : loss 0.528257429599762 - score 0.8333\n",
            "2019-07-23 20:16:32,188 TEST : loss 0.528257429599762 - score 0.8333\n",
            "2019-07-23 20:16:41,136 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:16:42,215 epoch 29 - iter 0/32 - loss 0.60335261\n",
            "2019-07-23 20:16:43,347 epoch 29 - iter 3/32 - loss 0.49700863\n",
            "2019-07-23 20:16:45,026 epoch 29 - iter 6/32 - loss 0.44312109\n",
            "2019-07-23 20:16:45,930 epoch 29 - iter 9/32 - loss 0.49605712\n",
            "2019-07-23 20:16:46,753 epoch 29 - iter 12/32 - loss 0.49892955\n",
            "2019-07-23 20:16:47,582 epoch 29 - iter 15/32 - loss 0.51422517\n",
            "2019-07-23 20:16:48,509 epoch 29 - iter 18/32 - loss 0.50863056\n",
            "2019-07-23 20:16:49,262 epoch 29 - iter 21/32 - loss 0.52096881\n",
            "2019-07-23 20:16:50,802 epoch 29 - iter 24/32 - loss 0.53206953\n",
            "2019-07-23 20:16:51,609 epoch 29 - iter 27/32 - loss 0.53403747\n",
            "2019-07-23 20:16:52,444 epoch 29 - iter 30/32 - loss 0.52791809\n",
            "2019-07-23 20:16:52,851 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:16:52,853 EPOCH 29 done: loss 0.5214 - lr 0.0125 - bad epochs 2\n",
            "2019-07-23 20:16:54,078 DEV : loss 0.5425286293029785 - score 0.8333\n",
            "2019-07-23 20:16:55,367 TEST : loss 0.5425286293029785 - score 0.8333\n",
            "2019-07-23 20:17:04,307 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:17:06,197 epoch 30 - iter 0/32 - loss 0.56723094\n",
            "2019-07-23 20:17:07,248 epoch 30 - iter 3/32 - loss 0.50943993\n",
            "2019-07-23 20:17:08,137 epoch 30 - iter 6/32 - loss 0.51740493\n",
            "2019-07-23 20:17:09,071 epoch 30 - iter 9/32 - loss 0.51954287\n",
            "2019-07-23 20:17:09,845 epoch 30 - iter 12/32 - loss 0.55299015\n",
            "2019-07-23 20:17:10,753 epoch 30 - iter 15/32 - loss 0.54743956\n",
            "2019-07-23 20:17:12,252 epoch 30 - iter 18/32 - loss 0.52883049\n",
            "2019-07-23 20:17:13,143 epoch 30 - iter 21/32 - loss 0.53182186\n",
            "2019-07-23 20:17:13,869 epoch 30 - iter 24/32 - loss 0.52845375\n",
            "2019-07-23 20:17:14,710 epoch 30 - iter 27/32 - loss 0.52413804\n",
            "2019-07-23 20:17:15,518 epoch 30 - iter 30/32 - loss 0.52943093\n",
            "2019-07-23 20:17:15,840 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:17:15,844 EPOCH 30 done: loss 0.5269 - lr 0.0125 - bad epochs 3\n",
            "2019-07-23 20:17:17,954 DEV : loss 0.5249475836753845 - score 0.8333\n",
            "2019-07-23 20:17:19,270 TEST : loss 0.5249475836753845 - score 0.8333\n",
            "2019-07-23 20:17:28,317 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:17:29,390 epoch 31 - iter 0/32 - loss 0.55802017\n",
            "2019-07-23 20:17:30,457 epoch 31 - iter 3/32 - loss 0.52116852\n",
            "2019-07-23 20:17:31,367 epoch 31 - iter 6/32 - loss 0.50974702\n",
            "2019-07-23 20:17:33,017 epoch 31 - iter 9/32 - loss 0.49199982\n",
            "2019-07-23 20:17:33,779 epoch 31 - iter 12/32 - loss 0.49840301\n",
            "2019-07-23 20:17:34,742 epoch 31 - iter 15/32 - loss 0.49124122\n",
            "2019-07-23 20:17:35,532 epoch 31 - iter 18/32 - loss 0.49362568\n",
            "2019-07-23 20:17:36,402 epoch 31 - iter 21/32 - loss 0.50131879\n",
            "2019-07-23 20:17:37,951 epoch 31 - iter 24/32 - loss 0.50627233\n",
            "2019-07-23 20:17:38,822 epoch 31 - iter 27/32 - loss 0.51392369\n",
            "2019-07-23 20:17:39,637 epoch 31 - iter 30/32 - loss 0.51920054\n",
            "2019-07-23 20:17:39,962 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:17:39,964 EPOCH 31 done: loss 0.5215 - lr 0.0125 - bad epochs 4\n",
            "2019-07-23 20:17:41,219 DEV : loss 0.5218908190727234 - score 0.8333\n",
            "2019-07-23 20:17:42,488 TEST : loss 0.5218908190727234 - score 0.8333\n",
            "2019-07-23 20:17:51,387 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:17:52,481 epoch 32 - iter 0/32 - loss 0.43022692\n",
            "2019-07-23 20:17:54,333 epoch 32 - iter 3/32 - loss 0.50401891\n",
            "2019-07-23 20:17:55,295 epoch 32 - iter 6/32 - loss 0.51353012\n",
            "2019-07-23 20:17:56,055 epoch 32 - iter 9/32 - loss 0.53761357\n",
            "2019-07-23 20:17:56,961 epoch 32 - iter 12/32 - loss 0.54868636\n",
            "2019-07-23 20:17:57,882 epoch 32 - iter 15/32 - loss 0.52172198\n",
            "2019-07-23 20:17:59,200 epoch 32 - iter 18/32 - loss 0.51684082\n",
            "2019-07-23 20:18:00,054 epoch 32 - iter 21/32 - loss 0.52685591\n",
            "2019-07-23 20:18:00,841 epoch 32 - iter 24/32 - loss 0.52594575\n",
            "2019-07-23 20:18:01,783 epoch 32 - iter 27/32 - loss 0.52760180\n",
            "2019-07-23 20:18:02,684 epoch 32 - iter 30/32 - loss 0.52518569\n",
            "2019-07-23 20:18:03,008 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:18:03,010 EPOCH 32 done: loss 0.5277 - lr 0.0125 - bad epochs 5\n",
            "2019-07-23 20:18:05,138 DEV : loss 0.5291627645492554 - score 0.8333\n",
            "2019-07-23 20:18:06,470 TEST : loss 0.5291627645492554 - score 0.8333\n",
            "Epoch    31: reducing learning rate of group 0 to 6.2500e-03.\n",
            "2019-07-23 20:18:15,472 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:18:16,494 epoch 33 - iter 0/32 - loss 0.62251246\n",
            "2019-07-23 20:18:17,542 epoch 33 - iter 3/32 - loss 0.58753023\n",
            "2019-07-23 20:18:18,580 epoch 33 - iter 6/32 - loss 0.56818253\n",
            "2019-07-23 20:18:20,226 epoch 33 - iter 9/32 - loss 0.54508492\n",
            "2019-07-23 20:18:21,117 epoch 33 - iter 12/32 - loss 0.54748830\n",
            "2019-07-23 20:18:21,923 epoch 33 - iter 15/32 - loss 0.55441927\n",
            "2019-07-23 20:18:22,966 epoch 33 - iter 18/32 - loss 0.54584875\n",
            "2019-07-23 20:18:23,671 epoch 33 - iter 21/32 - loss 0.54827321\n",
            "2019-07-23 20:18:25,198 epoch 33 - iter 24/32 - loss 0.53116582\n",
            "2019-07-23 20:18:26,077 epoch 33 - iter 27/32 - loss 0.53220278\n",
            "2019-07-23 20:18:26,914 epoch 33 - iter 30/32 - loss 0.52494554\n",
            "2019-07-23 20:18:27,278 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:18:27,280 EPOCH 33 done: loss 0.5318 - lr 0.0063 - bad epochs 0\n",
            "2019-07-23 20:18:28,576 DEV : loss 0.5234106779098511 - score 0.8333\n",
            "2019-07-23 20:18:29,862 TEST : loss 0.5234106779098511 - score 0.8333\n",
            "2019-07-23 20:18:38,826 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:18:39,907 epoch 34 - iter 0/32 - loss 0.61356449\n",
            "2019-07-23 20:18:41,733 epoch 34 - iter 3/32 - loss 0.58185837\n",
            "2019-07-23 20:18:42,596 epoch 34 - iter 6/32 - loss 0.54668913\n",
            "2019-07-23 20:18:43,272 epoch 34 - iter 9/32 - loss 0.54448402\n",
            "2019-07-23 20:18:44,343 epoch 34 - iter 12/32 - loss 0.52781865\n",
            "2019-07-23 20:18:45,277 epoch 34 - iter 15/32 - loss 0.51913806\n",
            "2019-07-23 20:18:46,778 epoch 34 - iter 18/32 - loss 0.52106851\n",
            "2019-07-23 20:18:47,716 epoch 34 - iter 21/32 - loss 0.52610317\n",
            "2019-07-23 20:18:48,507 epoch 34 - iter 24/32 - loss 0.52554150\n",
            "2019-07-23 20:18:49,268 epoch 34 - iter 27/32 - loss 0.51978817\n",
            "2019-07-23 20:18:50,155 epoch 34 - iter 30/32 - loss 0.52078480\n",
            "2019-07-23 20:18:50,481 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:18:50,483 EPOCH 34 done: loss 0.5179 - lr 0.0063 - bad epochs 1\n",
