main_notebook.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "from torch.utils.data import Dataset, DataLoader\n",
    "import matplotlib.pyplot as plt\n",
    "from importlib import reload\n",
    "from sklearn.metrics import f1_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "workers = 20\n",
    "btch_sz = 8000\n",
    "# ws = 0\n",
    "ws = 50\n",
    "sr = 0.5\n",
    "ngpu = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cuda:0\n"
     ]
    }
   ],
   "source": [
    "cuda = True if torch.cuda.is_available() else False\n",
    "FloatTensor = torch.cuda.FloatTensor if cuda else torch.FloatTensor\n",
    "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n",
    "print(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Data Preprocessing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv('data/train.csv')\n",
    "data_test = pd.read_csv('data/test.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "# data.iloc[:-100000].to_csv('data/train_new.csv', index=False)\n",
    "data.iloc[2400000:2500000].to_csv('data/test_new.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>signal</th>\n",
       "      <th>open_channels</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>open_channels_gessed</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>-1</th>\n",
       "      <td>160000</td>\n",
       "      <td>160000</td>\n",
       "      <td>160000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>28</td>\n",
       "      <td>28</td>\n",
       "      <td>28</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>239</td>\n",
       "      <td>239</td>\n",
       "      <td>239</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1643</td>\n",
       "      <td>1643</td>\n",
       "      <td>1643</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>7105</td>\n",
       "      <td>7105</td>\n",
       "      <td>7105</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>21740</td>\n",
       "      <td>21740</td>\n",
       "      <td>21740</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>44649</td>\n",
       "      <td>44649</td>\n",
       "      <td>44649</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>63692</td>\n",
       "      <td>63692</td>\n",
       "      <td>63692</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>58856</td>\n",
       "      <td>58856</td>\n",
       "      <td>58856</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>33411</td>\n",
       "      <td>33411</td>\n",
       "      <td>33411</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>8637</td>\n",
       "      <td>8637</td>\n",
       "      <td>8637</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                        time  signal  open_channels\n",
       "open_channels_gessed                               \n",
       "-1                    160000  160000         160000\n",
       " 1                        28      28             28\n",
       " 2                       239     239            239\n",
       " 3                      1643    1643           1643\n",
       " 4                      7105    7105           7105\n",
       " 5                     21740   21740          21740\n",
       " 6                     44649   44649          44649\n",
       " 7                     63692   63692          63692\n",
       " 8                     58856   58856          58856\n",
       " 9                     33411   33411          33411\n",
       " 10                     8637    8637           8637"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_new.groupby(['open_channels_gessed']).count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([7, 6, 4, 6, 8])"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "n_unique = train_new.groupby(['open_channels_gessed']).count().open_channels.values[1:]\n",
    "probs = n_unique/(len(train_new)-int(len(train_new)*0.4))\n",
    "np.random.choice(10, 5, p=probs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'data_test' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-42-b8e6fa67fbc9>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      2\u001b[0m     \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0maxvline\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m500000\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcolor\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;34m'red'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mxlim\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;36m5000000\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 4\u001b[0;31m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mplot\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata_test\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msignal\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      5\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshow\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'data_test' is not defined"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in range(10):\n",
    "    plt.axvline(500000*i, color='red')\n",
    "plt.xlim(0,5000000)\n",
    "plt.plot(data_test.signal)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# AAAAAAAAAAAAAAAAAAAAAAAA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "nb_train = 5\n",
    "nb_test = 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv('data/train_shifted_true.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = train.iloc[np.r_[0:3600000, 3850000:len(train)]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "train.to_csv('data/train_new.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "/* Put everything inside the global mpl namespace */\n",
       "window.mpl = {};\n",
       "\n",
       "\n",
       "mpl.get_websocket_type = function() {\n",
       "    if (typeof(WebSocket) !== 'undefined') {\n",
       "        return WebSocket;\n",
       "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
       "        return MozWebSocket;\n",
       "    } else {\n",
       "        alert('Your browser does not have WebSocket support. ' +\n",
       "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
       "              'Firefox 4 and 5 are also supported but you ' +\n",
       "              'have to enable WebSockets in about:config.');\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
       "    this.id = figure_id;\n",
       "\n",
       "    this.ws = websocket;\n",
       "\n",
       "    this.supports_binary = (this.ws.binaryType != undefined);\n",
       "\n",
       "    if (!this.supports_binary) {\n",
       "        var warnings = document.getElementById(\"mpl-warnings\");\n",
       "        if (warnings) {\n",
       "            warnings.style.display = 'block';\n",
       "            warnings.textContent = (\n",
       "                \"This browser does not support binary websocket messages. \" +\n",
       "                    \"Performance may be slow.\");\n",
       "        }\n",
       "    }\n",
       "\n",
       "    this.imageObj = new Image();\n",
       "\n",
       "    this.context = undefined;\n",
       "    this.message = undefined;\n",
       "    this.canvas = undefined;\n",
       "    this.rubberband_canvas = undefined;\n",
       "    this.rubberband_context = undefined;\n",
       "    this.format_dropdown = undefined;\n",
       "\n",
       "    this.image_mode = 'full';\n",
       "\n",
       "    this.root = $('<div/>');\n",
       "    this._root_extra_style(this.root)\n",
       "    this.root.attr('style', 'display: inline-block');\n",
       "\n",
       "    $(parent_element).append(this.root);\n",
       "\n",
       "    this._init_header(this);\n",
       "    this._init_canvas(this);\n",
       "    this._init_toolbar(this);\n",
       "\n",
       "    var fig = this;\n",
       "\n",
       "    this.waiting = false;\n",
       "\n",
       "    this.ws.onopen =  function () {\n",
       "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
       "            fig.send_message(\"send_image_mode\", {});\n",
       "            if (mpl.ratio != 1) {\n",
       "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
       "            }\n",
       "            fig.send_message(\"refresh\", {});\n",
       "        }\n",
       "\n",
       "    this.imageObj.onload = function() {\n",
       "            if (fig.image_mode == 'full') {\n",
       "                // Full images could contain transparency (where diff images\n",
       "                // almost always do), so we need to clear the canvas so that\n",
       "                // there is no ghosting.\n",
       "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
       "            }\n",
       "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
       "        };\n",
       "\n",
       "    this.imageObj.onunload = function() {\n",
       "        fig.ws.close();\n",
       "    }\n",
       "\n",
       "    this.ws.onmessage = this._make_on_message_function(this);\n",
       "\n",
       "    this.ondownload = ondownload;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_header = function() {\n",
       "    var titlebar = $(\n",
       "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
       "        'ui-helper-clearfix\"/>');\n",
       "    var titletext = $(\n",
       "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
       "        'text-align: center; padding: 3px;\"/>');\n",
       "    titlebar.append(titletext)\n",
       "    this.root.append(titlebar);\n",
       "    this.header = titletext[0];\n",
       "}\n",
       "\n",
       "\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_canvas = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var canvas_div = $('<div/>');\n",
       "\n",
       "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
       "\n",
       "    function canvas_keyboard_event(event) {\n",
       "        return fig.key_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
       "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
       "    this.canvas_div = canvas_div\n",
       "    this._canvas_extra_style(canvas_div)\n",
       "    this.root.append(canvas_div);\n",
       "\n",
       "    var canvas = $('<canvas/>');\n",
       "    canvas.addClass('mpl-canvas');\n",
       "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
       "\n",
       "    this.canvas = canvas[0];\n",
       "    this.context = canvas[0].getContext(\"2d\");\n",
       "\n",
       "    var backingStore = this.context.backingStorePixelRatio ||\n",
       "\tthis.context.webkitBackingStorePixelRatio ||\n",
       "\tthis.context.mozBackingStorePixelRatio ||\n",
       "\tthis.context.msBackingStorePixelRatio ||\n",
       "\tthis.context.oBackingStorePixelRatio ||\n",
       "\tthis.context.backingStorePixelRatio || 1;\n",
       "\n",
       "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
       "\n",
       "    var rubberband = $('<canvas/>');\n",
       "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
       "\n",
       "    var pass_mouse_events = true;\n",
       "\n",
       "    canvas_div.resizable({\n",
       "        start: function(event, ui) {\n",
       "            pass_mouse_events = false;\n",
       "        },\n",
       "        resize: function(event, ui) {\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "        stop: function(event, ui) {\n",
       "            pass_mouse_events = true;\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "    });\n",
       "\n",
       "    function mouse_event_fn(event) {\n",
       "        if (pass_mouse_events)\n",
       "            return fig.mouse_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    rubberband.mousedown('button_press', mouse_event_fn);\n",
       "    rubberband.mouseup('button_release', mouse_event_fn);\n",
       "    // Throttle sequential mouse events to 1 every 20ms.\n",
       "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
       "\n",
       "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
       "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
       "\n",
       "    canvas_div.on(\"wheel\", function (event) {\n",
       "        event = event.originalEvent;\n",
       "        event['data'] = 'scroll'\n",
       "        if (event.deltaY < 0) {\n",
       "            event.step = 1;\n",
       "        } else {\n",
       "            event.step = -1;\n",
       "        }\n",
       "        mouse_event_fn(event);\n",
       "    });\n",
       "\n",
       "    canvas_div.append(canvas);\n",
       "    canvas_div.append(rubberband);\n",
       "\n",
       "    this.rubberband = rubberband;\n",
       "    this.rubberband_canvas = rubberband[0];\n",
       "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
       "    this.rubberband_context.strokeStyle = \"#000000\";\n",
       "\n",
       "    this._resize_canvas = function(width, height) {\n",
       "        // Keep the size of the canvas, canvas container, and rubber band\n",