            "2019-07-23 20:18:52,657 DEV : loss 0.5293759107589722 - score 0.8333\n",
            "2019-07-23 20:18:53,996 TEST : loss 0.5293759107589722 - score 0.8333\n",
            "2019-07-23 20:19:02,953 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:19:04,217 epoch 35 - iter 0/32 - loss 0.55973417\n",
            "2019-07-23 20:19:05,077 epoch 35 - iter 3/32 - loss 0.50446724\n",
            "2019-07-23 20:19:05,817 epoch 35 - iter 6/32 - loss 0.58521772\n",
            "2019-07-23 20:19:07,517 epoch 35 - iter 9/32 - loss 0.55272823\n",
            "2019-07-23 20:19:08,295 epoch 35 - iter 12/32 - loss 0.54814622\n",
            "2019-07-23 20:19:09,158 epoch 35 - iter 15/32 - loss 0.53195479\n",
            "2019-07-23 20:19:10,161 epoch 35 - iter 18/32 - loss 0.52190022\n",
            "2019-07-23 20:19:10,813 epoch 35 - iter 21/32 - loss 0.51938232\n",
            "2019-07-23 20:19:12,540 epoch 35 - iter 24/32 - loss 0.51740806\n",
            "2019-07-23 20:19:13,208 epoch 35 - iter 27/32 - loss 0.52502911\n",
            "2019-07-23 20:19:13,985 epoch 35 - iter 30/32 - loss 0.52052155\n",
            "2019-07-23 20:19:14,391 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:19:14,396 EPOCH 35 done: loss 0.5187 - lr 0.0063 - bad epochs 2\n",
            "2019-07-23 20:19:15,675 DEV : loss 0.5297836661338806 - score 0.8333\n",
            "2019-07-23 20:19:16,964 TEST : loss 0.5297836661338806 - score 0.8333\n",
            "2019-07-23 20:19:25,805 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:19:27,009 epoch 36 - iter 0/32 - loss 0.69646776\n",
            "2019-07-23 20:19:28,881 epoch 36 - iter 3/32 - loss 0.59345237\n",
            "2019-07-23 20:19:29,838 epoch 36 - iter 6/32 - loss 0.53845478\n",
            "2019-07-23 20:19:30,640 epoch 36 - iter 9/32 - loss 0.52350458\n",
            "2019-07-23 20:19:31,586 epoch 36 - iter 12/32 - loss 0.52190953\n",
            "2019-07-23 20:19:32,491 epoch 36 - iter 15/32 - loss 0.53455563\n",
            "2019-07-23 20:19:33,966 epoch 36 - iter 18/32 - loss 0.52415608\n",
            "2019-07-23 20:19:34,802 epoch 36 - iter 21/32 - loss 0.52897878\n",
            "2019-07-23 20:19:35,590 epoch 36 - iter 24/32 - loss 0.52102546\n",
            "2019-07-23 20:19:36,350 epoch 36 - iter 27/32 - loss 0.52741684\n",
            "2019-07-23 20:19:37,185 epoch 36 - iter 30/32 - loss 0.52569521\n",
            "2019-07-23 20:19:37,530 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:19:37,533 EPOCH 36 done: loss 0.5262 - lr 0.0063 - bad epochs 3\n",
            "2019-07-23 20:19:38,756 DEV : loss 0.5236992835998535 - score 0.8333\n",
            "2019-07-23 20:19:40,900 TEST : loss 0.5236992835998535 - score 0.8333\n",
            "2019-07-23 20:19:49,592 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:19:50,752 epoch 37 - iter 0/32 - loss 0.58751744\n",
            "2019-07-23 20:19:51,900 epoch 37 - iter 3/32 - loss 0.60735674\n",
            "2019-07-23 20:19:52,779 epoch 37 - iter 6/32 - loss 0.57356138\n",
            "2019-07-23 20:19:53,656 epoch 37 - iter 9/32 - loss 0.53656156\n",
            "2019-07-23 20:19:55,384 epoch 37 - iter 12/32 - loss 0.53342005\n",
            "2019-07-23 20:19:56,209 epoch 37 - iter 15/32 - loss 0.53484758\n",
            "2019-07-23 20:19:57,150 epoch 37 - iter 18/32 - loss 0.53387844\n",
            "2019-07-23 20:19:58,086 epoch 37 - iter 21/32 - loss 0.52860238\n",
            "2019-07-23 20:19:58,993 epoch 37 - iter 24/32 - loss 0.52383013\n",
            "2019-07-23 20:20:00,503 epoch 37 - iter 27/32 - loss 0.51916630\n",
            "2019-07-23 20:20:01,330 epoch 37 - iter 30/32 - loss 0.52176372\n",
            "2019-07-23 20:20:01,670 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:20:01,672 EPOCH 37 done: loss 0.5222 - lr 0.0063 - bad epochs 4\n",
            "2019-07-23 20:20:02,979 DEV : loss 0.52812659740448 - score 0.8333\n",
            "2019-07-23 20:20:04,318 TEST : loss 0.52812659740448 - score 0.8333\n",
            "2019-07-23 20:20:13,539 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:20:14,638 epoch 38 - iter 0/32 - loss 0.45077848\n",
            "2019-07-23 20:20:15,776 epoch 38 - iter 3/32 - loss 0.49253254\n",
            "2019-07-23 20:20:17,515 epoch 38 - iter 6/32 - loss 0.52895423\n",
            "2019-07-23 20:20:18,414 epoch 38 - iter 9/32 - loss 0.52561251\n",
            "2019-07-23 20:20:19,392 epoch 38 - iter 12/32 - loss 0.52620154\n",
            "2019-07-23 20:20:20,260 epoch 38 - iter 15/32 - loss 0.51955187\n",
            "2019-07-23 20:20:21,763 epoch 38 - iter 18/32 - loss 0.53100011\n",
            "2019-07-23 20:20:22,472 epoch 38 - iter 21/32 - loss 0.53022187\n",
            "2019-07-23 20:20:23,290 epoch 38 - iter 24/32 - loss 0.53237302\n",
            "2019-07-23 20:20:24,162 epoch 38 - iter 27/32 - loss 0.52973854\n",
            "2019-07-23 20:20:24,938 epoch 38 - iter 30/32 - loss 0.52384119\n",
            "2019-07-23 20:20:25,271 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:20:25,273 EPOCH 38 done: loss 0.5189 - lr 0.0063 - bad epochs 5\n",
            "2019-07-23 20:20:26,579 DEV : loss 0.5276350378990173 - score 0.8333\n",
            "2019-07-23 20:20:27,874 TEST : loss 0.5276350378990173 - score 0.8333\n",
            "Epoch    37: reducing learning rate of group 0 to 3.1250e-03.\n",
            "2019-07-23 20:20:36,733 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:20:38,852 epoch 39 - iter 0/32 - loss 0.52572811\n",
            "2019-07-23 20:20:40,008 epoch 39 - iter 3/32 - loss 0.55714971\n",
            "2019-07-23 20:20:40,861 epoch 39 - iter 6/32 - loss 0.54455158\n",
            "2019-07-23 20:20:41,607 epoch 39 - iter 9/32 - loss 0.54745298\n",
            "2019-07-23 20:20:42,522 epoch 39 - iter 12/32 - loss 0.53292923\n",
            "2019-07-23 20:20:43,929 epoch 39 - iter 15/32 - loss 0.51931983\n",
            "2019-07-23 20:20:44,630 epoch 39 - iter 18/32 - loss 0.53967452\n",
            "2019-07-23 20:20:45,607 epoch 39 - iter 21/32 - loss 0.52874108\n",
            "2019-07-23 20:20:46,327 epoch 39 - iter 24/32 - loss 0.52612411\n",
            "2019-07-23 20:20:47,261 epoch 39 - iter 27/32 - loss 0.52483556\n",
            "2019-07-23 20:20:47,991 epoch 39 - iter 30/32 - loss 0.52345182\n",
            "2019-07-23 20:20:49,006 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:20:49,009 EPOCH 39 done: loss 0.5149 - lr 0.0031 - bad epochs 0\n",
            "2019-07-23 20:20:50,359 DEV : loss 0.5261865854263306 - score 0.8333\n",
            "2019-07-23 20:20:51,692 TEST : loss 0.5261865854263306 - score 0.8333\n",
            "2019-07-23 20:21:00,704 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:21:01,704 epoch 40 - iter 0/32 - loss 0.42332602\n",
            "2019-07-23 20:21:02,631 epoch 40 - iter 3/32 - loss 0.43975085\n",
            "2019-07-23 20:21:03,668 epoch 40 - iter 6/32 - loss 0.47369569\n",
            "2019-07-23 20:21:05,390 epoch 40 - iter 9/32 - loss 0.49878943\n",
            "2019-07-23 20:21:06,247 epoch 40 - iter 12/32 - loss 0.50771708\n",
            "2019-07-23 20:21:07,057 epoch 40 - iter 15/32 - loss 0.50904887\n",
            "2019-07-23 20:21:07,865 epoch 40 - iter 18/32 - loss 0.51396857\n",
            "2019-07-23 20:21:09,392 epoch 40 - iter 21/32 - loss 0.51375200\n",
            "2019-07-23 20:21:10,264 epoch 40 - iter 24/32 - loss 0.51056429\n",
            "2019-07-23 20:21:11,124 epoch 40 - iter 27/32 - loss 0.52029185\n",
            "2019-07-23 20:21:11,919 epoch 40 - iter 30/32 - loss 0.52475620\n",
            "2019-07-23 20:21:12,268 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:21:12,270 EPOCH 40 done: loss 0.5237 - lr 0.0031 - bad epochs 1\n",
            "2019-07-23 20:21:13,556 DEV : loss 0.5240964889526367 - score 0.8333\n",
            "2019-07-23 20:21:14,887 TEST : loss 0.5240964889526367 - score 0.8333\n",
            "2019-07-23 20:21:23,731 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:21:24,863 epoch 41 - iter 0/32 - loss 0.38255811\n",
            "2019-07-23 20:21:26,894 epoch 41 - iter 3/32 - loss 0.49717492\n",
            "2019-07-23 20:21:27,715 epoch 41 - iter 6/32 - loss 0.52265025\n",
            "2019-07-23 20:21:28,522 epoch 41 - iter 9/32 - loss 0.51488939\n",
            "2019-07-23 20:21:29,477 epoch 41 - iter 12/32 - loss 0.51777533\n",
            "2019-07-23 20:21:31,009 epoch 41 - iter 15/32 - loss 0.51081498\n",
            "2019-07-23 20:21:31,905 epoch 41 - iter 18/32 - loss 0.51307741\n",
            "2019-07-23 20:21:32,712 epoch 41 - iter 21/32 - loss 0.51911233\n",
            "2019-07-23 20:21:33,439 epoch 41 - iter 24/32 - loss 0.52254783\n",
            "2019-07-23 20:21:34,211 epoch 41 - iter 27/32 - loss 0.52435404\n",
            "2019-07-23 20:21:35,011 epoch 41 - iter 30/32 - loss 0.52353280\n",
            "2019-07-23 20:21:35,372 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:21:35,373 EPOCH 41 done: loss 0.5154 - lr 0.0031 - bad epochs 2\n",
            "2019-07-23 20:21:37,664 DEV : loss 0.5252335071563721 - score 0.8333\n",
            "2019-07-23 20:21:39,025 TEST : loss 0.5252335071563721 - score 0.8333\n",
            "2019-07-23 20:21:47,904 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:21:49,087 epoch 42 - iter 0/32 - loss 0.67340595\n",
            "2019-07-23 20:21:50,070 epoch 42 - iter 3/32 - loss 0.52513513\n",
            "2019-07-23 20:21:50,958 epoch 42 - iter 6/32 - loss 0.50658080\n",
            "2019-07-23 20:21:52,753 epoch 42 - iter 9/32 - loss 0.49941278\n",
            "2019-07-23 20:21:53,627 epoch 42 - iter 12/32 - loss 0.51529112\n",
            "2019-07-23 20:21:54,310 epoch 42 - iter 15/32 - loss 0.51901755\n",
            "2019-07-23 20:21:55,318 epoch 42 - iter 18/32 - loss 0.51120022\n",
            "2019-07-23 20:21:56,190 epoch 42 - iter 21/32 - loss 0.51203054\n",
            "2019-07-23 20:21:57,645 epoch 42 - iter 24/32 - loss 0.51937919\n",
            "2019-07-23 20:21:58,493 epoch 42 - iter 27/32 - loss 0.51618202\n",
            "2019-07-23 20:21:59,407 epoch 42 - iter 30/32 - loss 0.51720580\n",
            "2019-07-23 20:21:59,809 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:21:59,812 EPOCH 42 done: loss 0.5235 - lr 0.0031 - bad epochs 3\n",
            "2019-07-23 20:22:01,110 DEV : loss 0.5256131887435913 - score 0.8333\n",
            "2019-07-23 20:22:02,456 TEST : loss 0.5256131887435913 - score 0.8333\n",
            "2019-07-23 20:22:11,398 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:22:12,561 epoch 43 - iter 0/32 - loss 0.54702246\n",
            "2019-07-23 20:22:14,422 epoch 43 - iter 3/32 - loss 0.49503168\n",
            "2019-07-23 20:22:15,153 epoch 43 - iter 6/32 - loss 0.53304734\n",
            "2019-07-23 20:22:16,117 epoch 43 - iter 9/32 - loss 0.52681913\n",
            "2019-07-23 20:22:16,914 epoch 43 - iter 12/32 - loss 0.54166317\n",
            "2019-07-23 20:22:17,703 epoch 43 - iter 15/32 - loss 0.54789196\n",
            "2019-07-23 20:22:19,289 epoch 43 - iter 18/32 - loss 0.54378005\n",
            "2019-07-23 20:22:20,106 epoch 43 - iter 21/32 - loss 0.53984364\n",
            "2019-07-23 20:22:20,842 epoch 43 - iter 24/32 - loss 0.53724983\n",
            "2019-07-23 20:22:21,665 epoch 43 - iter 27/32 - loss 0.53996327\n",
            "2019-07-23 20:22:22,451 epoch 43 - iter 30/32 - loss 0.52473033\n",
            "2019-07-23 20:22:22,824 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:22:22,825 EPOCH 43 done: loss 0.5224 - lr 0.0031 - bad epochs 4\n",
            "2019-07-23 20:22:24,066 DEV : loss 0.5262536406517029 - score 0.8333\n",
            "2019-07-23 20:22:26,331 TEST : loss 0.5262536406517029 - score 0.8333\n",
            "2019-07-23 20:22:35,256 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:22:36,706 epoch 44 - iter 0/32 - loss 0.33550075\n",
            "2019-07-23 20:22:37,454 epoch 44 - iter 3/32 - loss 0.52014298\n",
            "2019-07-23 20:22:38,290 epoch 44 - iter 6/32 - loss 0.51205998\n",
            "2019-07-23 20:22:39,796 epoch 44 - iter 9/32 - loss 0.50957293\n",
            "2019-07-23 20:22:40,716 epoch 44 - iter 12/32 - loss 0.51146066\n",
            "2019-07-23 20:22:41,576 epoch 44 - iter 15/32 - loss 0.51663666\n",
            "2019-07-23 20:22:42,354 epoch 44 - iter 18/32 - loss 0.54074724\n",
            "2019-07-23 20:22:43,047 epoch 44 - iter 21/32 - loss 0.53909945\n",
            "2019-07-23 20:22:44,670 epoch 44 - iter 24/32 - loss 0.53443364\n",
            "2019-07-23 20:22:45,640 epoch 44 - iter 27/32 - loss 0.53391664\n",
            "2019-07-23 20:22:46,367 epoch 44 - iter 30/32 - loss 0.52785737\n",
            "2019-07-23 20:22:46,725 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:22:46,726 EPOCH 44 done: loss 0.5252 - lr 0.0031 - bad epochs 5\n",
            "2019-07-23 20:22:48,002 DEV : loss 0.5261768698692322 - score 0.8333\n",
            "2019-07-23 20:22:49,333 TEST : loss 0.5261768698692322 - score 0.8333\n",
            "Epoch    43: reducing learning rate of group 0 to 1.5625e-03.\n",
            "2019-07-23 20:22:58,266 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:22:59,377 epoch 45 - iter 0/32 - loss 0.50071222\n",
            "2019-07-23 20:23:00,297 epoch 45 - iter 3/32 - loss 0.54879398\n",
            "2019-07-23 20:23:02,229 epoch 45 - iter 6/32 - loss 0.50344855\n",
            "2019-07-23 20:23:03,009 epoch 45 - iter 9/32 - loss 0.49874748\n",
            "2019-07-23 20:23:03,847 epoch 45 - iter 12/32 - loss 0.52334333\n",
            "2019-07-23 20:23:04,743 epoch 45 - iter 15/32 - loss 0.52009664\n",
            "2019-07-23 20:23:05,515 epoch 45 - iter 18/32 - loss 0.51733411\n",
            "2019-07-23 20:23:06,247 epoch 45 - iter 21/32 - loss 0.51802208\n",
            "2019-07-23 20:23:07,898 epoch 45 - iter 24/32 - loss 0.51956162\n",
            "2019-07-23 20:23:08,753 epoch 45 - iter 27/32 - loss 0.52180975\n",
            "2019-07-23 20:23:09,609 epoch 45 - iter 30/32 - loss 0.52275058\n",
            "2019-07-23 20:23:09,945 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:23:09,947 EPOCH 45 done: loss 0.5148 - lr 0.0016 - bad epochs 0\n",
            "2019-07-23 20:23:11,203 DEV : loss 0.5267213582992554 - score 0.8333\n",
            "2019-07-23 20:23:12,512 TEST : loss 0.5267213582992554 - score 0.8333\n",
            "2019-07-23 20:23:21,453 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:23:23,567 epoch 46 - iter 0/32 - loss 0.73643506\n",
            "2019-07-23 20:23:24,518 epoch 46 - iter 3/32 - loss 0.53599095\n",
            "2019-07-23 20:23:25,310 epoch 46 - iter 6/32 - loss 0.45358242\n",
            "2019-07-23 20:23:26,111 epoch 46 - iter 9/32 - loss 0.45648091\n",
            "2019-07-23 20:23:26,877 epoch 46 - iter 12/32 - loss 0.47963200\n",
            "2019-07-23 20:23:28,359 epoch 46 - iter 15/32 - loss 0.50256901\n",
            "2019-07-23 20:23:29,234 epoch 46 - iter 18/32 - loss 0.50649111\n",
            "2019-07-23 20:23:30,036 epoch 46 - iter 21/32 - loss 0.51285580\n",
            "2019-07-23 20:23:30,936 epoch 46 - iter 24/32 - loss 0.51602924\n",
            "2019-07-23 20:23:31,764 epoch 46 - iter 27/32 - loss 0.51230968\n",