       "        // canvas in synch.\n",
       "        canvas_div.css('width', width)\n",
       "        canvas_div.css('height', height)\n",
       "\n",
       "        canvas.attr('width', width * mpl.ratio);\n",
       "        canvas.attr('height', height * mpl.ratio);\n",
       "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
       "\n",
       "        rubberband.attr('width', width);\n",
       "        rubberband.attr('height', height);\n",
       "    }\n",
       "\n",
       "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
       "    // upon first draw.\n",
       "    this._resize_canvas(600, 600);\n",
       "\n",
       "    // Disable right mouse context menu.\n",
       "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
       "        return false;\n",
       "    });\n",
       "\n",
       "    function set_focus () {\n",
       "        canvas.focus();\n",
       "        canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    window.setTimeout(set_focus, 100);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>');\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items) {\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) {\n",
       "            // put a spacer in here.\n",
       "            continue;\n",
       "        }\n",
       "        var button = $('<button/>');\n",
       "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
       "                        'ui-button-icon-only');\n",
       "        button.attr('role', 'button');\n",
       "        button.attr('aria-disabled', 'false');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "\n",
       "        var icon_img = $('<span/>');\n",
       "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
       "        icon_img.addClass(image);\n",
       "        icon_img.addClass('ui-corner-all');\n",
       "\n",
       "        var tooltip_span = $('<span/>');\n",
       "        tooltip_span.addClass('ui-button-text');\n",
       "        tooltip_span.html(tooltip);\n",
       "\n",
       "        button.append(icon_img);\n",
       "        button.append(tooltip_span);\n",
       "\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    var fmt_picker_span = $('<span/>');\n",
       "\n",
       "    var fmt_picker = $('<select/>');\n",
       "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
       "    fmt_picker_span.append(fmt_picker);\n",
       "    nav_element.append(fmt_picker_span);\n",
       "    this.format_dropdown = fmt_picker[0];\n",
       "\n",
       "    for (var ind in mpl.extensions) {\n",
       "        var fmt = mpl.extensions[ind];\n",
       "        var option = $(\n",
       "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
       "        fmt_picker.append(option);\n",
       "    }\n",
       "\n",
       "    // Add hover states to the ui-buttons\n",
       "    $( \".ui-button\" ).hover(\n",
       "        function() { $(this).addClass(\"ui-state-hover\");},\n",
       "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
       "    );\n",
       "\n",
       "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
       "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
       "    // which will in turn request a refresh of the image.\n",
       "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_message = function(type, properties) {\n",
       "    properties['type'] = type;\n",
       "    properties['figure_id'] = this.id;\n",
       "    this.ws.send(JSON.stringify(properties));\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_draw_message = function() {\n",
       "    if (!this.waiting) {\n",
       "        this.waiting = true;\n",
       "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
       "    }\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    var format_dropdown = fig.format_dropdown;\n",
       "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
       "    fig.ondownload(fig, format);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
       "    var size = msg['size'];\n",
       "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
       "        fig._resize_canvas(size[0], size[1]);\n",
       "        fig.send_message(\"refresh\", {});\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
       "    var x0 = msg['x0'] / mpl.ratio;\n",
       "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
       "    var x1 = msg['x1'] / mpl.ratio;\n",
       "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
       "    x0 = Math.floor(x0) + 0.5;\n",
       "    y0 = Math.floor(y0) + 0.5;\n",
       "    x1 = Math.floor(x1) + 0.5;\n",
       "    y1 = Math.floor(y1) + 0.5;\n",
       "    var min_x = Math.min(x0, x1);\n",
       "    var min_y = Math.min(y0, y1);\n",
       "    var width = Math.abs(x1 - x0);\n",
       "    var height = Math.abs(y1 - y0);\n",
       "\n",
       "    fig.rubberband_context.clearRect(\n",
       "        0, 0, fig.canvas.width / mpl.ratio, fig.canvas.height / mpl.ratio);\n",
       "\n",
       "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
       "    // Updates the figure title.\n",
       "    fig.header.textContent = msg['label'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
       "    var cursor = msg['cursor'];\n",
       "    switch(cursor)\n",
       "    {\n",
       "    case 0:\n",
       "        cursor = 'pointer';\n",
       "        break;\n",
       "    case 1:\n",
       "        cursor = 'default';\n",
       "        break;\n",
       "    case 2:\n",
       "        cursor = 'crosshair';\n",
       "        break;\n",
       "    case 3:\n",
       "        cursor = 'move';\n",
       "        break;\n",
       "    }\n",
       "    fig.rubberband_canvas.style.cursor = cursor;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
       "    fig.message.textContent = msg['message'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
       "    // Request the server to send over a new figure.\n",
       "    fig.send_draw_message();\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
       "    fig.image_mode = msg['mode'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Called whenever the canvas gets updated.\n",
       "    this.send_message(\"ack\", {});\n",
       "}\n",
       "\n",
       "// A function to construct a web socket function for onmessage handling.\n",
       "// Called in the figure constructor.\n",
       "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
       "    return function socket_on_message(evt) {\n",
       "        if (evt.data instanceof Blob) {\n",
       "            /* FIXME: We get \"Resource interpreted as Image but\n",
       "             * transferred with MIME type text/plain:\" errors on\n",
       "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
       "             * to be part of the websocket stream */\n",
       "            evt.data.type = \"image/png\";\n",
       "\n",
       "            /* Free the memory for the previous frames */\n",
       "            if (fig.imageObj.src) {\n",
       "                (window.URL || window.webkitURL).revokeObjectURL(\n",
       "                    fig.imageObj.src);\n",
       "            }\n",
       "\n",
       "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
       "                evt.data);\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
       "            fig.imageObj.src = evt.data;\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        var msg = JSON.parse(evt.data);\n",
       "        var msg_type = msg['type'];\n",
       "\n",
       "        // Call the  \"handle_{type}\" callback, which takes\n",
       "        // the figure and JSON message as its only arguments.\n",
       "        try {\n",
       "            var callback = fig[\"handle_\" + msg_type];\n",
       "        } catch (e) {\n",
       "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        if (callback) {\n",
       "            try {\n",
       "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
       "                callback(fig, msg);\n",
       "            } catch (e) {\n",
       "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
       "            }\n",
       "        }\n",
       "    };\n",
       "}\n",
       "\n",
       "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
       "mpl.findpos = function(e) {\n",
       "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
       "    var targ;\n",
       "    if (!e)\n",
       "        e = window.event;\n",
       "    if (e.target)\n",
       "        targ = e.target;\n",
       "    else if (e.srcElement)\n",
       "        targ = e.srcElement;\n",
       "    if (targ.nodeType == 3) // defeat Safari bug\n",
       "        targ = targ.parentNode;\n",
       "\n",
       "    // jQuery normalizes the pageX and pageY\n",
       "    // pageX,Y are the mouse positions relative to the document\n",
       "    // offset() returns the position of the element relative to the document\n",
       "    var x = e.pageX - $(targ).offset().left;\n",
       "    var y = e.pageY - $(targ).offset().top;\n",
       "\n",
       "    return {\"x\": x, \"y\": y};\n",
       "};\n",
       "\n",
       "/*\n",
       " * return a copy of an object with only non-object keys\n",
       " * we need this to avoid circular references\n",
       " * http://stackoverflow.com/a/24161582/3208463\n",
       " */\n",
       "function simpleKeys (original) {\n",
       "  return Object.keys(original).reduce(function (obj, key) {\n",
       "    if (typeof original[key] !== 'object')\n",
       "        obj[key] = original[key]\n",
       "    return obj;\n",
       "  }, {});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.mouse_event = function(event, name) {\n",
       "    var canvas_pos = mpl.findpos(event)\n",
       "\n",
       "    if (name === 'button_press')\n",
       "    {\n",
       "        this.canvas.focus();\n",
       "        this.canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    var x = canvas_pos.x * mpl.ratio;\n",
       "    var y = canvas_pos.y * mpl.ratio;\n",
       "\n",
       "    this.send_message(name, {x: x, y: y, button: event.button,\n",
       "                             step: event.step,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "\n",
       "    /* This prevents the web browser from automatically changing to\n",
       "     * the text insertion cursor when the button is pressed.  We want\n",
       "     * to control all of the cursor setting manually through the\n",
       "     * 'cursor' event from matplotlib */\n",
       "    event.preventDefault();\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    // Handle any extra behaviour associated with a key event\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.key_event = function(event, name) {\n",
       "\n",
       "    // Prevent repeat events\n",
       "    if (name == 'key_press')\n",
       "    {\n",
       "        if (event.which === this._key)\n",
       "            return;\n",
       "        else\n",
       "            this._key = event.which;\n",
       "    }\n",
       "    if (name == 'key_release')\n",
       "        this._key = null;\n",
       "\n",
       "    var value = '';\n",
       "    if (event.ctrlKey && event.which != 17)\n",
       "        value += \"ctrl+\";\n",
       "    if (event.altKey && event.which != 18)\n",
       "        value += \"alt+\";\n",
       "    if (event.shiftKey && event.which != 16)\n",
       "        value += \"shift+\";\n",
       "\n",
       "    value += 'k';\n",
       "    value += event.which.toString();\n",
       "\n",
       "    this._key_event_extra(event, name);\n",
       "\n",
       "    this.send_message(name, {key: value,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
       "    if (name == 'download') {\n",
       "        this.handle_save(this, null);\n",
       "    } else {\n",
       "        this.send_message(\"toolbar_button\", {name: name});\n",
       "    }\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
       "    this.message.textContent = tooltip;\n",
       "};\n",
       "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
       "\n",
       "mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
       "\n",
       "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
       "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
       "    // object with the appropriate methods. Currently this is a non binary\n",
       "    // socket, so there is still some room for performance tuning.\n",
       "    var ws = {};\n",
       "\n",
       "    ws.close = function() {\n",
       "        comm.close()\n",
       "    };\n",
       "    ws.send = function(m) {\n",
       "        //console.log('sending', m);\n",
       "        comm.send(m);\n",
       "    };\n",
       "    // Register the callback with on_msg.\n",
       "    comm.on_msg(function(msg) {\n",
       "        //console.log('receiving', msg['content']['data'], msg);\n",
       "        // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
       "        ws.onmessage(msg['content']['data'])\n",
       "    });\n",
       "    return ws;\n",
       "}\n",
       "\n",
       "mpl.mpl_figure_comm = function(comm, msg) {\n",
       "    // This is the function which gets called when the mpl process\n",
       "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
       "\n",
       "    var id = msg.content.data.id;\n",
       "    // Get hold of the div created by the display call when the Comm\n",
       "    // socket was opened in Python.\n",
       "    var element = $(\"#\" + id);\n",
       "    var ws_proxy = comm_websocket_adapter(comm)\n",
       "\n",
       "    function ondownload(figure, format) {\n",
       "        window.open(figure.imageObj.src);\n",
       "    }\n",
       "\n",
       "    var fig = new mpl.figure(id, ws_proxy,\n",
       "                           ondownload,\n",
       "                           element.get(0));\n",
       "\n",
       "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
       "    // web socket which is closed, not our websocket->open comm proxy.\n",
       "    ws_proxy.onopen();\n",
       "\n",
       "    fig.parent_element = element.get(0);\n",
       "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
       "    if (!fig.cell_info) {\n",
       "        console.error(\"Failed to find cell for figure\", id, fig);\n",
       "        return;\n",
       "    }\n",
       "\n",
       "    var output_index = fig.cell_info[2]\n",
       "    var cell = fig.cell_info[0];\n",
       "\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
       "    var width = fig.canvas.width/mpl.ratio\n",
       "    fig.root.unbind('remove')\n",
       "\n",
       "    // Update the output cell to use the data from the current canvas.\n",
       "    fig.push_to_output();\n",
       "    var dataURL = fig.canvas.toDataURL();\n",
       "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
       "    // the notebook keyboard shortcuts fail.\n",
       "    IPython.keyboard_manager.enable()\n",
       "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
       "    fig.close_ws(fig, msg);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.close_ws = function(fig, msg){\n",
       "    fig.send_message('closing', msg);\n",
       "    // fig.ws.close()\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
       "    // Turn the data on the canvas into data in the output cell.\n",
       "    var width = this.canvas.width/mpl.ratio\n",
       "    var dataURL = this.canvas.toDataURL();\n",
       "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Tell IPython that the notebook contents must change.\n",
       "    IPython.notebook.set_dirty(true);\n",
       "    this.send_message(\"ack\", {});\n",
       "    var fig = this;\n",
       "    // Wait a second, then push the new image to the DOM so\n",
       "    // that it is saved nicely (might be nice to debounce this).\n",
       "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>');\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items){\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) { continue; };\n",
       "\n",
       "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    // Add the status bar.\n",
       "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "\n",
       "    // Add the close button to the window.\n",
       "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
       "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
       "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
       "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
       "    buttongrp.append(button);\n",
       "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
       "    titlebar.prepend(buttongrp);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(el){\n",
       "    var fig = this\n",
       "    el.on(\"remove\", function(){\n",
       "\tfig.close_ws(fig, {});\n",
       "    });\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(el){\n",
       "    // this is important to make the div 'focusable\n",
       "    el.attr('tabindex', 0)\n",
       "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
       "    // off when our div gets focus\n",
       "\n",
       "    // location in version 3\n",
       "    if (IPython.notebook.keyboard_manager) {\n",
       "        IPython.notebook.keyboard_manager.register_events(el);\n",
       "    }\n",
       "    else {\n",
       "        // location in version 2\n",
       "        IPython.keyboard_manager.register_events(el);\n",
       "    }\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    var manager = IPython.notebook.keyboard_manager;\n",
       "    if (!manager)\n",
       "        manager = IPython.keyboard_manager;\n",
       "\n",
       "    // Check for shift+enter\n",
       "    if (event.shiftKey && event.which == 13) {\n",
       "        this.canvas_div.blur();\n",
       "        // select the cell after this one\n",
       "        var index = IPython.notebook.find_cell_index(this.cell_info[0]);\n",
       "        IPython.notebook.select(index + 1);\n",
       "    }\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    fig.ondownload(fig, null);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.find_output_cell = function(html_output) {\n",
       "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
       "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
       "    // IPython event is triggered only after the cells have been serialised, which for\n",
       "    // our purposes (turning an active figure into a static one), is too late.\n",
       "    var cells = IPython.notebook.get_cells();\n",
       "    var ncells = cells.length;\n",
       "    for (var i=0; i<ncells; i++) {\n",
       "        var cell = cells[i];\n",
       "        if (cell.cell_type === 'code'){\n",
       "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
       "                var data = cell.output_area.outputs[j];\n",
       "                if (data.data) {\n",
       "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
       "                    data = data.data;\n",
       "                }\n",
       "                if (data['text/html'] == html_output) {\n",
       "                    return [cell, data, j];\n",
       "                }\n",
       "            }\n",
       "        }\n",
       "    }\n",
       "}\n",
       "\n",
       "// Register the function which deals with the matplotlib target/channel.\n",
       "// The kernel may be null if the page has been refreshed.\n",
       "if (IPython.notebook.kernel != null) {\n",
       "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
       "}\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,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\" width=\"640\">"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib notebook\n",
    "plt.plot(train.signal)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = pd.read_csv('data/train_shifted_true.csv')\n",
    "test = pd.read_csv('data/test_shifted_true.csv')\n",
    "\n",
    "dev = train[(train.batch == 10)]\n",
    "dev.to_csv('data/dev.csv', index=False)\n",
    "\n",
    "# train = train[(train.batch == 5) | (train.batch == 10)]\n",
    "train = train[train.batch == nb_train]\n",
    "train.to_csv('data/train_new.csv', index=False)\n",
    "\n",
    "test = test[test.batch == nb_test]\n",
    "test.to_csv('data/test.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# AAAAAAAAAAAAAAAAAAAAAAAAAAAAA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>signal</th>\n",
       "      <th>batch</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>open_channels</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1923</td>\n",
       "      <td>1923</td>\n",
       "      <td>1923</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>23611</td>\n",
       "      <td>23611</td>\n",
       "      <td>23611</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>119295</td>\n",
       "      <td>119295</td>\n",
       "      <td>119295</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>295567</td>\n",
       "      <td>295567</td>\n",
       "      <td>295567</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>373428</td>\n",
       "      <td>373428</td>\n",
       "      <td>373428</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>186176</td>\n",
       "      <td>186176</td>\n",
       "      <td>186176</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 time  signal   batch\n",
       "open_channels                        \n",
       "0                1923    1923    1923\n",
       "1               23611   23611   23611\n",
       "2              119295  119295  119295\n",
       "3              295567  295567  295567\n",
       "4              373428  373428  373428\n",
       "5              186176  186176  186176"
      ]
     },
     "execution_count": 161,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.groupby('open_channels').count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>signal</th>\n",
       "      <th>open_channels</th>\n",
       "      <th>batch</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2500000</th>\n",
       "      <td>250.0001</td>\n",
       "      <td>2.855500</td>\n",
       "      <td>5</td>\n",
       "      <td>6.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2500001</th>\n",
       "      <td>250.0002</td>\n",
       "      <td>3.090700</td>\n",
       "      <td>5</td>\n",
       "      <td>6.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2500002</th>\n",
       "      <td>250.0003</td>\n",
       "      <td>3.527700</td>\n",
       "      <td>5</td>\n",
       "      <td>6.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2500003</th>\n",
       "      <td>250.0004</td>\n",
       "      <td>3.982200</td>\n",
       "      <td>5</td>\n",
       "      <td>6.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2500004</th>\n",
       "      <td>250.0005</td>\n",
       "      <td>3.336800</td>\n",
       "      <td>5</td>\n",
       "      <td>6.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4499995</th>\n",
       "      <td>449.9996</td>\n",
       "      <td>1.908458</td>\n",
       "      <td>4</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4499996</th>\n",
       "      <td>449.9997</td>\n",
       "      <td>2.065390</td>\n",
       "      <td>4</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4499997</th>\n",
       "      <td>449.9998</td>\n",
       "      <td>1.742521</td>\n",
       "      <td>4</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4499998</th>\n",
       "      <td>449.9999</td>\n",
       "      <td>1.588753</td>\n",
       "      <td>4</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4499999</th>\n",
       "      <td>450.0000</td>\n",
       "      <td>0.603784</td>\n",
       "      <td>3</td>\n",
       "      <td>9.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1000000 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             time    signal  open_channels  batch\n",
       "2500000  250.0001  2.855500              5    6.0\n",
       "2500001  250.0002  3.090700              5    6.0\n",
       "2500002  250.0003  3.527700              5    6.0\n",
       "2500003  250.0004  3.982200              5    6.0\n",
       "2500004  250.0005  3.336800              5    6.0\n",
       "...           ...       ...            ...    ...\n",
       "4499995  449.9996  1.908458              4    9.0\n",
       "4499996  449.9997  2.065390              4    9.0\n",
       "4499997  449.9998  1.742521              4    9.0\n",
       "4499998  449.9999  1.588753              4    9.0\n",
       "4499999  450.0000  0.603784              3    9.0\n",
       "\n",
       "[1000000 rows x 4 columns]"
      ]
     },
     "execution_count": 162,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def getTrainRangesIx():    \n",
    "    train_ranges={}\n",
    "    s0=0\n",
    "    for i in range(11):\n",
    "        l=500000 #length\n",
    "        if(i==1):l=100000\n",
    "        if(i==2):l=400000\n",
    "        train_ranges[i]=np.arange(s0,s0+l)\n",
    "        s0+=l\n",
    "    return train_ranges\n",
    "\n",
    "# nBatches = 10\n",
    "# nCh = [2,2,2,2,4,10,6,2,4,6,10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for b in range(nBatches):\n",
    "        print(\"batch Nr: \", b, \" nCh=\",nCh[b])\n",
    "        batch_ixs = getTrainRangesIx()[b]\n",
    "        batch = data.loc[batch_ixs].groupby(['open_channels']).count().signal.values\n",
    "        print(batch/len(batch_ixs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "signal = np.array([train.signal.values[1:], train.signal.values[:-1]])\n",
    "left_avr = np.mean(signal, axis=0)[:-3]\n",