            "2019-07-23 20:23:33,267 epoch 46 - iter 30/32 - loss 0.52133191\n",
            "2019-07-23 20:23:33,679 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:23:33,680 EPOCH 46 done: loss 0.5283 - lr 0.0016 - bad epochs 1\n",
            "2019-07-23 20:23:35,005 DEV : loss 0.5256251096725464 - score 0.8333\n",
            "2019-07-23 20:23:36,352 TEST : loss 0.5256251096725464 - score 0.8333\n",
            "2019-07-23 20:23:45,256 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:23:46,449 epoch 47 - iter 0/32 - loss 0.52282149\n",
            "2019-07-23 20:23:47,532 epoch 47 - iter 3/32 - loss 0.55527613\n",
            "2019-07-23 20:23:48,330 epoch 47 - iter 6/32 - loss 0.53701859\n",
            "2019-07-23 20:23:50,021 epoch 47 - iter 9/32 - loss 0.51646403\n",
            "2019-07-23 20:23:50,908 epoch 47 - iter 12/32 - loss 0.52467979\n",
            "2019-07-23 20:23:51,682 epoch 47 - iter 15/32 - loss 0.51367638\n",
            "2019-07-23 20:23:52,557 epoch 47 - iter 18/32 - loss 0.50884057\n",
            "2019-07-23 20:23:54,125 epoch 47 - iter 21/32 - loss 0.51298703\n",
            "2019-07-23 20:23:55,061 epoch 47 - iter 24/32 - loss 0.51770175\n",
            "2019-07-23 20:23:55,755 epoch 47 - iter 27/32 - loss 0.51832208\n",
            "2019-07-23 20:23:56,563 epoch 47 - iter 30/32 - loss 0.51702450\n",
            "2019-07-23 20:23:56,914 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:23:56,915 EPOCH 47 done: loss 0.5182 - lr 0.0016 - bad epochs 2\n",
            "2019-07-23 20:23:58,212 DEV : loss 0.5254979133605957 - score 0.8333\n",
            "2019-07-23 20:23:59,525 TEST : loss 0.5254979133605957 - score 0.8333\n",
            "2019-07-23 20:24:08,529 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:24:09,642 epoch 48 - iter 0/32 - loss 0.51519090\n",
            "2019-07-23 20:24:11,508 epoch 48 - iter 3/32 - loss 0.47853322\n",
            "2019-07-23 20:24:12,406 epoch 48 - iter 6/32 - loss 0.50433568\n",
            "2019-07-23 20:24:13,322 epoch 48 - iter 9/32 - loss 0.52964694\n",
            "2019-07-23 20:24:14,119 epoch 48 - iter 12/32 - loss 0.51653252\n",
            "2019-07-23 20:24:15,615 epoch 48 - iter 15/32 - loss 0.50791081\n",
            "2019-07-23 20:24:16,302 epoch 48 - iter 18/32 - loss 0.53012085\n",
            "2019-07-23 20:24:17,128 epoch 48 - iter 21/32 - loss 0.52249628\n",
            "2019-07-23 20:24:17,960 epoch 48 - iter 24/32 - loss 0.52866424\n",
            "2019-07-23 20:24:18,859 epoch 48 - iter 27/32 - loss 0.52069068\n",
            "2019-07-23 20:24:19,609 epoch 48 - iter 30/32 - loss 0.52353311\n",
            "2019-07-23 20:24:19,999 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:24:20,000 EPOCH 48 done: loss 0.5184 - lr 0.0016 - bad epochs 3\n",
            "2019-07-23 20:24:22,178 DEV : loss 0.526014506816864 - score 0.8333\n",
            "2019-07-23 20:24:23,524 TEST : loss 0.526014506816864 - score 0.8333\n",
            "2019-07-23 20:24:32,514 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:24:33,696 epoch 49 - iter 0/32 - loss 0.37898761\n",
            "2019-07-23 20:24:34,845 epoch 49 - iter 3/32 - loss 0.46785546\n",
            "2019-07-23 20:24:35,606 epoch 49 - iter 6/32 - loss 0.48538041\n",
            "2019-07-23 20:24:37,366 epoch 49 - iter 9/32 - loss 0.50596299\n",
            "2019-07-23 20:24:38,258 epoch 49 - iter 12/32 - loss 0.49871314\n",
            "2019-07-23 20:24:39,092 epoch 49 - iter 15/32 - loss 0.50918342\n",
            "2019-07-23 20:24:39,875 epoch 49 - iter 18/32 - loss 0.51423320\n",
            "2019-07-23 20:24:40,791 epoch 49 - iter 21/32 - loss 0.51156363\n",
            "2019-07-23 20:24:42,393 epoch 49 - iter 24/32 - loss 0.51795006\n",
            "2019-07-23 20:24:43,207 epoch 49 - iter 27/32 - loss 0.51425408\n",
            "2019-07-23 20:24:43,985 epoch 49 - iter 30/32 - loss 0.51745394\n",
            "2019-07-23 20:24:44,328 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:24:44,330 EPOCH 49 done: loss 0.5149 - lr 0.0016 - bad epochs 4\n",
            "2019-07-23 20:24:45,654 DEV : loss 0.5254129767417908 - score 0.8333\n",
            "2019-07-23 20:24:46,969 TEST : loss 0.5254129767417908 - score 0.8333\n",
            "2019-07-23 20:24:56,010 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:24:58,180 epoch 50 - iter 0/32 - loss 0.57546467\n",
            "2019-07-23 20:24:59,027 epoch 50 - iter 3/32 - loss 0.50450180\n",
            "2019-07-23 20:24:59,844 epoch 50 - iter 6/32 - loss 0.55370858\n",
            "2019-07-23 20:25:00,788 epoch 50 - iter 9/32 - loss 0.54439569\n",
            "2019-07-23 20:25:01,449 epoch 50 - iter 12/32 - loss 0.54100712\n",
            "2019-07-23 20:25:02,333 epoch 50 - iter 15/32 - loss 0.52893329\n",
            "2019-07-23 20:25:03,955 epoch 50 - iter 18/32 - loss 0.53147424\n",
            "2019-07-23 20:25:04,831 epoch 50 - iter 21/32 - loss 0.53069387\n",
            "2019-07-23 20:25:05,836 epoch 50 - iter 24/32 - loss 0.53156041\n",
            "2019-07-23 20:25:06,599 epoch 50 - iter 27/32 - loss 0.53120550\n",
            "2019-07-23 20:25:07,302 epoch 50 - iter 30/32 - loss 0.52089663\n",
            "2019-07-23 20:25:07,678 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:25:07,685 EPOCH 50 done: loss 0.5252 - lr 0.0016 - bad epochs 5\n",
            "2019-07-23 20:25:09,925 DEV : loss 0.525688111782074 - score 0.8333\n",
            "2019-07-23 20:25:11,312 TEST : loss 0.525688111782074 - score 0.8333\n",
            "Epoch    49: reducing learning rate of group 0 to 7.8125e-04.\n",
            "2019-07-23 20:25:20,185 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:25:21,302 epoch 51 - iter 0/32 - loss 0.56121475\n",
            "2019-07-23 20:25:22,353 epoch 51 - iter 3/32 - loss 0.53847475\n",
            "2019-07-23 20:25:23,270 epoch 51 - iter 6/32 - loss 0.55233681\n",
            "2019-07-23 20:25:24,109 epoch 51 - iter 9/32 - loss 0.53724294\n",
            "2019-07-23 20:25:25,858 epoch 51 - iter 12/32 - loss 0.52749379\n",
            "2019-07-23 20:25:26,738 epoch 51 - iter 15/32 - loss 0.52417541\n",
            "2019-07-23 20:25:27,641 epoch 51 - iter 18/32 - loss 0.52486210\n",
            "2019-07-23 20:25:28,519 epoch 51 - iter 21/32 - loss 0.51383000\n",
            "2019-07-23 20:25:29,960 epoch 51 - iter 24/32 - loss 0.52957647\n",
            "2019-07-23 20:25:30,808 epoch 51 - iter 27/32 - loss 0.52335892\n",
            "2019-07-23 20:25:31,578 epoch 51 - iter 30/32 - loss 0.51941565\n",
            "2019-07-23 20:25:31,929 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:25:31,930 EPOCH 51 done: loss 0.5165 - lr 0.0008 - bad epochs 0\n",
            "2019-07-23 20:25:33,209 DEV : loss 0.5259252190589905 - score 0.8333\n",
            "2019-07-23 20:25:34,530 TEST : loss 0.5259252190589905 - score 0.8333\n",
            "2019-07-23 20:25:43,555 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:25:44,696 epoch 52 - iter 0/32 - loss 0.49493212\n",
            "2019-07-23 20:25:46,381 epoch 52 - iter 3/32 - loss 0.51858553\n",
            "2019-07-23 20:25:47,561 epoch 52 - iter 6/32 - loss 0.48843156\n",
            "2019-07-23 20:25:48,404 epoch 52 - iter 9/32 - loss 0.49955581\n",
            "2019-07-23 20:25:49,299 epoch 52 - iter 12/32 - loss 0.51373339\n",
            "2019-07-23 20:25:50,188 epoch 52 - iter 15/32 - loss 0.50570985\n",
            "2019-07-23 20:25:51,598 epoch 52 - iter 18/32 - loss 0.51548012\n",
            "2019-07-23 20:25:52,676 epoch 52 - iter 21/32 - loss 0.50864917\n",
            "2019-07-23 20:25:53,364 epoch 52 - iter 24/32 - loss 0.50757580\n",
            "2019-07-23 20:25:54,206 epoch 52 - iter 27/32 - loss 0.50872203\n",
            "2019-07-23 20:25:54,932 epoch 52 - iter 30/32 - loss 0.51678490\n",
            "2019-07-23 20:25:55,304 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:25:55,305 EPOCH 52 done: loss 0.5126 - lr 0.0008 - bad epochs 1\n",
            "2019-07-23 20:25:57,729 DEV : loss 0.5263727307319641 - score 0.8333\n",
            "2019-07-23 20:25:59,197 TEST : loss 0.5263727307319641 - score 0.8333\n",
            "2019-07-23 20:26:08,571 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:26:09,907 epoch 53 - iter 0/32 - loss 0.63571948\n",
            "2019-07-23 20:26:10,887 epoch 53 - iter 3/32 - loss 0.54993255\n",
            "2019-07-23 20:26:11,856 epoch 53 - iter 6/32 - loss 0.51516946\n",
            "2019-07-23 20:26:12,793 epoch 53 - iter 9/32 - loss 0.51652423\n",
            "2019-07-23 20:26:14,380 epoch 53 - iter 12/32 - loss 0.53110973\n",
            "2019-07-23 20:26:15,213 epoch 53 - iter 15/32 - loss 0.51163440\n",
            "2019-07-23 20:26:16,181 epoch 53 - iter 18/32 - loss 0.52380828\n",
            "2019-07-23 20:26:17,091 epoch 53 - iter 21/32 - loss 0.52522847\n",
            "2019-07-23 20:26:18,695 epoch 53 - iter 24/32 - loss 0.53288838\n",
            "2019-07-23 20:26:19,669 epoch 53 - iter 27/32 - loss 0.52565563\n",
            "2019-07-23 20:26:20,443 epoch 53 - iter 30/32 - loss 0.51956074\n",
            "2019-07-23 20:26:20,850 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:26:20,853 EPOCH 53 done: loss 0.5276 - lr 0.0008 - bad epochs 2\n",
            "2019-07-23 20:26:22,291 DEV : loss 0.5263025164604187 - score 0.8333\n",
            "2019-07-23 20:26:23,718 TEST : loss 0.5263025164604187 - score 0.8333\n",
            "2019-07-23 20:26:33,104 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:26:34,242 epoch 54 - iter 0/32 - loss 0.39885530\n",
            "2019-07-23 20:26:35,264 epoch 54 - iter 3/32 - loss 0.45169710\n",
            "2019-07-23 20:26:37,143 epoch 54 - iter 6/32 - loss 0.47231399\n",
            "2019-07-23 20:26:38,133 epoch 54 - iter 9/32 - loss 0.48409487\n",
            "2019-07-23 20:26:39,047 epoch 54 - iter 12/32 - loss 0.51210301\n",
            "2019-07-23 20:26:39,996 epoch 54 - iter 15/32 - loss 0.50028793\n",
            "2019-07-23 20:26:41,441 epoch 54 - iter 18/32 - loss 0.49230357\n",
            "2019-07-23 20:26:42,469 epoch 54 - iter 21/32 - loss 0.48945669\n",
            "2019-07-23 20:26:43,164 epoch 54 - iter 24/32 - loss 0.50840734\n",
            "2019-07-23 20:26:44,055 epoch 54 - iter 27/32 - loss 0.51597175\n",
            "2019-07-23 20:26:44,948 epoch 54 - iter 30/32 - loss 0.51434253\n",
            "2019-07-23 20:26:45,371 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:26:45,373 EPOCH 54 done: loss 0.5299 - lr 0.0008 - bad epochs 3\n",
            "2019-07-23 20:26:46,677 DEV : loss 0.5252728462219238 - score 0.8333\n",
            "2019-07-23 20:26:48,987 TEST : loss 0.5252728462219238 - score 0.8333\n",
            "2019-07-23 20:26:58,172 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:26:59,234 epoch 55 - iter 0/32 - loss 0.49185288\n",
            "2019-07-23 20:27:00,600 epoch 55 - iter 3/32 - loss 0.50009235\n",
            "2019-07-23 20:27:01,527 epoch 55 - iter 6/32 - loss 0.56780163\n",
            "2019-07-23 20:27:02,408 epoch 55 - iter 9/32 - loss 0.56583146\n",
            "2019-07-23 20:27:04,274 epoch 55 - iter 12/32 - loss 0.55334389\n",
            "2019-07-23 20:27:05,015 epoch 55 - iter 15/32 - loss 0.54753877\n",
            "2019-07-23 20:27:05,807 epoch 55 - iter 18/32 - loss 0.52360790\n",
            "2019-07-23 20:27:06,802 epoch 55 - iter 21/32 - loss 0.52375214\n",
            "2019-07-23 20:27:07,656 epoch 55 - iter 24/32 - loss 0.51534131\n",
            "2019-07-23 20:27:09,159 epoch 55 - iter 27/32 - loss 0.52078427\n",
            "2019-07-23 20:27:09,988 epoch 55 - iter 30/32 - loss 0.52306543\n",
            "2019-07-23 20:27:10,411 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:27:10,412 EPOCH 55 done: loss 0.5262 - lr 0.0008 - bad epochs 4\n",
            "2019-07-23 20:27:11,773 DEV : loss 0.5253033638000488 - score 0.8333\n",
            "2019-07-23 20:27:13,165 TEST : loss 0.5253033638000488 - score 0.8333\n",
            "2019-07-23 20:27:22,371 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:27:23,416 epoch 56 - iter 0/32 - loss 0.41700375\n",
            "2019-07-23 20:27:24,372 epoch 56 - iter 3/32 - loss 0.52573249\n",
            "2019-07-23 20:27:26,172 epoch 56 - iter 6/32 - loss 0.49726685\n",
            "2019-07-23 20:27:27,237 epoch 56 - iter 9/32 - loss 0.49566107\n",
            "2019-07-23 20:27:28,022 epoch 56 - iter 12/32 - loss 0.50461383\n",
            "2019-07-23 20:27:28,910 epoch 56 - iter 15/32 - loss 0.50086040\n",
            "2019-07-23 20:27:29,672 epoch 56 - iter 18/32 - loss 0.50331163\n",
            "2019-07-23 20:27:31,229 epoch 56 - iter 21/32 - loss 0.49689222\n",
            "2019-07-23 20:27:32,094 epoch 56 - iter 24/32 - loss 0.50657665\n",
            "2019-07-23 20:27:33,038 epoch 56 - iter 27/32 - loss 0.52020342\n",
            "2019-07-23 20:27:33,869 epoch 56 - iter 30/32 - loss 0.51982121\n",
            "2019-07-23 20:27:34,227 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:27:34,228 EPOCH 56 done: loss 0.5219 - lr 0.0008 - bad epochs 5\n",
            "2019-07-23 20:27:35,581 DEV : loss 0.5250918865203857 - score 0.8333\n",
            "2019-07-23 20:27:36,940 TEST : loss 0.5250918865203857 - score 0.8333\n",
            "Epoch    55: reducing learning rate of group 0 to 3.9063e-04.\n",
            "2019-07-23 20:27:45,963 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:27:48,051 epoch 57 - iter 0/32 - loss 0.43387777\n",
            "2019-07-23 20:27:49,045 epoch 57 - iter 3/32 - loss 0.48017859\n",
            "2019-07-23 20:27:49,925 epoch 57 - iter 6/32 - loss 0.48831849\n",
            "2019-07-23 20:27:50,769 epoch 57 - iter 9/32 - loss 0.49332198\n",
            "2019-07-23 20:27:51,442 epoch 57 - iter 12/32 - loss 0.51110225\n",
            "2019-07-23 20:27:52,993 epoch 57 - iter 15/32 - loss 0.52640584\n",
            "2019-07-23 20:27:53,762 epoch 57 - iter 18/32 - loss 0.52792509\n",
            "2019-07-23 20:27:54,801 epoch 57 - iter 21/32 - loss 0.52864364\n",
            "2019-07-23 20:27:55,717 epoch 57 - iter 24/32 - loss 0.52432273\n",
            "2019-07-23 20:27:56,594 epoch 57 - iter 27/32 - loss 0.52321542\n",
            "2019-07-23 20:27:58,082 epoch 57 - iter 30/32 - loss 0.51737326\n",
            "2019-07-23 20:27:58,485 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:27:58,489 EPOCH 57 done: loss 0.5174 - lr 0.0004 - bad epochs 0\n",
            "2019-07-23 20:27:59,821 DEV : loss 0.5253920555114746 - score 0.8333\n",
            "2019-07-23 20:28:01,200 TEST : loss 0.5253920555114746 - score 0.8333\n",
            "2019-07-23 20:28:10,304 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:28:11,716 epoch 58 - iter 0/32 - loss 0.41252714\n",
            "2019-07-23 20:28:13,330 epoch 58 - iter 3/32 - loss 0.45155925\n",
            "2019-07-23 20:28:13,892 epoch 58 - iter 6/32 - loss 0.47707791\n",
            "2019-07-23 20:28:14,797 epoch 58 - iter 9/32 - loss 0.50117846\n",
            "2019-07-23 20:28:15,634 epoch 58 - iter 12/32 - loss 0.51141782\n",
            "2019-07-23 20:28:16,278 epoch 58 - iter 15/32 - loss 0.51663771\n",
            "2019-07-23 20:28:18,131 epoch 58 - iter 18/32 - loss 0.52967907\n",