    "right_avr = np.mean(signal, axis=0)[3:]\n",
    "\n",
    "train_new = pd.DataFrame({'left_avr': left_avr, 'right_avr': right_avr})\n",
    "train_new['signal'] = train.signal.values[2:-2]\n",
    "train_new['open_channels'] = train.open_channels.values[2:-2]\n",
    "\n",
    "train_new.to_csv('data/train_new.csv', index=False)\n",
    "\n",
    "train_new"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_0 = pd.read_csv('data/git_data/test_5.csv')\n",
    "# test_0.s2 = test_0.signal\n",
    "test_0.to_csv('data/test.csv', index=False)\n",
    "test_0"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Format Dataloaders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dataUtils.CustomDataset import IonSwitchingDataset\n",
    "\n",
    "ion_dataset_train = IonSwitchingDataset('data/train_new.csv', window_size=ws, slice_ratio=sr)\n",
    "ion_dataset_test = IonSwitchingDataset('data/dev.csv', window_size=ws, slice_ratio=sr)\n",
    "ion_dataset_submmit = IonSwitchingDataset('data/test_shifted_true.csv', window_size=ws, slice_ratio=sr, train=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataloader = DataLoader(ion_dataset_train, batch_size=btch_sz, shuffle=True, num_workers=workers)\n",
    "dataloader_test = DataLoader(ion_dataset_test, batch_size=25000, shuffle=False, num_workers=workers)\n",
    "dataloader_submmit = DataLoader(ion_dataset_submmit, batch_size=5000, shuffle=False, num_workers=workers)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dataUtils.CustomDataset import SequensDataset2\n",
    "\n",
    "def getTestRangesIx():\n",
    "    test_ranges={}\n",
    "    for i in range(12):\n",
    "        if i<10:\n",
    "            l=100000 #length\n",
    "            s= i*l   #start\n",
    "            test_ranges[i]=np.arange(s,s+l)\n",
    "        else:\n",
    "            s+=l\n",
    "            l=500000\n",
    "            test_ranges[i]=np.arange(s,s+l)\n",
    "    return test_ranges\n",
    "\n",
    "def getTrainRangesIx():    \n",
    "    train_ranges={}\n",
    "    s0=0\n",
    "    for i in range(11):\n",
    "        l=500000 #length\n",
    "        if(i==1):l=100000\n",
    "        if(i==2):l=400000\n",
    "        train_ranges[i]=np.arange(s0,s0+l)\n",
    "        s0+=l\n",
    "    return train_ranges\n",
    "\n",
    "sequence_length = 25\n",
    "hyp_probThr = 0.95\n",
    "nCh_test =  {0:2, 1:4, 2:6, 3:4, 4:2, 5:11, 6:6, 7:11, 8:4, 9:4, 10:5, 11:4}\n",
    "nCh_train = {0:2, 1:2, 2:2, 3:2, 4:4, 5:11, 6:6, 7:2,  8:4, 9:6, 10:11}\n",
    "bNr_test = 5\n",
    "bNr_train = 5\n",
    "\n",
    "batch_ixs = getTrainRangesIx()[bNr_train]\n",
    "# batch_ixs = getTestRangesIx()[bNr_train]\n",
    "\n",
    "SequensDataset = SequensDataset2('data/train_thrCK.csv', seqL=sequence_length, probThr = hyp_probThr,\n",
    "                                  num_classes=nCh_train[bNr_train], batch_ixs=batch_ixs,train=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataloader = DataLoader(SequensDataset, batch_size=btch_sz, num_workers=workers, shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch = next(iter(dataloader))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch['signal'].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(torch.Size([8000, 1, 51]), torch.Size([8000, 11]))"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch['input'].shape, batch['open_channels'].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "plt.plot(batch['input'][0][0])\n",
    "plt.plot(batch['input'][1][0])\n",
    "plt.plot(batch['input'][2][0])\n",
    "# plt.plot(np.where(batch['open_channels']==1)[1][:10])\n",
    "# plt.plot(batch['input'][0][1])\n",
    "# plt.plot(batch['input'][0][2])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "824907\n"
     ]
    }
   ],
   "source": [
    "import models.FCC as FCC\n",
    "reload(FCC)\n",
    "\n",
    "model = FCC.OpenChannelsClassifier_CNN2(1)\n",
    "# model = FCC.OpenChannelsClassifier_FC(1)\n",
    "# model = FCC.Seq_CNN(1)\n",
    "model.apply(FCC.weights_init)\n",
    "print(sum(p.numel() for p in model.parameters()))\n",
    "model.to(device)\n",
    "model = torch.nn.DataParallel(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([8000, 11])"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model(batch['input'].to(device).float().view(btch_sz, 1, ws+1)).shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Optimirez and Criterion Loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr = 0.001\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=lr)\n",
    "# optimizer = torch.optim.SGD(model.parameters(), lr=lr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_losses = np.array([])\n",
    "train_accurasy = np.array([])\n",
    "train_f1 = np.array([])\n",
    "test_accurasy = np.array([])\n",
    "test_f1 = np.array([])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_test = next(iter(dataloader_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training Loop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0/30] [4/594], loss: 2.3072, accyrasy: 0.4414, f1: 0.1875  f1_dev: 0.0003\n",
      "[0/30] [9/594], loss: 1.6806, accyrasy: 0.4424, f1: 0.1653  f1_dev: 0.0011\n",
      "[0/30] [14/594], loss: 1.2378, accyrasy: 0.5444, f1: 0.2382  f1_dev: 0.0016\n",
      "[0/30] [19/594], loss: 1.0035, accyrasy: 0.6115, f1: 0.3096  f1_dev: 0.0011\n",
      "[0/30] [24/594], loss: 0.8741, accyrasy: 0.6569, f1: 0.4187  f1_dev: 0.0012\n",
      "[0/30] [29/594], loss: 0.8046, accyrasy: 0.6773, f1: 0.4392  f1_dev: 0.0037\n",
      "[0/30] [34/594], loss: 0.7621, accyrasy: 0.6789, f1: 0.4511  f1_dev: 0.017\n",
      "[0/30] [39/594], loss: 0.7037, accyrasy: 0.7127, f1: 0.5113  f1_dev: 0.0899\n",
      "[0/30] [44/594], loss: 0.6679, accyrasy: 0.7279, f1: 0.5306  f1_dev: 0.0688\n",
      "[0/30] [49/594], loss: 0.6253, accyrasy: 0.7418, f1: 0.5521  f1_dev: 0.0403\n",
      "[0/30] [54/594], loss: 0.5974, accyrasy: 0.7431, f1: 0.5701  f1_dev: 0.0346\n",
      "[0/30] [59/594], loss: 0.5655, accyrasy: 0.7776, f1: 0.6314  f1_dev: 0.0198\n",
      "[0/30] [64/594], loss: 0.5224, accyrasy: 0.7977, f1: 0.6332  f1_dev: 0.0295\n",
      "[0/30] [69/594], loss: 0.4667, accyrasy: 0.8324, f1: 0.7264  f1_dev: 0.1594\n",
      "[0/30] [74/594], loss: 0.4312, accyrasy: 0.8438, f1: 0.6945  f1_dev: 0.1535\n",
      "[0/30] [79/594], loss: 0.4049, accyrasy: 0.8586, f1: 0.7442  f1_dev: 0.2973\n",
      "[0/30] [84/594], loss: 0.3642, accyrasy: 0.8749, f1: 0.7265  f1_dev: 0.2313\n",
      "[0/30] [89/594], loss: 0.3856, accyrasy: 0.8226, f1: 0.6998  f1_dev: 0.3067\n",
      "[0/30] [94/594], loss: 0.3602, accyrasy: 0.8724, f1: 0.7739  f1_dev: 0.2674\n",
      "[0/30] [99/594], loss: 0.3235, accyrasy: 0.8894, f1: 0.8006  f1_dev: 0.3584\n",
      "[0/30] [104/594], loss: 0.293, accyrasy: 0.9034, f1: 0.8133  f1_dev: 0.3675\n",
      "[0/30] [109/594], loss: 0.2876, accyrasy: 0.9075, f1: 0.8382  f1_dev: 0.3571\n",
      "[0/30] [114/594], loss: 0.2658, accyrasy: 0.916, f1: 0.8514  f1_dev: 0.3809\n",
      "[0/30] [119/594], loss: 0.2538, accyrasy: 0.9133, f1: 0.8427  f1_dev: 0.3811\n",
      "[0/30] [124/594], loss: 0.2566, accyrasy: 0.9014, f1: 0.8291  f1_dev: 0.3858\n",
      "[0/30] [129/594], loss: 0.2466, accyrasy: 0.9241, f1: 0.8652  f1_dev: 0.3372\n",
      "[0/30] [134/594], loss: 0.236, accyrasy: 0.9191, f1: 0.8364  f1_dev: 0.3561\n",
      "[0/30] [139/594], loss: 0.2294, accyrasy: 0.925, f1: 0.8665  f1_dev: 0.4164\n",
      "[0/30] [144/594], loss: 0.2306, accyrasy: 0.9265, f1: 0.87  f1_dev: 0.3\n",
      "[0/30] [149/594], loss: 0.2318, accyrasy: 0.9243, f1: 0.85  f1_dev: 0.3697\n",
      "[0/30] [154/594], loss: 0.2375, accyrasy: 0.9273, f1: 0.8548  f1_dev: 0.4031\n",
      "[0/30] [159/594], loss: 0.215, accyrasy: 0.9376, f1: 0.8872  f1_dev: 0.4594\n",
      "[0/30] [164/594], loss: 0.1979, accyrasy: 0.9316, f1: 0.8602  f1_dev: 0.4632\n",
      "[0/30] [169/594], loss: 0.1986, accyrasy: 0.937, f1: 0.8887  f1_dev: 0.4348\n",
      "[0/30] [174/594], loss: 0.1888, accyrasy: 0.9353, f1: 0.8943  f1_dev: 0.4668\n",
      "[0/30] [179/594], loss: 0.1907, accyrasy: 0.9385, f1: 0.892  f1_dev: 0.4865\n",
      "[0/30] [184/594], loss: 0.1893, accyrasy: 0.9286, f1: 0.8575  f1_dev: 0.4842\n",
      "[0/30] [189/594], loss: 0.181, accyrasy: 0.9419, f1: 0.8984  f1_dev: 0.4532\n",
      "[0/30] [194/594], loss: 0.1735, accyrasy: 0.9426, f1: 0.9004  f1_dev: 0.4717\n",
      "[0/30] [199/594], loss: 0.1796, accyrasy: 0.9466, f1: 0.9055  f1_dev: 0.404\n",
      "[0/30] [204/594], loss: 0.1853, accyrasy: 0.9466, f1: 0.9058  f1_dev: 0.4499\n",
      "[0/30] [209/594], loss: 0.176, accyrasy: 0.9357, f1: 0.8418  f1_dev: 0.4607\n",
      "[0/30] [214/594], loss: 0.1705, accyrasy: 0.9416, f1: 0.898  f1_dev: 0.4936\n",
      "[0/30] [219/594], loss: 0.1616, accyrasy: 0.9481, f1: 0.9131  f1_dev: 0.4956\n",
      "[0/30] [224/594], loss: 0.17, accyrasy: 0.9325, f1: 0.8635  f1_dev: 0.4974\n",
      "[0/30] [229/594], loss: 0.1689, accyrasy: 0.9374, f1: 0.8873  f1_dev: 0.512\n",
      "[0/30] [234/594], loss: 0.1496, accyrasy: 0.9444, f1: 0.9033  f1_dev: 0.512\n",
      "[0/30] [239/594], loss: 0.1609, accyrasy: 0.9427, f1: 0.8895  f1_dev: 0.5039\n",
      "[0/30] [244/594], loss: 0.1455, accyrasy: 0.9517, f1: 0.9111  f1_dev: 0.5109\n",
      "[0/30] [249/594], loss: 0.1485, accyrasy: 0.9485, f1: 0.9067  f1_dev: 0.5058\n",
      "[0/30] [254/594], loss: 0.1422, accyrasy: 0.9456, f1: 0.9074  f1_dev: 0.4971\n",
      "[0/30] [259/594], loss: 0.1463, accyrasy: 0.9421, f1: 0.8886  f1_dev: 0.512\n",
      "[0/30] [264/594], loss: 0.1416, accyrasy: 0.9547, f1: 0.9172  f1_dev: 0.5146\n",
      "[0/30] [269/594], loss: 0.1371, accyrasy: 0.9537, f1: 0.9136  f1_dev: 0.5219\n",
      "[0/30] [274/594], loss: 0.1388, accyrasy: 0.9459, f1: 0.8996  f1_dev: 0.5349\n",
      "[0/30] [279/594], loss: 0.1551, accyrasy: 0.9496, f1: 0.9095  f1_dev: 0.5295\n",
      "[0/30] [284/594], loss: 0.1436, accyrasy: 0.9527, f1: 0.9101  f1_dev: 0.4793\n",
      "[0/30] [289/594], loss: 0.1499, accyrasy: 0.9511, f1: 0.9115  f1_dev: 0.5557\n",
      "[0/30] [294/594], loss: 0.1347, accyrasy: 0.9536, f1: 0.9179  f1_dev: 0.5609\n",
      "[0/30] [299/594], loss: 0.1294, accyrasy: 0.9525, f1: 0.9168  f1_dev: 0.5555\n",
      "[0/30] [304/594], loss: 0.1312, accyrasy: 0.9557, f1: 0.919  f1_dev: 0.575\n",
      "[0/30] [309/594], loss: 0.1317, accyrasy: 0.9531, f1: 0.9149  f1_dev: 0.56\n",
      "[0/30] [314/594], loss: 0.1484, accyrasy: 0.9386, f1: 0.8875  f1_dev: 0.5471\n",
      "[0/30] [319/594], loss: 0.1453, accyrasy: 0.9435, f1: 0.9054  f1_dev: 0.5729\n",
      "[0/30] [324/594], loss: 0.1384, accyrasy: 0.9546, f1: 0.9247  f1_dev: 0.566\n",
      "[0/30] [329/594], loss: 0.1366, accyrasy: 0.9511, f1: 0.9014  f1_dev: 0.5163\n",
      "[0/30] [334/594], loss: 0.1334, accyrasy: 0.9563, f1: 0.9282  f1_dev: 0.5767\n",
      "[0/30] [339/594], loss: 0.1271, accyrasy: 0.9507, f1: 0.9092  f1_dev: 0.5475\n",
      "[0/30] [344/594], loss: 0.1364, accyrasy: 0.9421, f1: 0.8985  f1_dev: 0.5327\n",
      "[0/30] [349/594], loss: 0.1437, accyrasy: 0.9426, f1: 0.9033  f1_dev: 0.5793\n",
      "[0/30] [354/594], loss: 0.1349, accyrasy: 0.9531, f1: 0.9176  f1_dev: 0.5738\n",
      "[0/30] [359/594], loss: 0.1262, accyrasy: 0.9541, f1: 0.9131  f1_dev: 0.5725\n",
      "[0/30] [364/594], loss: 0.127, accyrasy: 0.9579, f1: 0.9246  f1_dev: 0.5618\n",
      "[0/30] [369/594], loss: 0.1257, accyrasy: 0.9577, f1: 0.9193  f1_dev: 0.5651\n",
      "[0/30] [374/594], loss: 0.1245, accyrasy: 0.9527, f1: 0.9158  f1_dev: 0.5684\n",
      "[0/30] [379/594], loss: 0.1223, accyrasy: 0.9611, f1: 0.9274  f1_dev: 0.5711\n",
      "[0/30] [384/594], loss: 0.1216, accyrasy: 0.9531, f1: 0.9188  f1_dev: 0.5763\n",
      "[0/30] [389/594], loss: 0.127, accyrasy: 0.951, f1: 0.9027  f1_dev: 0.5747\n",
      "[0/30] [394/594], loss: 0.1246, accyrasy: 0.9574, f1: 0.9173  f1_dev: 0.5355\n",
      "[0/30] [399/594], loss: 0.1239, accyrasy: 0.9531, f1: 0.9048  f1_dev: 0.539\n",
      "[0/30] [404/594], loss: 0.1269, accyrasy: 0.9514, f1: 0.9032  f1_dev: 0.5772\n",
      "[0/30] [409/594], loss: 0.1195, accyrasy: 0.9585, f1: 0.9239  f1_dev: 0.5815\n",
      "[0/30] [414/594], loss: 0.1195, accyrasy: 0.961, f1: 0.9234  f1_dev: 0.5729\n",
      "[0/30] [419/594], loss: 0.1205, accyrasy: 0.957, f1: 0.9215  f1_dev: 0.5665\n",
      "[0/30] [424/594], loss: 0.1356, accyrasy: 0.9429, f1: 0.8701  f1_dev: 0.5664\n",
      "[0/30] [429/594], loss: 0.1317, accyrasy: 0.9589, f1: 0.9123  f1_dev: 0.5135\n",
      "[0/30] [434/594], loss: 0.1397, accyrasy: 0.9445, f1: 0.8865  f1_dev: 0.5763\n",
      "[0/30] [439/594], loss: 0.1365, accyrasy: 0.9447, f1: 0.8743  f1_dev: 0.5409\n",
      "[0/30] [444/594], loss: 0.1443, accyrasy: 0.9564, f1: 0.9216  f1_dev: 0.4617\n",
      "[0/30] [449/594], loss: 0.1551, accyrasy: 0.9395, f1: 0.8802  f1_dev: 0.5615\n",
      "[0/30] [454/594], loss: 0.1396, accyrasy: 0.9457, f1: 0.8832  f1_dev: 0.5785\n",
      "[0/30] [459/594], loss: 0.1276, accyrasy: 0.9491, f1: 0.8945  f1_dev: 0.5753\n",
      "[0/30] [464/594], loss: 0.1292, accyrasy: 0.9506, f1: 0.9074  f1_dev: 0.5758\n",
      "[0/30] [469/594], loss: 0.1388, accyrasy: 0.9579, f1: 0.9265  f1_dev: 0.5617\n",
      "[0/30] [474/594], loss: 0.1283, accyrasy: 0.9585, f1: 0.9216  f1_dev: 0.5765\n",
      "[0/30] [479/594], loss: 0.1352, accyrasy: 0.9483, f1: 0.8976  f1_dev: 0.581\n",
      "[0/30] [484/594], loss: 0.1207, accyrasy: 0.954, f1: 0.9125  f1_dev: 0.5801\n",
      "[0/30] [489/594], loss: 0.1292, accyrasy: 0.9533, f1: 0.906  f1_dev: 0.5585\n",
      "[0/30] [494/594], loss: 0.1313, accyrasy: 0.9605, f1: 0.9343  f1_dev: 0.5892\n",
      "[0/30] [499/594], loss: 0.1231, accyrasy: 0.9569, f1: 0.9165  f1_dev: 0.5775\n",
      "[0/30] [504/594], loss: 0.1236, accyrasy: 0.9566, f1: 0.9222  f1_dev: 0.572\n",
      "[0/30] [509/594], loss: 0.1285, accyrasy: 0.9509, f1: 0.9135  f1_dev: 0.5779\n",
      "[0/30] [514/594], loss: 0.1302, accyrasy: 0.9595, f1: 0.9236  f1_dev: 0.5949\n",
      "[0/30] [519/594], loss: 0.117, accyrasy: 0.9525, f1: 0.9125  f1_dev: 0.5598\n",
      "[0/30] [524/594], loss: 0.129, accyrasy: 0.954, f1: 0.9101  f1_dev: 0.5236\n",
      "[0/30] [529/594], loss: 0.1477, accyrasy: 0.9417, f1: 0.8726  f1_dev: 0.5606\n",
      "[0/30] [534/594], loss: 0.1396, accyrasy: 0.944, f1: 0.9095  f1_dev: 0.5955\n",
      "[0/30] [539/594], loss: 0.1371, accyrasy: 0.9584, f1: 0.918  f1_dev: 0.5631\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0/30] [544/594], loss: 0.1524, accyrasy: 0.9297, f1: 0.8694  f1_dev: 0.5735\n",
      "[0/30] [549/594], loss: 0.1283, accyrasy: 0.9581, f1: 0.9195  f1_dev: 0.5998\n",
      "[0/30] [554/594], loss: 0.1376, accyrasy: 0.949, f1: 0.901  f1_dev: 0.6142\n",
      "[0/30] [559/594], loss: 0.1313, accyrasy: 0.9535, f1: 0.9117  f1_dev: 0.5933\n",
      "[0/30] [564/594], loss: 0.1264, accyrasy: 0.9537, f1: 0.9114  f1_dev: 0.5549\n",
      "[0/30] [569/594], loss: 0.1228, accyrasy: 0.9566, f1: 0.9193  f1_dev: 0.5932\n",
      "[0/30] [574/594], loss: 0.1194, accyrasy: 0.958, f1: 0.9216  f1_dev: 0.575\n",
      "[0/30] [579/594], loss: 0.1221, accyrasy: 0.952, f1: 0.9167  f1_dev: 0.5792\n",
      "[0/30] [584/594], loss: 0.1216, accyrasy: 0.9561, f1: 0.9179  f1_dev: 0.6147\n",
      "[0/30] [589/594], loss: 0.1172, accyrasy: 0.9616, f1: 0.9275  f1_dev: 0.5784\n",
      "[1/30] [4/594], loss: 0.1144, accyrasy: 0.9587, f1: 0.9233  f1_dev: 0.5892\n",
      "[1/30] [9/594], loss: 0.119, accyrasy: 0.9574, f1: 0.9116  f1_dev: 0.5637\n",
      "[1/30] [14/594], loss: 0.1182, accyrasy: 0.9599, f1: 0.9215  f1_dev: 0.616\n",
      "[1/30] [19/594], loss: 0.1148, accyrasy: 0.9576, f1: 0.9243  f1_dev: 0.5956\n",
      "[1/30] [24/594], loss: 0.1151, accyrasy: 0.9559, f1: 0.9161  f1_dev: 0.5846\n",
      "[1/30] [29/594], loss: 0.1159, accyrasy: 0.9556, f1: 0.9202  f1_dev: 0.5881\n",
      "[1/30] [34/594], loss: 0.117, accyrasy: 0.9537, f1: 0.9233  f1_dev: 0.5809\n",
      "[1/30] [39/594], loss: 0.1189, accyrasy: 0.9627, f1: 0.9351  f1_dev: 0.5908\n",
      "[1/30] [44/594], loss: 0.1213, accyrasy: 0.953, f1: 0.9154  f1_dev: 0.6062\n",
      "[1/30] [49/594], loss: 0.1147, accyrasy: 0.955, f1: 0.9153  f1_dev: 0.5996\n",
      "[1/30] [54/594], loss: 0.1123, accyrasy: 0.9587, f1: 0.9243  f1_dev: 0.5928\n",
      "[1/30] [59/594], loss: 0.1112, accyrasy: 0.9597, f1: 0.9225  f1_dev: 0.6073\n",
      "[1/30] [64/594], loss: 0.1126, accyrasy: 0.9573, f1: 0.9216  f1_dev: 0.5975\n",
      "[1/30] [69/594], loss: 0.115, accyrasy: 0.9531, f1: 0.9178  f1_dev: 0.5967\n",
      "[1/30] [74/594], loss: 0.1251, accyrasy: 0.9553, f1: 0.9066  f1_dev: 0.5951\n",
      "[1/30] [79/594], loss: 0.1187, accyrasy: 0.9579, f1: 0.9193  f1_dev: 0.5772\n",
      "[1/30] [84/594], loss: 0.1283, accyrasy: 0.9465, f1: 0.8773  f1_dev: 0.5871\n",
      "[1/30] [89/594], loss: 0.1233, accyrasy: 0.9533, f1: 0.9211  f1_dev: 0.5924\n",
      "[1/30] [94/594], loss: 0.1266, accyrasy: 0.9576, f1: 0.9233  f1_dev: 0.5376\n",
      "[1/30] [99/594], loss: 0.138, accyrasy: 0.9489, f1: 0.9064  f1_dev: 0.575\n",
      "[1/30] [104/594], loss: 0.1347, accyrasy: 0.938, f1: 0.877  f1_dev: 0.5841\n",
      "[1/30] [109/594], loss: 0.1275, accyrasy: 0.9493, f1: 0.9034  f1_dev: 0.6153\n",
      "[1/30] [114/594], loss: 0.1335, accyrasy: 0.9494, f1: 0.9031  f1_dev: 0.6045\n",
      "[1/30] [119/594], loss: 0.1189, accyrasy: 0.9589, f1: 0.9231  f1_dev: 0.5979\n",
      "[1/30] [124/594], loss: 0.1151, accyrasy: 0.957, f1: 0.927  f1_dev: 0.5919\n",
      "[1/30] [129/594], loss: 0.1127, accyrasy: 0.9594, f1: 0.9168  f1_dev: 0.6118\n",
      "[1/30] [134/594], loss: 0.1243, accyrasy: 0.9481, f1: 0.9021  f1_dev: 0.5903\n",
      "[1/30] [139/594], loss: 0.1156, accyrasy: 0.9617, f1: 0.9301  f1_dev: 0.587\n",
      "[1/30] [144/594], loss: 0.1164, accyrasy: 0.9561, f1: 0.9176  f1_dev: 0.608\n",
      "[1/30] [149/594], loss: 0.1172, accyrasy: 0.9567, f1: 0.9225  f1_dev: 0.6308\n",
      "[1/30] [154/594], loss: 0.1245, accyrasy: 0.9531, f1: 0.9132  f1_dev: 0.5839\n",
      "[1/30] [159/594], loss: 0.1192, accyrasy: 0.961, f1: 0.9274  f1_dev: 0.5929\n",
      "[1/30] [164/594], loss: 0.1218, accyrasy: 0.9485, f1: 0.9064  f1_dev: 0.6127\n",
      "[1/30] [169/594], loss: 0.1224, accyrasy: 0.9509, f1: 0.9187  f1_dev: 0.6142\n",
      "[1/30] [174/594], loss: 0.1215, accyrasy: 0.9557, f1: 0.9142  f1_dev: 0.6006\n",
      "[1/30] [179/594], loss: 0.1173, accyrasy: 0.9564, f1: 0.9193  f1_dev: 0.5684\n",
      "[1/30] [184/594], loss: 0.1161, accyrasy: 0.9566, f1: 0.9199  f1_dev: 0.5983\n",
      "[1/30] [189/594], loss: 0.1137, accyrasy: 0.9577, f1: 0.9208  f1_dev: 0.5915\n",
      "[1/30] [194/594], loss: 0.1249, accyrasy: 0.9506, f1: 0.9101  f1_dev: 0.5974\n",
      "[1/30] [199/594], loss: 0.1138, accyrasy: 0.9575, f1: 0.9222  f1_dev: 0.6038\n",
      "[1/30] [204/594], loss: 0.1274, accyrasy: 0.9544, f1: 0.917  f1_dev: 0.5756\n",
      "[1/30] [209/594], loss: 0.124, accyrasy: 0.9535, f1: 0.8917  f1_dev: 0.583\n",
      "[1/30] [214/594], loss: 0.1183, accyrasy: 0.9561, f1: 0.9086  f1_dev: 0.5904\n",
      "[1/30] [219/594], loss: 0.1097, accyrasy: 0.9587, f1: 0.9165  f1_dev: 0.5911\n",
      "[1/30] [224/594], loss: 0.1179, accyrasy: 0.9529, f1: 0.9086  f1_dev: 0.5928\n",
      "[1/30] [229/594], loss: 0.1183, accyrasy: 0.9561, f1: 0.9059  f1_dev: 0.6239\n",
      "[1/30] [234/594], loss: 0.1144, accyrasy: 0.9584, f1: 0.9214  f1_dev: 0.6121\n",
      "[1/30] [239/594], loss: 0.1197, accyrasy: 0.955, f1: 0.9174  f1_dev: 0.5805\n",
      "[1/30] [244/594], loss: 0.1214, accyrasy: 0.9585, f1: 0.9246  f1_dev: 0.5879\n",
      "[1/30] [249/594], loss: 0.1231, accyrasy: 0.9536, f1: 0.9206  f1_dev: 0.6184\n",
      "[1/30] [254/594], loss: 0.1218, accyrasy: 0.9547, f1: 0.9238  f1_dev: 0.6186\n",
      "[1/30] [259/594], loss: 0.1122, accyrasy: 0.9584, f1: 0.9234  f1_dev: 0.5883\n",
      "[1/30] [264/594], loss: 0.1117, accyrasy: 0.9605, f1: 0.9292  f1_dev: 0.6067\n",
      "[1/30] [269/594], loss: 0.1216, accyrasy: 0.9635, f1: 0.9279  f1_dev: 0.6062\n",
      "[1/30] [274/594], loss: 0.1132, accyrasy: 0.9605, f1: 0.9226  f1_dev: 0.6062\n",
      "[1/30] [279/594], loss: 0.1139, accyrasy: 0.9604, f1: 0.9262  f1_dev: 0.6201\n",
      "[1/30] [284/594], loss: 0.1108, accyrasy: 0.9576, f1: 0.9236  f1_dev: 0.5979\n",
      "[1/30] [289/594], loss: 0.1125, accyrasy: 0.9574, f1: 0.9147  f1_dev: 0.6112\n",
      "[1/30] [294/594], loss: 0.1183, accyrasy: 0.9539, f1: 0.9162  f1_dev: 0.5932\n",
      "[1/30] [299/594], loss: 0.1134, accyrasy: 0.9584, f1: 0.9189  f1_dev: 0.6164\n",
      "[1/30] [304/594], loss: 0.1196, accyrasy: 0.9569, f1: 0.9224  f1_dev: 0.5973\n",
      "[1/30] [309/594], loss: 0.1204, accyrasy: 0.9525, f1: 0.9041  f1_dev: 0.5856\n",
      "[1/30] [314/594], loss: 0.117, accyrasy: 0.9627, f1: 0.93  f1_dev: 0.558\n",
      "[1/30] [319/594], loss: 0.1203, accyrasy: 0.9596, f1: 0.9275  f1_dev: 0.5926\n",
      "[1/30] [324/594], loss: 0.1245, accyrasy: 0.9534, f1: 0.9185  f1_dev: 0.6097\n",
      "[1/30] [329/594], loss: 0.1213, accyrasy: 0.9546, f1: 0.9127  f1_dev: 0.6095\n",
      "[1/30] [334/594], loss: 0.1246, accyrasy: 0.9563, f1: 0.9215  f1_dev: 0.5753\n",
      "[1/30] [339/594], loss: 0.114, accyrasy: 0.9611, f1: 0.9307  f1_dev: 0.6005\n",
      "[1/30] [344/594], loss: 0.1123, accyrasy: 0.9559, f1: 0.9218  f1_dev: 0.629\n",
      "[1/30] [349/594], loss: 0.1156, accyrasy: 0.9591, f1: 0.9276  f1_dev: 0.6207\n",
      "[1/30] [354/594], loss: 0.1119, accyrasy: 0.9591, f1: 0.9267  f1_dev: 0.6037\n",
      "[1/30] [359/594], loss: 0.1181, accyrasy: 0.9559, f1: 0.9098  f1_dev: 0.6149\n",
      "[1/30] [364/594], loss: 0.1197, accyrasy: 0.9494, f1: 0.9136  f1_dev: 0.6109\n",
      "[1/30] [369/594], loss: 0.1211, accyrasy: 0.959, f1: 0.9216  f1_dev: 0.5875\n",
      "[1/30] [374/594], loss: 0.1269, accyrasy: 0.957, f1: 0.9135  f1_dev: 0.6043\n",
      "[1/30] [379/594], loss: 0.1133, accyrasy: 0.9606, f1: 0.9218  f1_dev: 0.6049\n",
      "[1/30] [384/594], loss: 0.1158, accyrasy: 0.9595, f1: 0.923  f1_dev: 0.5535\n",
      "[1/30] [389/594], loss: 0.119, accyrasy: 0.954, f1: 0.8942  f1_dev: 0.5141\n",
      "[1/30] [394/594], loss: 0.1222, accyrasy: 0.9549, f1: 0.9132  f1_dev: 0.5847\n",
      "[1/30] [399/594], loss: 0.1191, accyrasy: 0.9591, f1: 0.9066  f1_dev: 0.5956\n",
      "[1/30] [404/594], loss: 0.1241, accyrasy: 0.9439, f1: 0.8549  f1_dev: 0.5906\n",
      "[1/30] [409/594], loss: 0.1269, accyrasy: 0.9475, f1: 0.8977  f1_dev: 0.6065\n",
      "[1/30] [414/594], loss: 0.117, accyrasy: 0.9601, f1: 0.9088  f1_dev: 0.6037\n",
      "[1/30] [419/594], loss: 0.1108, accyrasy: 0.9563, f1: 0.9159  f1_dev: 0.6176\n",
      "[1/30] [424/594], loss: 0.1131, accyrasy: 0.9586, f1: 0.9225  f1_dev: 0.5736\n",
      "[1/30] [429/594], loss: 0.1238, accyrasy: 0.9521, f1: 0.9053  f1_dev: 0.5273\n",
      "[1/30] [434/594], loss: 0.1281, accyrasy: 0.9557, f1: 0.9172  f1_dev: 0.59\n",
      "[1/30] [439/594], loss: 0.1153, accyrasy: 0.9597, f1: 0.9201  f1_dev: 0.5907\n",
      "[1/30] [444/594], loss: 0.1162, accyrasy: 0.9636, f1: 0.9292  f1_dev: 0.5483\n",
      "[1/30] [449/594], loss: 0.1215, accyrasy: 0.9589, f1: 0.9232  f1_dev: 0.5801\n",
      "[1/30] [454/594], loss: 0.1147, accyrasy: 0.9617, f1: 0.9157  f1_dev: 0.5241\n",
      "[1/30] [459/594], loss: 0.1304, accyrasy: 0.9559, f1: 0.9231  f1_dev: 0.5596\n",
      "[1/30] [464/594], loss: 0.1209, accyrasy: 0.9556, f1: 0.915  f1_dev: 0.612\n",
      "[1/30] [469/594], loss: 0.1229, accyrasy: 0.9571, f1: 0.9152  f1_dev: 0.6025\n",
      "[1/30] [474/594], loss: 0.1194, accyrasy: 0.9561, f1: 0.925  f1_dev: 0.59\n",
      "[1/30] [479/594], loss: 0.1191, accyrasy: 0.9544, f1: 0.9133  f1_dev: 0.5827\n",
      "[1/30] [484/594], loss: 0.115, accyrasy: 0.9557, f1: 0.922  f1_dev: 0.6099\n",
      "[1/30] [489/594], loss: 0.1138, accyrasy: 0.955, f1: 0.9145  f1_dev: 0.6164\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1/30] [494/594], loss: 0.1197, accyrasy: 0.955, f1: 0.9156  f1_dev: 0.6034\n",
      "[1/30] [499/594], loss: 0.1129, accyrasy: 0.9603, f1: 0.9291  f1_dev: 0.6072\n",
      "[1/30] [504/594], loss: 0.1227, accyrasy: 0.9566, f1: 0.9275  f1_dev: 0.6215\n",
      "[1/30] [509/594], loss: 0.1187, accyrasy: 0.955, f1: 0.9119  f1_dev: 0.6065\n",
      "[1/30] [514/594], loss: 0.1211, accyrasy: 0.9516, f1: 0.9111  f1_dev: 0.6036\n",
      "[1/30] [519/594], loss: 0.1265, accyrasy: 0.9534, f1: 0.9229  f1_dev: 0.6112\n",
      "[1/30] [524/594], loss: 0.1202, accyrasy: 0.961, f1: 0.9224  f1_dev: 0.6074\n",
      "[1/30] [529/594], loss: 0.1143, accyrasy: 0.9604, f1: 0.9258  f1_dev: 0.6069\n",
      "[1/30] [534/594], loss: 0.113, accyrasy: 0.9589, f1: 0.9256  f1_dev: 0.6065\n",
      "[1/30] [539/594], loss: 0.1118, accyrasy: 0.9615, f1: 0.9305  f1_dev: 0.6292\n",
      "[1/30] [544/594], loss: 0.1139, accyrasy: 0.9571, f1: 0.9154  f1_dev: 0.6009\n",
      "[1/30] [549/594], loss: 0.1071, accyrasy: 0.9653, f1: 0.9277  f1_dev: 0.5847\n",
      "[1/30] [554/594], loss: 0.1095, accyrasy: 0.9581, f1: 0.9235  f1_dev: 0.6042\n",
      "[1/30] [559/594], loss: 0.1155, accyrasy: 0.955, f1: 0.9154  f1_dev: 0.6025\n",
      "[1/30] [564/594], loss: 0.1204, accyrasy: 0.9564, f1: 0.9204  f1_dev: 0.6205\n",
      "[1/30] [569/594], loss: 0.1181, accyrasy: 0.9577, f1: 0.9201  f1_dev: 0.6072\n",
      "[1/30] [574/594], loss: 0.1292, accyrasy: 0.9571, f1: 0.9201  f1_dev: 0.5035\n",
      "[1/30] [579/594], loss: 0.1502, accyrasy: 0.9541, f1: 0.9155  f1_dev: 0.5897\n",
      "[1/30] [584/594], loss: 0.1354, accyrasy: 0.9545, f1: 0.918  f1_dev: 0.6007\n",
      "[1/30] [589/594], loss: 0.128, accyrasy: 0.958, f1: 0.9224  f1_dev: 0.6069\n",
      "[2/30] [4/594], loss: 0.1224, accyrasy: 0.9536, f1: 0.9085  f1_dev: 0.6202\n",
      "[2/30] [9/594], loss: 0.1206, accyrasy: 0.9565, f1: 0.9186  f1_dev: 0.6116\n",
      "[2/30] [14/594], loss: 0.1185, accyrasy: 0.9576, f1: 0.925  f1_dev: 0.5969\n",