            "2019-07-23 20:28:18,913 epoch 58 - iter 21/32 - loss 0.51848958\n",
            "2019-07-23 20:28:19,919 epoch 58 - iter 24/32 - loss 0.50724922\n",
            "2019-07-23 20:28:20,795 epoch 58 - iter 27/32 - loss 0.50890825\n",
            "2019-07-23 20:28:21,574 epoch 58 - iter 30/32 - loss 0.51407300\n",
            "2019-07-23 20:28:21,937 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:28:21,939 EPOCH 58 done: loss 0.5135 - lr 0.0004 - bad epochs 1\n",
            "2019-07-23 20:28:23,250 DEV : loss 0.5254231691360474 - score 0.8333\n",
            "2019-07-23 20:28:24,616 TEST : loss 0.5254231691360474 - score 0.8333\n",
            "2019-07-23 20:28:33,806 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:28:34,833 epoch 59 - iter 0/32 - loss 0.65126336\n",
            "2019-07-23 20:28:36,862 epoch 59 - iter 3/32 - loss 0.57301362\n",
            "2019-07-23 20:28:37,778 epoch 59 - iter 6/32 - loss 0.52391627\n",
            "2019-07-23 20:28:38,567 epoch 59 - iter 9/32 - loss 0.54835161\n",
            "2019-07-23 20:28:39,716 epoch 59 - iter 12/32 - loss 0.53486529\n",
            "2019-07-23 20:28:40,443 epoch 59 - iter 15/32 - loss 0.52916432\n",
            "2019-07-23 20:28:42,056 epoch 59 - iter 18/32 - loss 0.53568628\n",
            "2019-07-23 20:28:43,107 epoch 59 - iter 21/32 - loss 0.53019052\n",
            "2019-07-23 20:28:43,934 epoch 59 - iter 24/32 - loss 0.53054937\n",
            "2019-07-23 20:28:44,707 epoch 59 - iter 27/32 - loss 0.52562472\n",
            "2019-07-23 20:28:45,462 epoch 59 - iter 30/32 - loss 0.52149885\n",
            "2019-07-23 20:28:45,803 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:28:45,804 EPOCH 59 done: loss 0.5217 - lr 0.0004 - bad epochs 2\n",
            "2019-07-23 20:28:48,036 DEV : loss 0.5255278944969177 - score 0.8333\n",
            "2019-07-23 20:28:49,427 TEST : loss 0.5255278944969177 - score 0.8333\n",
            "2019-07-23 20:28:58,455 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:28:59,571 epoch 60 - iter 0/32 - loss 0.50579834\n",
            "2019-07-23 20:29:00,436 epoch 60 - iter 3/32 - loss 0.49858665\n",
            "2019-07-23 20:29:01,412 epoch 60 - iter 6/32 - loss 0.53323765\n",
            "2019-07-23 20:29:02,307 epoch 60 - iter 9/32 - loss 0.51388754\n",
            "2019-07-23 20:29:03,961 epoch 60 - iter 12/32 - loss 0.52050602\n",
            "2019-07-23 20:29:04,766 epoch 60 - iter 15/32 - loss 0.50961955\n",
            "2019-07-23 20:29:05,554 epoch 60 - iter 18/32 - loss 0.51184797\n",
            "2019-07-23 20:29:06,421 epoch 60 - iter 21/32 - loss 0.51163413\n",
            "2019-07-23 20:29:07,965 epoch 60 - iter 24/32 - loss 0.51058418\n",
            "2019-07-23 20:29:08,927 epoch 60 - iter 27/32 - loss 0.51189107\n",
            "2019-07-23 20:29:09,754 epoch 60 - iter 30/32 - loss 0.51664487\n",
            "2019-07-23 20:29:10,122 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:29:10,124 EPOCH 60 done: loss 0.5223 - lr 0.0004 - bad epochs 3\n",
            "2019-07-23 20:29:11,452 DEV : loss 0.5254830121994019 - score 0.8333\n",
            "2019-07-23 20:29:12,831 TEST : loss 0.5254830121994019 - score 0.8333\n",
            "2019-07-23 20:29:21,968 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:29:23,112 epoch 61 - iter 0/32 - loss 0.48792285\n",
            "2019-07-23 20:29:24,953 epoch 61 - iter 3/32 - loss 0.45281640\n",
            "2019-07-23 20:29:25,780 epoch 61 - iter 6/32 - loss 0.49263652\n",
            "2019-07-23 20:29:26,627 epoch 61 - iter 9/32 - loss 0.53392921\n",
            "2019-07-23 20:29:27,623 epoch 61 - iter 12/32 - loss 0.54430509\n",
            "2019-07-23 20:29:28,433 epoch 61 - iter 15/32 - loss 0.53907311\n",
            "2019-07-23 20:29:29,361 epoch 61 - iter 18/32 - loss 0.52509220\n",
            "2019-07-23 20:29:30,773 epoch 61 - iter 21/32 - loss 0.51807924\n",
            "2019-07-23 20:29:31,625 epoch 61 - iter 24/32 - loss 0.51654927\n",
            "2019-07-23 20:29:32,517 epoch 61 - iter 27/32 - loss 0.51482630\n",
            "2019-07-23 20:29:33,415 epoch 61 - iter 30/32 - loss 0.52284652\n",
            "2019-07-23 20:29:33,788 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:29:33,792 EPOCH 61 done: loss 0.5255 - lr 0.0004 - bad epochs 4\n",
            "2019-07-23 20:29:35,102 DEV : loss 0.5255597829818726 - score 0.8333\n",
            "2019-07-23 20:29:37,382 TEST : loss 0.5255597829818726 - score 0.8333\n",
            "2019-07-23 20:29:46,616 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:29:47,960 epoch 62 - iter 0/32 - loss 0.41138664\n",
            "2019-07-23 20:29:48,865 epoch 62 - iter 3/32 - loss 0.51750743\n",
            "2019-07-23 20:29:49,818 epoch 62 - iter 6/32 - loss 0.51161058\n",
            "2019-07-23 20:29:50,760 epoch 62 - iter 9/32 - loss 0.52162771\n",
            "2019-07-23 20:29:51,620 epoch 62 - iter 12/32 - loss 0.51529576\n",
            "2019-07-23 20:29:53,391 epoch 62 - iter 15/32 - loss 0.53392678\n",
            "2019-07-23 20:29:54,254 epoch 62 - iter 18/32 - loss 0.53375239\n",
            "2019-07-23 20:29:55,076 epoch 62 - iter 21/32 - loss 0.52037543\n",
            "2019-07-23 20:29:55,856 epoch 62 - iter 24/32 - loss 0.51125155\n",
            "2019-07-23 20:29:56,704 epoch 62 - iter 27/32 - loss 0.50757449\n",
            "2019-07-23 20:29:58,224 epoch 62 - iter 30/32 - loss 0.51211116\n",
            "2019-07-23 20:29:58,594 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:29:58,597 EPOCH 62 done: loss 0.5160 - lr 0.0004 - bad epochs 5\n",
            "2019-07-23 20:29:59,924 DEV : loss 0.5253886580467224 - score 0.8333\n",
            "2019-07-23 20:30:01,296 TEST : loss 0.5253886580467224 - score 0.8333\n",
            "Epoch    61: reducing learning rate of group 0 to 1.9531e-04.\n",
            "2019-07-23 20:30:10,375 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:30:11,439 epoch 63 - iter 0/32 - loss 0.48204410\n",
            "2019-07-23 20:30:12,323 epoch 63 - iter 3/32 - loss 0.48998653\n",
            "2019-07-23 20:30:14,112 epoch 63 - iter 6/32 - loss 0.52030800\n",
            "2019-07-23 20:30:15,135 epoch 63 - iter 9/32 - loss 0.49827109\n",
            "2019-07-23 20:30:16,062 epoch 63 - iter 12/32 - loss 0.50848610\n",
            "2019-07-23 20:30:16,898 epoch 63 - iter 15/32 - loss 0.49613437\n",
            "2019-07-23 20:30:17,863 epoch 63 - iter 18/32 - loss 0.50167271\n",
            "2019-07-23 20:30:19,537 epoch 63 - iter 21/32 - loss 0.50431397\n",
            "2019-07-23 20:30:20,450 epoch 63 - iter 24/32 - loss 0.51145795\n",
            "2019-07-23 20:30:21,272 epoch 63 - iter 27/32 - loss 0.51244622\n",
            "2019-07-23 20:30:21,968 epoch 63 - iter 30/32 - loss 0.51765800\n",
            "2019-07-23 20:30:22,362 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:30:22,368 EPOCH 63 done: loss 0.5147 - lr 0.0002 - bad epochs 0\n",
            "2019-07-23 20:30:23,711 DEV : loss 0.5255306959152222 - score 0.8333\n",
            "2019-07-23 20:30:25,935 TEST : loss 0.5255306959152222 - score 0.8333\n",
            "2019-07-23 20:30:34,851 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:30:35,944 epoch 64 - iter 0/32 - loss 0.51049358\n",
            "2019-07-23 20:30:36,857 epoch 64 - iter 3/32 - loss 0.54296656\n",
            "2019-07-23 20:30:37,957 epoch 64 - iter 6/32 - loss 0.56286517\n",
            "2019-07-23 20:30:38,844 epoch 64 - iter 9/32 - loss 0.53131162\n",
            "2019-07-23 20:30:40,813 epoch 64 - iter 12/32 - loss 0.51605125\n",
            "2019-07-23 20:30:41,674 epoch 64 - iter 15/32 - loss 0.52081151\n",