      "[2/30] [19/594], loss: 0.1232, accyrasy: 0.9591, f1: 0.9224  f1_dev: 0.6059\n",
      "[2/30] [24/594], loss: 0.1205, accyrasy: 0.9589, f1: 0.9196  f1_dev: 0.5724\n",
      "[2/30] [29/594], loss: 0.1306, accyrasy: 0.9581, f1: 0.9173  f1_dev: 0.5827\n",
      "[2/30] [34/594], loss: 0.1289, accyrasy: 0.958, f1: 0.9182  f1_dev: 0.5911\n",
      "[2/30] [39/594], loss: 0.1148, accyrasy: 0.9606, f1: 0.9088  f1_dev: 0.6018\n",
      "[2/30] [44/594], loss: 0.116, accyrasy: 0.9623, f1: 0.9287  f1_dev: 0.624\n",
      "[2/30] [49/594], loss: 0.1188, accyrasy: 0.9499, f1: 0.8994  f1_dev: 0.6021\n",
      "[2/30] [54/594], loss: 0.1073, accyrasy: 0.9617, f1: 0.9216  f1_dev: 0.6085\n",
      "[2/30] [59/594], loss: 0.1059, accyrasy: 0.9614, f1: 0.9191  f1_dev: 0.6068\n",
      "[2/30] [64/594], loss: 0.1061, accyrasy: 0.963, f1: 0.9336  f1_dev: 0.6317\n",
      "[2/30] [69/594], loss: 0.1205, accyrasy: 0.9466, f1: 0.8888  f1_dev: 0.5937\n",
      "[2/30] [74/594], loss: 0.1216, accyrasy: 0.9526, f1: 0.9101  f1_dev: 0.6061\n",
      "[2/30] [79/594], loss: 0.1237, accyrasy: 0.9471, f1: 0.8761  f1_dev: 0.5989\n",
      "[2/30] [84/594], loss: 0.1147, accyrasy: 0.9624, f1: 0.9279  f1_dev: 0.5791\n",
      "[2/30] [89/594], loss: 0.1224, accyrasy: 0.9609, f1: 0.9333  f1_dev: 0.5518\n",
      "[2/30] [94/594], loss: 0.1256, accyrasy: 0.9493, f1: 0.8968  f1_dev: 0.609\n",
      "[2/30] [99/594], loss: 0.1266, accyrasy: 0.9464, f1: 0.8795  f1_dev: 0.5932\n",
      "[2/30] [104/594], loss: 0.124, accyrasy: 0.949, f1: 0.9083  f1_dev: 0.6087\n",
      "[2/30] [109/594], loss: 0.1088, accyrasy: 0.9641, f1: 0.924  f1_dev: 0.627\n",
      "[2/30] [114/594], loss: 0.1218, accyrasy: 0.957, f1: 0.9298  f1_dev: 0.6108\n",
      "[2/30] [119/594], loss: 0.118, accyrasy: 0.9611, f1: 0.9224  f1_dev: 0.6198\n",
      "[2/30] [124/594], loss: 0.1128, accyrasy: 0.9545, f1: 0.913  f1_dev: 0.6066\n",
      "[2/30] [129/594], loss: 0.1187, accyrasy: 0.9509, f1: 0.8896  f1_dev: 0.5992\n",
      "[2/30] [134/594], loss: 0.1181, accyrasy: 0.9601, f1: 0.9241  f1_dev: 0.5962\n",
      "[2/30] [139/594], loss: 0.1075, accyrasy: 0.9575, f1: 0.9265  f1_dev: 0.6286\n",
      "[2/30] [144/594], loss: 0.1068, accyrasy: 0.9601, f1: 0.9237  f1_dev: 0.6104\n",
      "[2/30] [149/594], loss: 0.1068, accyrasy: 0.9617, f1: 0.9297  f1_dev: 0.6061\n",
      "[2/30] [154/594], loss: 0.1061, accyrasy: 0.9594, f1: 0.9209  f1_dev: 0.5998\n",
      "[2/30] [159/594], loss: 0.1153, accyrasy: 0.9543, f1: 0.9161  f1_dev: 0.6096\n",
      "[2/30] [164/594], loss: 0.1106, accyrasy: 0.9599, f1: 0.9254  f1_dev: 0.6223\n",
      "[2/30] [169/594], loss: 0.1253, accyrasy: 0.9511, f1: 0.9133  f1_dev: 0.6207\n",
      "[2/30] [174/594], loss: 0.1205, accyrasy: 0.9573, f1: 0.9214  f1_dev: 0.5913\n",
      "[2/30] [179/594], loss: 0.1176, accyrasy: 0.952, f1: 0.9172  f1_dev: 0.5955\n",
      "[2/30] [184/594], loss: 0.1176, accyrasy: 0.9607, f1: 0.9298  f1_dev: 0.6017\n",
      "[2/30] [189/594], loss: 0.1149, accyrasy: 0.9613, f1: 0.9309  f1_dev: 0.6238\n",
      "[2/30] [194/594], loss: 0.1097, accyrasy: 0.9593, f1: 0.9262  f1_dev: 0.6072\n",
      "[2/30] [199/594], loss: 0.1086, accyrasy: 0.961, f1: 0.9236  f1_dev: 0.5993\n",
      "[2/30] [204/594], loss: 0.1093, accyrasy: 0.9545, f1: 0.9119  f1_dev: 0.6045\n",
      "[2/30] [209/594], loss: 0.1131, accyrasy: 0.9584, f1: 0.9229  f1_dev: 0.6313\n",
      "[2/30] [214/594], loss: 0.119, accyrasy: 0.9529, f1: 0.9204  f1_dev: 0.5996\n",
      "[2/30] [219/594], loss: 0.1167, accyrasy: 0.9577, f1: 0.9193  f1_dev: 0.5778\n",
      "[2/30] [224/594], loss: 0.1203, accyrasy: 0.9566, f1: 0.9207  f1_dev: 0.594\n",
      "[2/30] [229/594], loss: 0.1156, accyrasy: 0.9593, f1: 0.9258  f1_dev: 0.6266\n",
      "[2/30] [234/594], loss: 0.1129, accyrasy: 0.9575, f1: 0.9205  f1_dev: 0.623\n",
      "[2/30] [239/594], loss: 0.1123, accyrasy: 0.9584, f1: 0.9243  f1_dev: 0.5894\n",
      "[2/30] [244/594], loss: 0.1135, accyrasy: 0.9557, f1: 0.9073  f1_dev: 0.5806\n",
      "[2/30] [249/594], loss: 0.1138, accyrasy: 0.9539, f1: 0.9073  f1_dev: 0.5601\n",
      "[2/30] [254/594], loss: 0.1059, accyrasy: 0.9616, f1: 0.931  f1_dev: 0.6195\n",
      "[2/30] [259/594], loss: 0.1094, accyrasy: 0.9571, f1: 0.9114  f1_dev: 0.6237\n",
      "[2/30] [264/594], loss: 0.1107, accyrasy: 0.9601, f1: 0.9265  f1_dev: 0.6096\n",
      "[2/30] [269/594], loss: 0.1218, accyrasy: 0.9497, f1: 0.8837  f1_dev: 0.6334\n",
      "[2/30] [274/594], loss: 0.1116, accyrasy: 0.958, f1: 0.9231  f1_dev: 0.5839\n",
      "[2/30] [279/594], loss: 0.1166, accyrasy: 0.9614, f1: 0.9335  f1_dev: 0.6035\n",
      "[2/30] [284/594], loss: 0.1087, accyrasy: 0.9606, f1: 0.9251  f1_dev: 0.6176\n",
      "[2/30] [289/594], loss: 0.1124, accyrasy: 0.9585, f1: 0.9203  f1_dev: 0.6207\n",
      "[2/30] [294/594], loss: 0.1087, accyrasy: 0.958, f1: 0.9225  f1_dev: 0.6285\n",
      "[2/30] [299/594], loss: 0.1142, accyrasy: 0.9623, f1: 0.9306  f1_dev: 0.6347\n",
      "[2/30] [304/594], loss: 0.1096, accyrasy: 0.962, f1: 0.9302  f1_dev: 0.599\n",
      "[2/30] [309/594], loss: 0.109, accyrasy: 0.9616, f1: 0.932  f1_dev: 0.6097\n",
      "[2/30] [314/594], loss: 0.1102, accyrasy: 0.9561, f1: 0.9135  f1_dev: 0.6196\n",
      "[2/30] [319/594], loss: 0.1144, accyrasy: 0.9591, f1: 0.9115  f1_dev: 0.6124\n",
      "[2/30] [324/594], loss: 0.1176, accyrasy: 0.9574, f1: 0.9192  f1_dev: 0.5468\n",
      "[2/30] [329/594], loss: 0.1264, accyrasy: 0.961, f1: 0.9278  f1_dev: 0.5755\n",
      "[2/30] [334/594], loss: 0.1157, accyrasy: 0.9634, f1: 0.9347  f1_dev: 0.588\n",
      "[2/30] [339/594], loss: 0.1082, accyrasy: 0.9614, f1: 0.9329  f1_dev: 0.6235\n",
      "[2/30] [344/594], loss: 0.1106, accyrasy: 0.96, f1: 0.9246  f1_dev: 0.6254\n",
      "[2/30] [349/594], loss: 0.1088, accyrasy: 0.9623, f1: 0.9236  f1_dev: 0.6183\n",
      "[2/30] [354/594], loss: 0.108, accyrasy: 0.9599, f1: 0.9198  f1_dev: 0.6331\n",
      "[2/30] [359/594], loss: 0.1045, accyrasy: 0.9579, f1: 0.9291  f1_dev: 0.6425\n",
      "[2/30] [364/594], loss: 0.1128, accyrasy: 0.9604, f1: 0.9329  f1_dev: 0.6284\n",
      "[2/30] [369/594], loss: 0.1088, accyrasy: 0.9597, f1: 0.9258  f1_dev: 0.6202\n",
      "[2/30] [374/594], loss: 0.1243, accyrasy: 0.9553, f1: 0.9177  f1_dev: 0.6166\n",
      "[2/30] [379/594], loss: 0.1137, accyrasy: 0.9553, f1: 0.9084  f1_dev: 0.6294\n",
      "[2/30] [384/594], loss: 0.1119, accyrasy: 0.9557, f1: 0.9156  f1_dev: 0.5883\n",
      "[2/30] [389/594], loss: 0.1124, accyrasy: 0.962, f1: 0.9347  f1_dev: 0.6102\n",
      "[2/30] [394/594], loss: 0.1173, accyrasy: 0.9551, f1: 0.9191  f1_dev: 0.6293\n",
      "[2/30] [399/594], loss: 0.1154, accyrasy: 0.9594, f1: 0.9233  f1_dev: 0.6079\n",
      "[2/30] [404/594], loss: 0.1148, accyrasy: 0.957, f1: 0.9144  f1_dev: 0.601\n",
      "[2/30] [409/594], loss: 0.1195, accyrasy: 0.9603, f1: 0.9311  f1_dev: 0.6429\n",
      "[2/30] [414/594], loss: 0.1254, accyrasy: 0.9555, f1: 0.9143  f1_dev: 0.6149\n",
      "[2/30] [419/594], loss: 0.1372, accyrasy: 0.9511, f1: 0.9128  f1_dev: 0.5989\n",
      "[2/30] [424/594], loss: 0.128, accyrasy: 0.9581, f1: 0.9256  f1_dev: 0.5859\n",
      "[2/30] [429/594], loss: 0.1151, accyrasy: 0.958, f1: 0.9186  f1_dev: 0.6337\n",
      "[2/30] [434/594], loss: 0.1152, accyrasy: 0.956, f1: 0.9185  f1_dev: 0.6306\n",
      "[2/30] [439/594], loss: 0.1204, accyrasy: 0.956, f1: 0.9197  f1_dev: 0.6358\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2/30] [444/594], loss: 0.1124, accyrasy: 0.9591, f1: 0.9201  f1_dev: 0.6152\n",
      "[2/30] [449/594], loss: 0.1106, accyrasy: 0.9597, f1: 0.9277  f1_dev: 0.5862\n",
      "[2/30] [454/594], loss: 0.1166, accyrasy: 0.9613, f1: 0.922  f1_dev: 0.6295\n",
      "[2/30] [459/594], loss: 0.1176, accyrasy: 0.9579, f1: 0.9158  f1_dev: 0.6374\n",
      "[2/30] [464/594], loss: 0.1136, accyrasy: 0.9614, f1: 0.9281  f1_dev: 0.6564\n",
      "[2/30] [469/594], loss: 0.1105, accyrasy: 0.9577, f1: 0.9193  f1_dev: 0.6224\n",
      "[2/30] [474/594], loss: 0.1062, accyrasy: 0.9617, f1: 0.9278  f1_dev: 0.6469\n",
      "[2/30] [479/594], loss: 0.1088, accyrasy: 0.9617, f1: 0.9281  f1_dev: 0.6352\n",
      "[2/30] [484/594], loss: 0.1088, accyrasy: 0.9625, f1: 0.9345  f1_dev: 0.6375\n",
      "[2/30] [489/594], loss: 0.1068, accyrasy: 0.9617, f1: 0.9293  f1_dev: 0.6038\n",
      "[2/30] [494/594], loss: 0.117, accyrasy: 0.9589, f1: 0.9264  f1_dev: 0.5751\n",
      "[2/30] [499/594], loss: 0.1282, accyrasy: 0.9576, f1: 0.9197  f1_dev: 0.6201\n",
      "[2/30] [504/594], loss: 0.1244, accyrasy: 0.9421, f1: 0.8822  f1_dev: 0.6196\n",
      "[2/30] [509/594], loss: 0.1288, accyrasy: 0.9509, f1: 0.8928  f1_dev: 0.5748\n",
      "[2/30] [514/594], loss: 0.1195, accyrasy: 0.9604, f1: 0.925  f1_dev: 0.593\n",
      "[2/30] [519/594], loss: 0.1154, accyrasy: 0.957, f1: 0.9197  f1_dev: 0.6478\n",
      "[2/30] [524/594], loss: 0.1167, accyrasy: 0.957, f1: 0.9161  f1_dev: 0.6461\n",
      "[2/30] [529/594], loss: 0.1193, accyrasy: 0.9589, f1: 0.916  f1_dev: 0.6537\n",
      "[2/30] [534/594], loss: 0.1176, accyrasy: 0.9541, f1: 0.9175  f1_dev: 0.6523\n",
      "[2/30] [539/594], loss: 0.114, accyrasy: 0.9547, f1: 0.9192  f1_dev: 0.6334\n",
      "[2/30] [544/594], loss: 0.1142, accyrasy: 0.9619, f1: 0.9299  f1_dev: 0.6024\n",
      "[2/30] [549/594], loss: 0.1169, accyrasy: 0.9587, f1: 0.9314  f1_dev: 0.6576\n",
      "[2/30] [554/594], loss: 0.1109, accyrasy: 0.9633, f1: 0.9325  f1_dev: 0.6409\n",
      "[2/30] [559/594], loss: 0.106, accyrasy: 0.9579, f1: 0.9227  f1_dev: 0.6121\n",
      "[2/30] [564/594], loss: 0.1119, accyrasy: 0.9623, f1: 0.9326  f1_dev: 0.6208\n",
      "[2/30] [569/594], loss: 0.1137, accyrasy: 0.9581, f1: 0.9239  f1_dev: 0.6097\n",
      "[2/30] [574/594], loss: 0.1105, accyrasy: 0.9624, f1: 0.9252  f1_dev: 0.6327\n",
      "[2/30] [579/594], loss: 0.1105, accyrasy: 0.9625, f1: 0.9321  f1_dev: 0.6388\n",
      "[2/30] [584/594], loss: 0.1059, accyrasy: 0.9615, f1: 0.9245  f1_dev: 0.6411\n",
      "[2/30] [589/594], loss: 0.1095, accyrasy: 0.9599, f1: 0.929  f1_dev: 0.6318\n",
      "[3/30] [4/594], loss: 0.1155, accyrasy: 0.9571, f1: 0.9163  f1_dev: 0.6314\n",
      "[3/30] [9/594], loss: 0.1138, accyrasy: 0.9579, f1: 0.9179  f1_dev: 0.6501\n",
      "[3/30] [14/594], loss: 0.1136, accyrasy: 0.9607, f1: 0.9265  f1_dev: 0.6305\n",
      "[3/30] [19/594], loss: 0.1092, accyrasy: 0.9585, f1: 0.918  f1_dev: 0.6218\n",
      "[3/30] [24/594], loss: 0.1089, accyrasy: 0.96, f1: 0.9232  f1_dev: 0.6271\n",
      "[3/30] [29/594], loss: 0.105, accyrasy: 0.9577, f1: 0.9181  f1_dev: 0.6602\n",
      "[3/30] [34/594], loss: 0.1095, accyrasy: 0.9587, f1: 0.9276  f1_dev: 0.5956\n",
      "[3/30] [39/594], loss: 0.1135, accyrasy: 0.9599, f1: 0.9191  f1_dev: 0.6562\n",
      "[3/30] [44/594], loss: 0.1054, accyrasy: 0.9604, f1: 0.9277  f1_dev: 0.6533\n",
      "[3/30] [49/594], loss: 0.1098, accyrasy: 0.9606, f1: 0.9265  f1_dev: 0.6354\n",
      "[3/30] [54/594], loss: 0.1069, accyrasy: 0.9569, f1: 0.9163  f1_dev: 0.6335\n",
      "[3/30] [59/594], loss: 0.1169, accyrasy: 0.9611, f1: 0.925  f1_dev: 0.6261\n",
      "[3/30] [64/594], loss: 0.11, accyrasy: 0.9616, f1: 0.9291  f1_dev: 0.635\n",
      "[3/30] [69/594], loss: 0.1083, accyrasy: 0.96, f1: 0.9246  f1_dev: 0.6203\n",
      "[3/30] [74/594], loss: 0.1084, accyrasy: 0.9595, f1: 0.9252  f1_dev: 0.6196\n",
      "[3/30] [79/594], loss: 0.1092, accyrasy: 0.9607, f1: 0.9251  f1_dev: 0.6519\n",
      "[3/30] [84/594], loss: 0.1069, accyrasy: 0.9621, f1: 0.9308  f1_dev: 0.6206\n",
      "[3/30] [89/594], loss: 0.1047, accyrasy: 0.9566, f1: 0.9191  f1_dev: 0.6386\n",
      "[3/30] [94/594], loss: 0.1027, accyrasy: 0.9606, f1: 0.9242  f1_dev: 0.6196\n",
      "[3/30] [99/594], loss: 0.1121, accyrasy: 0.9519, f1: 0.9007  f1_dev: 0.6691\n",
      "[3/30] [104/594], loss: 0.1058, accyrasy: 0.9589, f1: 0.9231  f1_dev: 0.6578\n",
      "[3/30] [109/594], loss: 0.1112, accyrasy: 0.9597, f1: 0.9254  f1_dev: 0.6148\n",
      "[3/30] [114/594], loss: 0.1093, accyrasy: 0.96, f1: 0.9262  f1_dev: 0.6232\n",
      "[3/30] [119/594], loss: 0.1108, accyrasy: 0.9531, f1: 0.915  f1_dev: 0.6311\n",
      "[3/30] [124/594], loss: 0.1084, accyrasy: 0.9564, f1: 0.9232  f1_dev: 0.6482\n",
      "[3/30] [129/594], loss: 0.1108, accyrasy: 0.9606, f1: 0.9312  f1_dev: 0.6597\n",
      "[3/30] [134/594], loss: 0.1122, accyrasy: 0.9574, f1: 0.9229  f1_dev: 0.6312\n",
      "[3/30] [139/594], loss: 0.115, accyrasy: 0.9583, f1: 0.9228  f1_dev: 0.6424\n",
      "[3/30] [144/594], loss: 0.1208, accyrasy: 0.9526, f1: 0.9168  f1_dev: 0.6449\n",
      "[3/30] [149/594], loss: 0.1112, accyrasy: 0.9601, f1: 0.9242  f1_dev: 0.6334\n",
      "[3/30] [154/594], loss: 0.1106, accyrasy: 0.9596, f1: 0.9261  f1_dev: 0.6305\n",
      "[3/30] [159/594], loss: 0.1201, accyrasy: 0.9535, f1: 0.9143  f1_dev: 0.6456\n",
      "[3/30] [164/594], loss: 0.1106, accyrasy: 0.9616, f1: 0.928  f1_dev: 0.6307\n",
      "[3/30] [169/594], loss: 0.108, accyrasy: 0.9577, f1: 0.9163  f1_dev: 0.639\n",
      "[3/30] [174/594], loss: 0.1065, accyrasy: 0.962, f1: 0.9258  f1_dev: 0.641\n",
      "[3/30] [179/594], loss: 0.1097, accyrasy: 0.9629, f1: 0.9287  f1_dev: 0.6519\n",
      "[3/30] [184/594], loss: 0.1227, accyrasy: 0.952, f1: 0.8712  f1_dev: 0.5955\n",
      "[3/30] [189/594], loss: 0.1222, accyrasy: 0.9566, f1: 0.9166  f1_dev: 0.6034\n",
      "[3/30] [194/594], loss: 0.111, accyrasy: 0.961, f1: 0.9293  f1_dev: 0.634\n",
      "[3/30] [199/594], loss: 0.1084, accyrasy: 0.9625, f1: 0.9331  f1_dev: 0.6462\n",
      "[3/30] [204/594], loss: 0.1117, accyrasy: 0.9627, f1: 0.93  f1_dev: 0.624\n",
      "[3/30] [209/594], loss: 0.1133, accyrasy: 0.9467, f1: 0.9055  f1_dev: 0.6043\n",
      "[3/30] [214/594], loss: 0.1137, accyrasy: 0.9561, f1: 0.9148  f1_dev: 0.6158\n",
      "[3/30] [219/594], loss: 0.1067, accyrasy: 0.9626, f1: 0.9333  f1_dev: 0.6288\n",
      "[3/30] [224/594], loss: 0.1122, accyrasy: 0.951, f1: 0.9034  f1_dev: 0.6492\n",
      "[3/30] [229/594], loss: 0.1186, accyrasy: 0.9517, f1: 0.8924  f1_dev: 0.6117\n",
      "[3/30] [234/594], loss: 0.113, accyrasy: 0.962, f1: 0.9241  f1_dev: 0.6206\n",
      "[3/30] [239/594], loss: 0.1092, accyrasy: 0.9635, f1: 0.9282  f1_dev: 0.6297\n",
      "[3/30] [244/594], loss: 0.1079, accyrasy: 0.9617, f1: 0.9313  f1_dev: 0.6541\n",
      "[3/30] [249/594], loss: 0.1195, accyrasy: 0.9563, f1: 0.902  f1_dev: 0.6294\n",
      "[3/30] [254/594], loss: 0.1324, accyrasy: 0.9594, f1: 0.9277  f1_dev: 0.6146\n",
      "[3/30] [259/594], loss: 0.1135, accyrasy: 0.9586, f1: 0.9106  f1_dev: 0.6246\n",
      "[3/30] [264/594], loss: 0.1117, accyrasy: 0.9589, f1: 0.9176  f1_dev: 0.6787\n",
      "[3/30] [269/594], loss: 0.1087, accyrasy: 0.9563, f1: 0.9156  f1_dev: 0.6481\n",
      "[3/30] [274/594], loss: 0.1134, accyrasy: 0.9545, f1: 0.909  f1_dev: 0.6178\n",
      "[3/30] [279/594], loss: 0.1065, accyrasy: 0.9613, f1: 0.9258  f1_dev: 0.6535\n",
      "[3/30] [284/594], loss: 0.1078, accyrasy: 0.9614, f1: 0.9222  f1_dev: 0.6525\n",
      "[3/30] [289/594], loss: 0.1085, accyrasy: 0.9603, f1: 0.9261  f1_dev: 0.6066\n",
      "[3/30] [294/594], loss: 0.1162, accyrasy: 0.9546, f1: 0.9132  f1_dev: 0.6304\n",
      "[3/30] [299/594], loss: 0.1138, accyrasy: 0.9555, f1: 0.906  f1_dev: 0.6239\n",
      "[3/30] [304/594], loss: 0.1183, accyrasy: 0.942, f1: 0.8717  f1_dev: 0.6281\n",
      "[3/30] [309/594], loss: 0.11, accyrasy: 0.9563, f1: 0.9254  f1_dev: 0.6163\n",
      "[3/30] [314/594], loss: 0.113, accyrasy: 0.9635, f1: 0.9296  f1_dev: 0.6056\n",
      "[3/30] [319/594], loss: 0.1112, accyrasy: 0.9596, f1: 0.9258  f1_dev: 0.6009\n",
      "[3/30] [324/594], loss: 0.1108, accyrasy: 0.9597, f1: 0.9172  f1_dev: 0.6386\n",
      "[3/30] [329/594], loss: 0.1015, accyrasy: 0.9669, f1: 0.9334  f1_dev: 0.6381\n",
      "[3/30] [334/594], loss: 0.1126, accyrasy: 0.9555, f1: 0.9107  f1_dev: 0.6154\n",
      "[3/30] [339/594], loss: 0.1051, accyrasy: 0.957, f1: 0.919  f1_dev: 0.6562\n",
      "[3/30] [344/594], loss: 0.1143, accyrasy: 0.9611, f1: 0.9198  f1_dev: 0.6587\n",
      "[3/30] [349/594], loss: 0.1151, accyrasy: 0.9575, f1: 0.9198  f1_dev: 0.6465\n",
      "[3/30] [354/594], loss: 0.108, accyrasy: 0.9586, f1: 0.9209  f1_dev: 0.6545\n",
      "[3/30] [359/594], loss: 0.1079, accyrasy: 0.961, f1: 0.931  f1_dev: 0.6427\n",
      "[3/30] [364/594], loss: 0.1108, accyrasy: 0.959, f1: 0.9293  f1_dev: 0.6529\n",
      "[3/30] [369/594], loss: 0.1113, accyrasy: 0.9525, f1: 0.9126  f1_dev: 0.6369\n",
      "[3/30] [374/594], loss: 0.109, accyrasy: 0.9654, f1: 0.9355  f1_dev: 0.6216\n",
      "[3/30] [379/594], loss: 0.1152, accyrasy: 0.9619, f1: 0.9277  f1_dev: 0.5709\n",
      "[3/30] [384/594], loss: 0.1133, accyrasy: 0.9603, f1: 0.9256  f1_dev: 0.6436\n",
      "[3/30] [389/594], loss: 0.1112, accyrasy: 0.9606, f1: 0.9271  f1_dev: 0.6674\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[3/30] [394/594], loss: 0.1112, accyrasy: 0.9607, f1: 0.9303  f1_dev: 0.6575\n",
      "[3/30] [399/594], loss: 0.1132, accyrasy: 0.957, f1: 0.9171  f1_dev: 0.5878\n",
      "[3/30] [404/594], loss: 0.1116, accyrasy: 0.9561, f1: 0.9177  f1_dev: 0.6662\n",
      "[3/30] [409/594], loss: 0.1104, accyrasy: 0.9626, f1: 0.9293  f1_dev: 0.6724\n",
      "[3/30] [414/594], loss: 0.1135, accyrasy: 0.9601, f1: 0.9242  f1_dev: 0.6454\n",
      "[3/30] [419/594], loss: 0.1162, accyrasy: 0.9514, f1: 0.904  f1_dev: 0.6129\n",
      "[3/30] [424/594], loss: 0.1106, accyrasy: 0.962, f1: 0.9283  f1_dev: 0.6227\n",
      "[3/30] [429/594], loss: 0.1262, accyrasy: 0.9584, f1: 0.9174  f1_dev: 0.6641\n",
      "[3/30] [434/594], loss: 0.1161, accyrasy: 0.9599, f1: 0.921  f1_dev: 0.6644\n",
      "[3/30] [439/594], loss: 0.1143, accyrasy: 0.9621, f1: 0.9298  f1_dev: 0.6213\n",
      "[3/30] [444/594], loss: 0.1107, accyrasy: 0.96, f1: 0.9302  f1_dev: 0.6456\n",
      "[3/30] [449/594], loss: 0.1114, accyrasy: 0.9601, f1: 0.9216  f1_dev: 0.6354\n",
      "[3/30] [454/594], loss: 0.1056, accyrasy: 0.9601, f1: 0.9301  f1_dev: 0.6172\n",
      "[3/30] [459/594], loss: 0.1116, accyrasy: 0.9609, f1: 0.9252  f1_dev: 0.6609\n",
      "[3/30] [464/594], loss: 0.1087, accyrasy: 0.9576, f1: 0.925  f1_dev: 0.6091\n",
      "[3/30] [469/594], loss: 0.1138, accyrasy: 0.9597, f1: 0.9278  f1_dev: 0.609\n",
      "[3/30] [474/594], loss: 0.1066, accyrasy: 0.9625, f1: 0.9307  f1_dev: 0.6488\n",
      "[3/30] [479/594], loss: 0.1064, accyrasy: 0.9579, f1: 0.922  f1_dev: 0.6175\n",
      "[3/30] [484/594], loss: 0.1121, accyrasy: 0.9596, f1: 0.9256  f1_dev: 0.6201\n",
      "[3/30] [489/594], loss: 0.1118, accyrasy: 0.9635, f1: 0.9337  f1_dev: 0.6734\n",
      "[3/30] [494/594], loss: 0.1165, accyrasy: 0.9594, f1: 0.9203  f1_dev: 0.6192\n",
      "[3/30] [499/594], loss: 0.1067, accyrasy: 0.9624, f1: 0.9306  f1_dev: 0.6405\n",
      "[3/30] [504/594], loss: 0.1101, accyrasy: 0.9565, f1: 0.9209  f1_dev: 0.6535\n",
      "[3/30] [509/594], loss: 0.108, accyrasy: 0.9544, f1: 0.9195  f1_dev: 0.6271\n",
      "[3/30] [514/594], loss: 0.1127, accyrasy: 0.9554, f1: 0.9138  f1_dev: 0.6061\n",
      "[3/30] [519/594], loss: 0.108, accyrasy: 0.9615, f1: 0.9255  f1_dev: 0.6311\n",
      "[3/30] [524/594], loss: 0.1081, accyrasy: 0.9589, f1: 0.9264  f1_dev: 0.6372\n",
      "[3/30] [529/594], loss: 0.1043, accyrasy: 0.9643, f1: 0.9345  f1_dev: 0.629\n",
      "[3/30] [534/594], loss: 0.1107, accyrasy: 0.9505, f1: 0.9058  f1_dev: 0.6731\n",
      "[3/30] [539/594], loss: 0.1151, accyrasy: 0.9501, f1: 0.9049  f1_dev: 0.6401\n",
      "[3/30] [544/594], loss: 0.1142, accyrasy: 0.9556, f1: 0.9167  f1_dev: 0.6027\n",
      "[3/30] [549/594], loss: 0.1077, accyrasy: 0.9584, f1: 0.9209  f1_dev: 0.6344\n",
      "[3/30] [554/594], loss: 0.1059, accyrasy: 0.9619, f1: 0.9313  f1_dev: 0.6336\n",
      "[3/30] [559/594], loss: 0.1158, accyrasy: 0.9564, f1: 0.9219  f1_dev: 0.597\n",
      "[3/30] [564/594], loss: 0.1143, accyrasy: 0.9571, f1: 0.9068  f1_dev: 0.6367\n",
      "[3/30] [569/594], loss: 0.105, accyrasy: 0.9646, f1: 0.9347  f1_dev: 0.6301\n",
      "[3/30] [574/594], loss: 0.1026, accyrasy: 0.9633, f1: 0.9322  f1_dev: 0.6448\n",
      "[3/30] [579/594], loss: 0.1044, accyrasy: 0.9581, f1: 0.9208  f1_dev: 0.6526\n",
      "[3/30] [584/594], loss: 0.1031, accyrasy: 0.9626, f1: 0.9345  f1_dev: 0.6481\n",
      "[3/30] [589/594], loss: 0.103, accyrasy: 0.9641, f1: 0.9297  f1_dev: 0.6305\n",
      "[4/30] [4/594], loss: 0.1135, accyrasy: 0.9586, f1: 0.9195  f1_dev: 0.6506\n",
      "[4/30] [9/594], loss: 0.1111, accyrasy: 0.9605, f1: 0.9307  f1_dev: 0.6117\n",
      "[4/30] [14/594], loss: 0.1112, accyrasy: 0.9593, f1: 0.9273  f1_dev: 0.6537\n",
      "[4/30] [19/594], loss: 0.106, accyrasy: 0.9585, f1: 0.9215  f1_dev: 0.6504\n",
      "[4/30] [24/594], loss: 0.1012, accyrasy: 0.9665, f1: 0.9365  f1_dev: 0.6292\n",
      "[4/30] [29/594], loss: 0.1078, accyrasy: 0.9589, f1: 0.9179  f1_dev: 0.6262\n",
      "[4/30] [34/594], loss: 0.1217, accyrasy: 0.9397, f1: 0.8861  f1_dev: 0.6312\n",
      "[4/30] [39/594], loss: 0.115, accyrasy: 0.9564, f1: 0.9168  f1_dev: 0.6362\n",
      "[4/30] [44/594], loss: 0.1085, accyrasy: 0.9599, f1: 0.9259  f1_dev: 0.6342\n",
      "[4/30] [49/594], loss: 0.1082, accyrasy: 0.9595, f1: 0.9306  f1_dev: 0.6467\n",
      "[4/30] [54/594], loss: 0.1105, accyrasy: 0.9581, f1: 0.9152  f1_dev: 0.649\n",
      "[4/30] [59/594], loss: 0.1118, accyrasy: 0.9593, f1: 0.9126  f1_dev: 0.6152\n",
      "[4/30] [64/594], loss: 0.1096, accyrasy: 0.9607, f1: 0.9318  f1_dev: 0.6411\n",
      "[4/30] [69/594], loss: 0.1115, accyrasy: 0.9566, f1: 0.9208  f1_dev: 0.6716\n",
      "[4/30] [74/594], loss: 0.1116, accyrasy: 0.953, f1: 0.8989  f1_dev: 0.6356\n",
      "[4/30] [79/594], loss: 0.1058, accyrasy: 0.957, f1: 0.915  f1_dev: 0.6193\n",
      "[4/30] [84/594], loss: 0.1058, accyrasy: 0.9595, f1: 0.9259  f1_dev: 0.627\n",
      "[4/30] [89/594], loss: 0.1086, accyrasy: 0.9614, f1: 0.9284  f1_dev: 0.6239\n",
      "[4/30] [94/594], loss: 0.1089, accyrasy: 0.9609, f1: 0.927  f1_dev: 0.6516\n",
      "[4/30] [99/594], loss: 0.1024, accyrasy: 0.9605, f1: 0.9275  f1_dev: 0.6369\n",
      "[4/30] [104/594], loss: 0.1062, accyrasy: 0.9589, f1: 0.9183  f1_dev: 0.6592\n",
      "[4/30] [109/594], loss: 0.1109, accyrasy: 0.9576, f1: 0.916  f1_dev: 0.6454\n",
      "[4/30] [114/594], loss: 0.1129, accyrasy: 0.9594, f1: 0.9249  f1_dev: 0.6416\n",
      "[4/30] [119/594], loss: 0.1106, accyrasy: 0.9626, f1: 0.9257  f1_dev: 0.6219\n",
      "[4/30] [124/594], loss: 0.1095, accyrasy: 0.9583, f1: 0.9231  f1_dev: 0.6451\n",
      "[4/30] [129/594], loss: 0.1024, accyrasy: 0.962, f1: 0.9268  f1_dev: 0.6565\n",
      "[4/30] [134/594], loss: 0.1044, accyrasy: 0.9649, f1: 0.9342  f1_dev: 0.6123\n",
      "[4/30] [139/594], loss: 0.1017, accyrasy: 0.9634, f1: 0.9303  f1_dev: 0.6288\n",
      "[4/30] [144/594], loss: 0.1036, accyrasy: 0.9595, f1: 0.9295  f1_dev: 0.6504\n",
      "[4/30] [149/594], loss: 0.1034, accyrasy: 0.965, f1: 0.936  f1_dev: 0.6415\n",
      "[4/30] [154/594], loss: 0.1081, accyrasy: 0.9606, f1: 0.9241  f1_dev: 0.6262\n",
      "[4/30] [159/594], loss: 0.1049, accyrasy: 0.9613, f1: 0.9239  f1_dev: 0.6224\n",
      "[4/30] [164/594], loss: 0.1072, accyrasy: 0.9605, f1: 0.9275  f1_dev: 0.6258\n",
      "[4/30] [169/594], loss: 0.1036, accyrasy: 0.9656, f1: 0.9325  f1_dev: 0.6492\n",
      "[4/30] [174/594], loss: 0.1089, accyrasy: 0.9596, f1: 0.9306  f1_dev: 0.6591\n",
      "[4/30] [179/594], loss: 0.1077, accyrasy: 0.9529, f1: 0.9095  f1_dev: 0.6625\n",
      "[4/30] [184/594], loss: 0.106, accyrasy: 0.9613, f1: 0.9245  f1_dev: 0.6323\n",
      "[4/30] [189/594], loss: 0.1071, accyrasy: 0.959, f1: 0.9251  f1_dev: 0.6436\n",
      "[4/30] [194/594], loss: 0.1083, accyrasy: 0.9567, f1: 0.9166  f1_dev: 0.6562\n",
      "[4/30] [199/594], loss: 0.1103, accyrasy: 0.9614, f1: 0.9264  f1_dev: 0.67\n",
      "[4/30] [204/594], loss: 0.1066, accyrasy: 0.9599, f1: 0.927  f1_dev: 0.638\n",
      "[4/30] [209/594], loss: 0.1143, accyrasy: 0.9534, f1: 0.9068  f1_dev: 0.6295\n",
      "[4/30] [214/594], loss: 0.1081, accyrasy: 0.9613, f1: 0.9232  f1_dev: 0.6301\n",
      "[4/30] [219/594], loss: 0.106, accyrasy: 0.9586, f1: 0.9126  f1_dev: 0.6287\n",
      "[4/30] [224/594], loss: 0.1106, accyrasy: 0.9554, f1: 0.9001  f1_dev: 0.6313\n",
      "[4/30] [229/594], loss: 0.1105, accyrasy: 0.9594, f1: 0.9237  f1_dev: 0.6517\n",
      "[4/30] [234/594], loss: 0.1043, accyrasy: 0.9625, f1: 0.9304  f1_dev: 0.5994\n",
      "[4/30] [239/594], loss: 0.1082, accyrasy: 0.9605, f1: 0.9265  f1_dev: 0.6247\n",
      "[4/30] [244/594], loss: 0.1059, accyrasy: 0.9611, f1: 0.9256  f1_dev: 0.6216\n",
      "[4/30] [249/594], loss: 0.1119, accyrasy: 0.9609, f1: 0.923  f1_dev: 0.6292\n",
      "[4/30] [254/594], loss: 0.1037, accyrasy: 0.9633, f1: 0.9351  f1_dev: 0.6589\n",
      "[4/30] [259/594], loss: 0.1074, accyrasy: 0.9599, f1: 0.9257  f1_dev: 0.6617\n",
      "[4/30] [264/594], loss: 0.1115, accyrasy: 0.9635, f1: 0.9361  f1_dev: 0.6626\n",
      "[4/30] [269/594], loss: 0.1082, accyrasy: 0.9616, f1: 0.9266  f1_dev: 0.6319\n",
      "[4/30] [274/594], loss: 0.1048, accyrasy: 0.961, f1: 0.928  f1_dev: 0.6483\n",
      "[4/30] [279/594], loss: 0.1117, accyrasy: 0.9624, f1: 0.9335  f1_dev: 0.6509\n",
      "[4/30] [284/594], loss: 0.1237, accyrasy: 0.9525, f1: 0.9158  f1_dev: 0.653\n",
      "[4/30] [289/594], loss: 0.1082, accyrasy: 0.9553, f1: 0.9184  f1_dev: 0.6181\n",
      "[4/30] [294/594], loss: 0.1056, accyrasy: 0.9625, f1: 0.9303  f1_dev: 0.6455\n",
      "[4/30] [299/594], loss: 0.109, accyrasy: 0.9621, f1: 0.9279  f1_dev: 0.6565\n",
      "[4/30] [304/594], loss: 0.106, accyrasy: 0.962, f1: 0.9225  f1_dev: 0.647\n",
      "[4/30] [309/594], loss: 0.1003, accyrasy: 0.9634, f1: 0.9291  f1_dev: 0.6594\n",
      "[4/30] [314/594], loss: 0.1056, accyrasy: 0.9624, f1: 0.9311  f1_dev: 0.635\n",
      "[4/30] [319/594], loss: 0.102, accyrasy: 0.9663, f1: 0.932  f1_dev: 0.6636\n",
      "[4/30] [324/594], loss: 0.1067, accyrasy: 0.9561, f1: 0.9061  f1_dev: 0.6087\n",