            "2019-07-23 20:30:42,520 epoch 64 - iter 18/32 - loss 0.51996120\n",
            "2019-07-23 20:30:43,458 epoch 64 - iter 21/32 - loss 0.52386710\n",
            "2019-07-23 20:30:44,256 epoch 64 - iter 24/32 - loss 0.52120733\n",
            "2019-07-23 20:30:45,070 epoch 64 - iter 27/32 - loss 0.52009740\n",
            "2019-07-23 20:30:46,590 epoch 64 - iter 30/32 - loss 0.51667528\n",
            "2019-07-23 20:30:46,973 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:30:46,974 EPOCH 64 done: loss 0.5176 - lr 0.0002 - bad epochs 1\n",
            "2019-07-23 20:30:48,322 DEV : loss 0.5256408452987671 - score 0.8333\n",
            "2019-07-23 20:30:49,705 TEST : loss 0.5256408452987671 - score 0.8333\n",
            "2019-07-23 20:30:58,978 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:31:00,315 epoch 65 - iter 0/32 - loss 0.65667474\n",
            "2019-07-23 20:31:01,214 epoch 65 - iter 3/32 - loss 0.54822194\n",
            "2019-07-23 20:31:02,908 epoch 65 - iter 6/32 - loss 0.55009072\n",
            "2019-07-23 20:31:03,923 epoch 65 - iter 9/32 - loss 0.54453444\n",
            "2019-07-23 20:31:04,852 epoch 65 - iter 12/32 - loss 0.56550852\n",
            "2019-07-23 20:31:05,618 epoch 65 - iter 15/32 - loss 0.55684212\n",
            "2019-07-23 20:31:06,435 epoch 65 - iter 18/32 - loss 0.54533851\n",
            "2019-07-23 20:31:08,097 epoch 65 - iter 21/32 - loss 0.53510972\n",
            "2019-07-23 20:31:08,953 epoch 65 - iter 24/32 - loss 0.52542166\n",
            "2019-07-23 20:31:09,751 epoch 65 - iter 27/32 - loss 0.51920887\n",
            "2019-07-23 20:31:10,706 epoch 65 - iter 30/32 - loss 0.51697150\n",
            "2019-07-23 20:31:11,098 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:31:11,100 EPOCH 65 done: loss 0.5185 - lr 0.0002 - bad epochs 2\n",
            "2019-07-23 20:31:12,443 DEV : loss 0.5256524085998535 - score 0.8333\n",
            "2019-07-23 20:31:13,808 TEST : loss 0.5256524085998535 - score 0.8333\n",
            "2019-07-23 20:31:22,818 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:31:23,951 epoch 66 - iter 0/32 - loss 0.59627736\n",
            "2019-07-23 20:31:25,941 epoch 66 - iter 3/32 - loss 0.47500651\n",
            "2019-07-23 20:31:26,946 epoch 66 - iter 6/32 - loss 0.47969865\n",
            "2019-07-23 20:31:27,863 epoch 66 - iter 9/32 - loss 0.50855870\n",
            "2019-07-23 20:31:28,820 epoch 66 - iter 12/32 - loss 0.52527084\n",
            "2019-07-23 20:31:30,373 epoch 66 - iter 15/32 - loss 0.52583583\n",
            "2019-07-23 20:31:31,188 epoch 66 - iter 18/32 - loss 0.53413671\n",
            "2019-07-23 20:31:32,053 epoch 66 - iter 21/32 - loss 0.53811292\n",
            "2019-07-23 20:31:32,867 epoch 66 - iter 24/32 - loss 0.52388730\n",
            "2019-07-23 20:31:33,707 epoch 66 - iter 27/32 - loss 0.51296288\n",
            "2019-07-23 20:31:34,502 epoch 66 - iter 30/32 - loss 0.52310173\n",
            "2019-07-23 20:31:34,847 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:31:34,848 EPOCH 66 done: loss 0.5195 - lr 0.0002 - bad epochs 3\n",
            "2019-07-23 20:31:37,120 DEV : loss 0.5257201194763184 - score 0.8333\n",
            "2019-07-23 20:31:38,541 TEST : loss 0.5257201194763184 - score 0.8333\n",
            "2019-07-23 20:31:47,673 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:31:48,768 epoch 67 - iter 0/32 - loss 0.52896541\n",
            "2019-07-23 20:31:49,900 epoch 67 - iter 3/32 - loss 0.55236156\n",
            "2019-07-23 20:31:50,714 epoch 67 - iter 6/32 - loss 0.51796997\n",
            "2019-07-23 20:31:52,420 epoch 67 - iter 9/32 - loss 0.51372840\n",
            "2019-07-23 20:31:53,502 epoch 67 - iter 12/32 - loss 0.50244138\n",
            "2019-07-23 20:31:54,260 epoch 67 - iter 15/32 - loss 0.50540506\n",
            "2019-07-23 20:31:55,024 epoch 67 - iter 18/32 - loss 0.52536502\n",
            "2019-07-23 20:31:56,574 epoch 67 - iter 21/32 - loss 0.52161521\n",
            "2019-07-23 20:31:57,366 epoch 67 - iter 24/32 - loss 0.51357284\n",
            "2019-07-23 20:31:58,326 epoch 67 - iter 27/32 - loss 0.51050631\n",
            "2019-07-23 20:31:59,219 epoch 67 - iter 30/32 - loss 0.51641302\n",
            "2019-07-23 20:31:59,550 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:31:59,551 EPOCH 67 done: loss 0.5129 - lr 0.0002 - bad epochs 4\n",
            "2019-07-23 20:32:00,912 DEV : loss 0.5258643627166748 - score 0.8333\n",
            "2019-07-23 20:32:02,255 TEST : loss 0.5258643627166748 - score 0.8333\n",
            "2019-07-23 20:32:11,383 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:32:12,594 epoch 68 - iter 0/32 - loss 0.54882646\n",
            "2019-07-23 20:32:14,629 epoch 68 - iter 3/32 - loss 0.43912799\n",
            "2019-07-23 20:32:15,410 epoch 68 - iter 6/32 - loss 0.46690912\n",
            "2019-07-23 20:32:16,333 epoch 68 - iter 9/32 - loss 0.48818274\n",
            "2019-07-23 20:32:17,182 epoch 68 - iter 12/32 - loss 0.50178451\n",
            "2019-07-23 20:32:17,956 epoch 68 - iter 15/32 - loss 0.50120921\n",
            "2019-07-23 20:32:19,607 epoch 68 - iter 18/32 - loss 0.51888802\n",
            "2019-07-23 20:32:20,431 epoch 68 - iter 21/32 - loss 0.52234259\n",
            "2019-07-23 20:32:21,297 epoch 68 - iter 24/32 - loss 0.50765350\n",
            "2019-07-23 20:32:22,156 epoch 68 - iter 27/32 - loss 0.51137578\n",
            "2019-07-23 20:32:22,962 epoch 68 - iter 30/32 - loss 0.52047583\n",
            "2019-07-23 20:32:23,313 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:32:23,315 EPOCH 68 done: loss 0.5179 - lr 0.0002 - bad epochs 5\n",
            "2019-07-23 20:32:25,563 DEV : loss 0.5259045958518982 - score 0.8333\n",
            "2019-07-23 20:32:26,952 TEST : loss 0.5259045958518982 - score 0.8333\n",
            "Epoch    67: reducing learning rate of group 0 to 9.7656e-05.\n",
            "2019-07-23 20:32:36,046 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:32:36,048 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:32:36,054 learning rate too small - quitting training!\n",
            "2019-07-23 20:32:36,055 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:32:44,765 ----------------------------------------------------------------------------------------------------\n",
            "2019-07-23 20:32:44,766 Testing using best model ...\n",
            "2019-07-23 20:32:44,767 loading file fall-train/best-model.pt\n",
            "2019-07-23 20:32:49,653 0.8333\t0.8333\t0.8333\n",
            "2019-07-23 20:32:49,654 \n",
            "MICRO_AVG: acc 0.7143 - f1-score 0.8333\n",
            "MACRO_AVG: acc 0.6647 - f1-score 0.7913\n",
            "0          tp: 50 - fp: 10 - fn: 3 - tn: 15 - precision: 0.8333 - recall: 0.9434 - accuracy: 0.7937 - f1-score: 0.8849\n",
            "1          tp: 15 - fp: 3 - fn: 10 - tn: 50 - precision: 0.8333 - recall: 0.6000 - accuracy: 0.5357 - f1-score: 0.6977\n",
            "2019-07-23 20:32:49,659 ----------------------------------------------------------------------------------------------------\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'dev_loss_history': [tensor(0.6339, device='cuda:0'),\n",
              "  tensor(0.6207, device='cuda:0'),\n",
              "  tensor(0.6689, device='cuda:0'),\n",
              "  tensor(0.5678, device='cuda:0'),\n",
              "  tensor(0.6289, device='cuda:0'),\n",
              "  tensor(0.5477, device='cuda:0'),\n",
              "  tensor(0.7136, device='cuda:0'),\n",
              "  tensor(0.5339, device='cuda:0'),\n",
              "  tensor(0.5686, device='cuda:0'),\n",
              "  tensor(0.6336, device='cuda:0'),\n",
              "  tensor(0.5893, device='cuda:0'),\n",
              "  tensor(0.5736, device='cuda:0'),\n",
              "  tensor(0.5254, device='cuda:0'),\n",
              "  tensor(0.5463, device='cuda:0'),\n",
              "  tensor(0.5204, device='cuda:0'),\n",
              "  tensor(0.5240, device='cuda:0'),\n",
              "  tensor(0.5568, device='cuda:0'),\n",
              "  tensor(0.5204, device='cuda:0'),\n",
              "  tensor(0.5429, device='cuda:0'),\n",
              "  tensor(0.5210, device='cuda:0'),\n",
              "  tensor(0.5266, device='cuda:0'),\n",
              "  tensor(0.5421, device='cuda:0'),\n",
              "  tensor(0.5210, device='cuda:0'),\n",
              "  tensor(0.5615, device='cuda:0'),\n",
              "  tensor(0.5354, device='cuda:0'),\n",
              "  tensor(0.5283, device='cuda:0'),\n",
              "  tensor(0.5248, device='cuda:0'),\n",
              "  tensor(0.5283, device='cuda:0'),\n",
              "  tensor(0.5425, device='cuda:0'),\n",
              "  tensor(0.5249, device='cuda:0'),\n",
              "  tensor(0.5219, device='cuda:0'),\n",
              "  tensor(0.5292, device='cuda:0'),\n",
              "  tensor(0.5234, device='cuda:0'),\n",
              "  tensor(0.5294, device='cuda:0'),\n",
              "  tensor(0.5298, device='cuda:0'),\n",
              "  tensor(0.5237, device='cuda:0'),\n",
              "  tensor(0.5281, device='cuda:0'),\n",
              "  tensor(0.5276, device='cuda:0'),\n",
              "  tensor(0.5262, device='cuda:0'),\n",
              "  tensor(0.5241, device='cuda:0'),\n",
              "  tensor(0.5252, device='cuda:0'),\n",
              "  tensor(0.5256, device='cuda:0'),\n",
              "  tensor(0.5263, device='cuda:0'),\n",
              "  tensor(0.5262, device='cuda:0'),\n",
              "  tensor(0.5267, device='cuda:0'),\n",
              "  tensor(0.5256, device='cuda:0'),\n",
              "  tensor(0.5255, device='cuda:0'),\n",
              "  tensor(0.5260, device='cuda:0'),\n",
              "  tensor(0.5254, device='cuda:0'),\n",
              "  tensor(0.5257, device='cuda:0'),\n",
              "  tensor(0.5259, device='cuda:0'),\n",
              "  tensor(0.5264, device='cuda:0'),\n",
              "  tensor(0.5263, device='cuda:0'),\n",
              "  tensor(0.5253, device='cuda:0'),\n",
              "  tensor(0.5253, device='cuda:0'),\n",
              "  tensor(0.5251, device='cuda:0'),\n",
              "  tensor(0.5254, device='cuda:0'),\n",
              "  tensor(0.5254, device='cuda:0'),\n",
              "  tensor(0.5255, device='cuda:0'),\n",
              "  tensor(0.5255, device='cuda:0'),\n",
              "  tensor(0.5256, device='cuda:0'),\n",
              "  tensor(0.5254, device='cuda:0'),\n",
              "  tensor(0.5255, device='cuda:0'),\n",
              "  tensor(0.5256, device='cuda:0'),\n",
              "  tensor(0.5257, device='cuda:0'),\n",
              "  tensor(0.5257, device='cuda:0'),\n",
              "  tensor(0.5259, device='cuda:0'),\n",
              "  tensor(0.5259, device='cuda:0')],\n",
              " 'dev_score_history': [0.6795,\n",
              "  0.6795,\n",
              "  0.6795,\n",
              "  0.7436,\n",
              "  0.7949,\n",
              "  0.6795,\n",
              "  0.6795,\n",
              "  0.8333,\n",
              "  0.6795,\n",
              "  0.6795,\n",
              "  0.6795,\n",
              "  0.6795,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8205,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8077,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333,\n",
              "  0.8333],\n",
              " 'test_score': 0.8333,\n",
              " 'train_loss_history': [0.6307457312941551,\n",
              "  0.6268004029989243,\n",
              "  0.611404137685895,\n",
              "  0.6068434910848737,\n",
              "  0.591563792899251,\n",
              "  0.5912030069157481,\n",
              "  0.5646141683682799,\n",
              "  0.5597005821764469,\n",
              "  0.5493824519217014,\n",
              "  0.5637702802196145,\n",
              "  0.5634729452431202,\n",
              "  0.5442605288699269,\n",
              "  0.549583344720304,\n",
              "  0.5447301166132092,\n",
              "  0.5308881802484393,\n",
              "  0.5433246660977602,\n",
              "  0.5352271134033799,\n",
              "  0.5372859435155988,\n",
              "  0.5432275095954537,\n",
              "  0.5353682432323694,\n",
              "  0.5294457292184234,\n",
              "  0.5310280248522758,\n",
              "  0.5318589964881539,\n",
              "  0.5330985533073545,\n",
              "  0.5220737494528294,\n",
              "  0.5382002471014857,\n",
              "  0.5307350102812052,\n",
              "  0.5229055732488632,\n",
              "  0.5214240876957774,\n",
              "  0.5269381990656257,\n",
              "  0.5215317970141768,\n",
              "  0.5276898508891463,\n",
              "  0.5318283122032881,\n",
              "  0.5178960934281349,\n",
              "  0.5187460333108902,\n",
              "  0.5262481328099966,\n",
              "  0.522157565690577,\n",
              "  0.518870634958148,\n",
              "  0.5149079505354166,\n",
              "  0.5236698612570763,\n",
              "  0.5153531739488244,\n",
              "  0.5235016057267785,\n",
              "  0.522446702234447,\n",
              "  0.5251729227602482,\n",
              "  0.5147639522328973,\n",
              "  0.5282929893583059,\n",
              "  0.5181664600968361,\n",
              "  0.5184130650013685,\n",
              "  0.5149418190121651,\n",
              "  0.5251589557155967,\n",
              "  0.5165025060996413,\n",
              "  0.5125883016735315,\n",
              "  0.5275941919535398,\n",
              "  0.5298729706555605,\n",
              "  0.5262252921238542,\n",
              "  0.5219339309260249,\n",
              "  0.5173761313781142,\n",
              "  0.5134787270799279,\n",
              "  0.5217456445097923,\n",
              "  0.5223123403266072,\n",
              "  0.5255416193976998,\n",
              "  0.5160467037931085,\n",
              "  0.5147049427032471,\n",
              "  0.5176337528973818,\n",
              "  0.5184633890166879,\n",
              "  0.5195271279662848,\n",
              "  0.5129460915923119,\n",
              "  0.5178976291790605]}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aiX7lPeXey1o",
        "colab_type": "text"
      },
      "source": [
        "## Plot Learning Curves"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LhcympJee0eP",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 659
        },
        "outputId": "6d087675-f847-47fe-93ea-040bdd1d5383"
      },
      "source": [
        "%matplotlib inline\n",
        "\n",
        "# 8. plot training curves (optional)\n",
        "from flair.visual.training_curves import Plotter\n",
        "plotter = Plotter()\n",
        "plotter.plot_training_curves('fall-train/loss.tsv')\n",
        "plotter.plot_weights('fall-train/weights.txt')"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAV8AAAKCCAYAAAB2/drtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nKaCCr6IoSgr8fy4t2+PQkz2xAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 360x648 with 8 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SAslIwicieHQ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}