      "[4/30] [329/594], loss: 0.1067, accyrasy: 0.9596, f1: 0.9283  f1_dev: 0.6055\n",
      "[4/30] [334/594], loss: 0.1165, accyrasy: 0.9599, f1: 0.9237  f1_dev: 0.6489\n",
      "[4/30] [339/594], loss: 0.1216, accyrasy: 0.9499, f1: 0.9133  f1_dev: 0.6202\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[4/30] [344/594], loss: 0.1169, accyrasy: 0.9633, f1: 0.931  f1_dev: 0.6438\n",
      "[4/30] [349/594], loss: 0.1184, accyrasy: 0.9585, f1: 0.9201  f1_dev: 0.7221\n",
      "[4/30] [354/594], loss: 0.122, accyrasy: 0.9563, f1: 0.9131  f1_dev: 0.6839\n",
      "[4/30] [359/594], loss: 0.1101, accyrasy: 0.9629, f1: 0.9358  f1_dev: 0.6326\n",
      "[4/30] [364/594], loss: 0.1151, accyrasy: 0.9571, f1: 0.924  f1_dev: 0.6511\n",
      "[4/30] [369/594], loss: 0.1053, accyrasy: 0.9605, f1: 0.9278  f1_dev: 0.6657\n",
      "[4/30] [374/594], loss: 0.1049, accyrasy: 0.9624, f1: 0.9311  f1_dev: 0.6459\n",
      "[4/30] [379/594], loss: 0.0997, accyrasy: 0.961, f1: 0.9237  f1_dev: 0.6717\n",
      "[4/30] [384/594], loss: 0.1024, accyrasy: 0.9635, f1: 0.9262  f1_dev: 0.6487\n",
      "[4/30] [389/594], loss: 0.1031, accyrasy: 0.9599, f1: 0.9258  f1_dev: 0.6462\n",
      "[4/30] [394/594], loss: 0.1051, accyrasy: 0.9589, f1: 0.9263  f1_dev: 0.6526\n",
      "[4/30] [399/594], loss: 0.1025, accyrasy: 0.9581, f1: 0.918  f1_dev: 0.6255\n",
      "[4/30] [404/594], loss: 0.1057, accyrasy: 0.9646, f1: 0.9274  f1_dev: 0.592\n",
      "[4/30] [409/594], loss: 0.1205, accyrasy: 0.9594, f1: 0.9248  f1_dev: 0.6048\n",
      "[4/30] [414/594], loss: 0.1137, accyrasy: 0.9605, f1: 0.9324  f1_dev: 0.6089\n",
      "[4/30] [419/594], loss: 0.1121, accyrasy: 0.9599, f1: 0.925  f1_dev: 0.6691\n",
      "[4/30] [424/594], loss: 0.1073, accyrasy: 0.9639, f1: 0.9351  f1_dev: 0.7056\n",
      "[4/30] [429/594], loss: 0.1036, accyrasy: 0.963, f1: 0.9313  f1_dev: 0.6716\n",
      "[4/30] [434/594], loss: 0.1093, accyrasy: 0.9607, f1: 0.9264  f1_dev: 0.644\n",
      "[4/30] [439/594], loss: 0.1215, accyrasy: 0.9584, f1: 0.9233  f1_dev: 0.6244\n",
      "[4/30] [444/594], loss: 0.1056, accyrasy: 0.963, f1: 0.9246  f1_dev: 0.6208\n",
      "[4/30] [449/594], loss: 0.11, accyrasy: 0.9589, f1: 0.9259  f1_dev: 0.6521\n",
      "[4/30] [454/594], loss: 0.1068, accyrasy: 0.9599, f1: 0.9225  f1_dev: 0.6426\n",
      "[4/30] [459/594], loss: 0.1059, accyrasy: 0.9634, f1: 0.9243  f1_dev: 0.6351\n",
      "[4/30] [464/594], loss: 0.1082, accyrasy: 0.9635, f1: 0.9308  f1_dev: 0.6349\n",
      "[4/30] [469/594], loss: 0.1085, accyrasy: 0.9573, f1: 0.9264  f1_dev: 0.6665\n",
      "[4/30] [474/594], loss: 0.1107, accyrasy: 0.9544, f1: 0.9111  f1_dev: 0.6578\n",
      "[4/30] [479/594], loss: 0.1143, accyrasy: 0.9591, f1: 0.9325  f1_dev: 0.6636\n",
      "[4/30] [484/594], loss: 0.1092, accyrasy: 0.962, f1: 0.9341  f1_dev: 0.6456\n",
      "[4/30] [489/594], loss: 0.109, accyrasy: 0.9577, f1: 0.9182  f1_dev: 0.6243\n",
      "[4/30] [494/594], loss: 0.1129, accyrasy: 0.9591, f1: 0.9148  f1_dev: 0.6157\n",
      "[4/30] [499/594], loss: 0.1117, accyrasy: 0.9571, f1: 0.9263  f1_dev: 0.6737\n",
      "[4/30] [504/594], loss: 0.1212, accyrasy: 0.9549, f1: 0.915  f1_dev: 0.7083\n",
      "[4/30] [509/594], loss: 0.1227, accyrasy: 0.957, f1: 0.9175  f1_dev: 0.6889\n",
      "[4/30] [514/594], loss: 0.1149, accyrasy: 0.9543, f1: 0.9147  f1_dev: 0.611\n",
      "[4/30] [519/594], loss: 0.1149, accyrasy: 0.958, f1: 0.919  f1_dev: 0.6562\n",
      "[4/30] [524/594], loss: 0.1045, accyrasy: 0.963, f1: 0.9329  f1_dev: 0.6379\n",
      "[4/30] [529/594], loss: 0.1105, accyrasy: 0.9604, f1: 0.9264  f1_dev: 0.6482\n",
      "[4/30] [534/594], loss: 0.104, accyrasy: 0.9627, f1: 0.9299  f1_dev: 0.6486\n",
      "[4/30] [539/594], loss: 0.1055, accyrasy: 0.961, f1: 0.9269  f1_dev: 0.6577\n",
      "[4/30] [544/594], loss: 0.103, accyrasy: 0.9599, f1: 0.9219  f1_dev: 0.6501\n",
      "[4/30] [549/594], loss: 0.1049, accyrasy: 0.9627, f1: 0.9343  f1_dev: 0.6419\n",
      "[4/30] [554/594], loss: 0.1042, accyrasy: 0.9586, f1: 0.922  f1_dev: 0.636\n",
      "[4/30] [559/594], loss: 0.1026, accyrasy: 0.9557, f1: 0.9179  f1_dev: 0.6416\n",
      "[4/30] [564/594], loss: 0.119, accyrasy: 0.952, f1: 0.8757  f1_dev: 0.6018\n",
      "[4/30] [569/594], loss: 0.1185, accyrasy: 0.9579, f1: 0.9152  f1_dev: 0.5835\n",
      "[4/30] [574/594], loss: 0.1146, accyrasy: 0.9579, f1: 0.9245  f1_dev: 0.6274\n",
      "[4/30] [579/594], loss: 0.1126, accyrasy: 0.9554, f1: 0.9094  f1_dev: 0.6659\n",
      "[4/30] [584/594], loss: 0.1086, accyrasy: 0.9593, f1: 0.924  f1_dev: 0.6269\n",
      "[4/30] [589/594], loss: 0.1106, accyrasy: 0.9607, f1: 0.9201  f1_dev: 0.652\n",
      "[5/30] [4/594], loss: 0.1071, accyrasy: 0.9601, f1: 0.9279  f1_dev: 0.6168\n",
      "[5/30] [9/594], loss: 0.1059, accyrasy: 0.9611, f1: 0.9278  f1_dev: 0.6708\n",
      "[5/30] [14/594], loss: 0.1023, accyrasy: 0.9611, f1: 0.9247  f1_dev: 0.6493\n",
      "[5/30] [19/594], loss: 0.1043, accyrasy: 0.9603, f1: 0.9272  f1_dev: 0.6384\n",
      "[5/30] [24/594], loss: 0.1017, accyrasy: 0.9671, f1: 0.9423  f1_dev: 0.6446\n",
      "[5/30] [29/594], loss: 0.1053, accyrasy: 0.964, f1: 0.936  f1_dev: 0.6589\n",
      "[5/30] [34/594], loss: 0.1029, accyrasy: 0.964, f1: 0.9337  f1_dev: 0.6276\n",
      "[5/30] [39/594], loss: 0.1047, accyrasy: 0.961, f1: 0.9263  f1_dev: 0.6605\n",
      "[5/30] [44/594], loss: 0.1036, accyrasy: 0.9547, f1: 0.9108  f1_dev: 0.6685\n",
      "[5/30] [49/594], loss: 0.1145, accyrasy: 0.9576, f1: 0.9183  f1_dev: 0.6479\n",
      "[5/30] [54/594], loss: 0.1126, accyrasy: 0.9607, f1: 0.9223  f1_dev: 0.6342\n",
      "[5/30] [59/594], loss: 0.1099, accyrasy: 0.9606, f1: 0.9288  f1_dev: 0.6498\n",
      "[5/30] [64/594], loss: 0.1068, accyrasy: 0.9576, f1: 0.9135  f1_dev: 0.6854\n",
      "[5/30] [69/594], loss: 0.1075, accyrasy: 0.9589, f1: 0.9189  f1_dev: 0.6601\n",
      "[5/30] [74/594], loss: 0.1054, accyrasy: 0.9593, f1: 0.9224  f1_dev: 0.649\n",
      "[5/30] [79/594], loss: 0.1051, accyrasy: 0.9585, f1: 0.9174  f1_dev: 0.6687\n",
      "[5/30] [84/594], loss: 0.1048, accyrasy: 0.9631, f1: 0.9355  f1_dev: 0.6627\n",
      "[5/30] [89/594], loss: 0.1036, accyrasy: 0.9607, f1: 0.9298  f1_dev: 0.6517\n",
      "[5/30] [94/594], loss: 0.1068, accyrasy: 0.9533, f1: 0.9118  f1_dev: 0.6352\n",
      "[5/30] [99/594], loss: 0.1061, accyrasy: 0.9591, f1: 0.9241  f1_dev: 0.6335\n",
      "[5/30] [104/594], loss: 0.0998, accyrasy: 0.9633, f1: 0.9297  f1_dev: 0.6492\n",
      "[5/30] [109/594], loss: 0.1029, accyrasy: 0.9643, f1: 0.9293  f1_dev: 0.6365\n",
      "[5/30] [114/594], loss: 0.1091, accyrasy: 0.96, f1: 0.9252  f1_dev: 0.6553\n",
      "[5/30] [119/594], loss: 0.102, accyrasy: 0.9615, f1: 0.93  f1_dev: 0.6493\n",
      "[5/30] [124/594], loss: 0.1053, accyrasy: 0.9604, f1: 0.9188  f1_dev: 0.6238\n",
      "[5/30] [129/594], loss: 0.1278, accyrasy: 0.9525, f1: 0.9086  f1_dev: 0.6541\n",
      "[5/30] [134/594], loss: 0.1262, accyrasy: 0.9543, f1: 0.8916  f1_dev: 0.63\n",
      "[5/30] [139/594], loss: 0.1219, accyrasy: 0.9523, f1: 0.9243  f1_dev: 0.6755\n",
      "[5/30] [144/594], loss: 0.1182, accyrasy: 0.9615, f1: 0.927  f1_dev: 0.6646\n",
      "[5/30] [149/594], loss: 0.1228, accyrasy: 0.9591, f1: 0.9262  f1_dev: 0.6897\n",
      "[5/30] [154/594], loss: 0.1253, accyrasy: 0.9489, f1: 0.888  f1_dev: 0.6605\n",
      "[5/30] [159/594], loss: 0.1121, accyrasy: 0.9586, f1: 0.9159  f1_dev: 0.5858\n",
      "[5/30] [164/594], loss: 0.1137, accyrasy: 0.9604, f1: 0.924  f1_dev: 0.6765\n",
      "[5/30] [169/594], loss: 0.1063, accyrasy: 0.9626, f1: 0.9327  f1_dev: 0.6518\n",
      "[5/30] [174/594], loss: 0.1057, accyrasy: 0.9641, f1: 0.9353  f1_dev: 0.6326\n",
      "[5/30] [179/594], loss: 0.1079, accyrasy: 0.9606, f1: 0.9347  f1_dev: 0.6386\n",
      "[5/30] [184/594], loss: 0.1172, accyrasy: 0.9523, f1: 0.8916  f1_dev: 0.6791\n",
      "[5/30] [189/594], loss: 0.1111, accyrasy: 0.9604, f1: 0.9155  f1_dev: 0.6627\n",
      "[5/30] [194/594], loss: 0.1122, accyrasy: 0.9611, f1: 0.925  f1_dev: 0.6391\n",
      "[5/30] [199/594], loss: 0.1089, accyrasy: 0.9616, f1: 0.9273  f1_dev: 0.6199\n",
      "[5/30] [204/594], loss: 0.1063, accyrasy: 0.9603, f1: 0.9183  f1_dev: 0.6707\n",
      "[5/30] [209/594], loss: 0.1077, accyrasy: 0.9605, f1: 0.9241  f1_dev: 0.6385\n",
      "[5/30] [214/594], loss: 0.1086, accyrasy: 0.9605, f1: 0.9257  f1_dev: 0.6224\n",
      "[5/30] [219/594], loss: 0.1056, accyrasy: 0.9616, f1: 0.9294  f1_dev: 0.6553\n",
      "[5/30] [224/594], loss: 0.1078, accyrasy: 0.96, f1: 0.9297  f1_dev: 0.6579\n",
      "[5/30] [229/594], loss: 0.1154, accyrasy: 0.9617, f1: 0.9342  f1_dev: 0.6842\n",
      "[5/30] [234/594], loss: 0.1073, accyrasy: 0.959, f1: 0.9232  f1_dev: 0.6435\n",
      "[5/30] [239/594], loss: 0.1128, accyrasy: 0.963, f1: 0.9275  f1_dev: 0.5868\n",
      "[5/30] [244/594], loss: 0.1177, accyrasy: 0.9557, f1: 0.9172  f1_dev: 0.6763\n",
      "[5/30] [249/594], loss: 0.1082, accyrasy: 0.9579, f1: 0.9138  f1_dev: 0.6741\n",
      "[5/30] [254/594], loss: 0.1004, accyrasy: 0.9623, f1: 0.9288  f1_dev: 0.6517\n",
      "[5/30] [259/594], loss: 0.1048, accyrasy: 0.9554, f1: 0.9177  f1_dev: 0.6646\n",
      "[5/30] [264/594], loss: 0.1003, accyrasy: 0.9624, f1: 0.9309  f1_dev: 0.6425\n",
      "[5/30] [269/594], loss: 0.1095, accyrasy: 0.9557, f1: 0.9212  f1_dev: 0.6266\n",
      "[5/30] [274/594], loss: 0.1053, accyrasy: 0.9655, f1: 0.9319  f1_dev: 0.6572\n",
      "[5/30] [279/594], loss: 0.1076, accyrasy: 0.9627, f1: 0.9308  f1_dev: 0.6185\n",
      "[5/30] [284/594], loss: 0.1056, accyrasy: 0.9607, f1: 0.9219  f1_dev: 0.635\n",
      "[5/30] [289/594], loss: 0.1073, accyrasy: 0.9605, f1: 0.9017  f1_dev: 0.617\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[5/30] [294/594], loss: 0.1074, accyrasy: 0.9629, f1: 0.9259  f1_dev: 0.6732\n",
      "[5/30] [299/594], loss: 0.1049, accyrasy: 0.9599, f1: 0.9298  f1_dev: 0.658\n",
      "[5/30] [304/594], loss: 0.1099, accyrasy: 0.9591, f1: 0.9126  f1_dev: 0.6359\n",
      "[5/30] [309/594], loss: 0.1095, accyrasy: 0.9596, f1: 0.9218  f1_dev: 0.6612\n",
      "[5/30] [314/594], loss: 0.1031, accyrasy: 0.9569, f1: 0.9166  f1_dev: 0.665\n",
      "[5/30] [319/594], loss: 0.1104, accyrasy: 0.9586, f1: 0.9228  f1_dev: 0.6266\n",
      "[5/30] [324/594], loss: 0.1052, accyrasy: 0.9621, f1: 0.9277  f1_dev: 0.6414\n",
      "[5/30] [329/594], loss: 0.1062, accyrasy: 0.9627, f1: 0.9284  f1_dev: 0.6703\n",
      "[5/30] [334/594], loss: 0.1031, accyrasy: 0.9601, f1: 0.9278  f1_dev: 0.6263\n",
      "[5/30] [339/594], loss: 0.1035, accyrasy: 0.9613, f1: 0.929  f1_dev: 0.6441\n",
      "[5/30] [344/594], loss: 0.1084, accyrasy: 0.9601, f1: 0.9227  f1_dev: 0.6865\n",
      "[5/30] [349/594], loss: 0.1049, accyrasy: 0.9585, f1: 0.9214  f1_dev: 0.647\n",
      "[5/30] [354/594], loss: 0.1039, accyrasy: 0.9641, f1: 0.9255  f1_dev: 0.6103\n",
      "[5/30] [359/594], loss: 0.1031, accyrasy: 0.959, f1: 0.9277  f1_dev: 0.6337\n",
      "[5/30] [364/594], loss: 0.1033, accyrasy: 0.9635, f1: 0.9241  f1_dev: 0.6298\n",
      "[5/30] [369/594], loss: 0.1139, accyrasy: 0.9645, f1: 0.9359  f1_dev: 0.6282\n",
      "[5/30] [374/594], loss: 0.1026, accyrasy: 0.9617, f1: 0.9282  f1_dev: 0.6763\n",
      "[5/30] [379/594], loss: 0.1015, accyrasy: 0.959, f1: 0.9186  f1_dev: 0.6572\n",
      "[5/30] [384/594], loss: 0.1021, accyrasy: 0.9634, f1: 0.9353  f1_dev: 0.6498\n",
      "[5/30] [389/594], loss: 0.1083, accyrasy: 0.9594, f1: 0.9274  f1_dev: 0.6411\n",
      "[5/30] [394/594], loss: 0.0973, accyrasy: 0.962, f1: 0.9245  f1_dev: 0.6764\n",
      "[5/30] [399/594], loss: 0.1008, accyrasy: 0.9624, f1: 0.9328  f1_dev: 0.6739\n",
      "[5/30] [404/594], loss: 0.1006, accyrasy: 0.9619, f1: 0.9309  f1_dev: 0.6646\n",
      "[5/30] [409/594], loss: 0.1057, accyrasy: 0.958, f1: 0.9231  f1_dev: 0.6643\n",
      "[5/30] [414/594], loss: 0.1018, accyrasy: 0.9615, f1: 0.9317  f1_dev: 0.6745\n",
      "[5/30] [419/594], loss: 0.1091, accyrasy: 0.9591, f1: 0.9261  f1_dev: 0.6795\n",
      "[5/30] [424/594], loss: 0.1026, accyrasy: 0.9606, f1: 0.9223  f1_dev: 0.6632\n",
      "[5/30] [429/594], loss: 0.1015, accyrasy: 0.959, f1: 0.9218  f1_dev: 0.672\n",
      "[5/30] [434/594], loss: 0.1063, accyrasy: 0.963, f1: 0.921  f1_dev: 0.625\n",
      "[5/30] [439/594], loss: 0.1089, accyrasy: 0.9596, f1: 0.9207  f1_dev: 0.6334\n",
      "[5/30] [444/594], loss: 0.1012, accyrasy: 0.9609, f1: 0.92  f1_dev: 0.6314\n",
      "[5/30] [449/594], loss: 0.1068, accyrasy: 0.9544, f1: 0.9198  f1_dev: 0.6489\n",
      "[5/30] [454/594], loss: 0.1025, accyrasy: 0.9623, f1: 0.9285  f1_dev: 0.6626\n",
      "[5/30] [459/594], loss: 0.1032, accyrasy: 0.9613, f1: 0.9244  f1_dev: 0.6629\n",
      "[5/30] [464/594], loss: 0.1006, accyrasy: 0.9603, f1: 0.9258  f1_dev: 0.6568\n",
      "[5/30] [469/594], loss: 0.1052, accyrasy: 0.9553, f1: 0.9091  f1_dev: 0.6665\n",
      "[5/30] [474/594], loss: 0.1123, accyrasy: 0.9481, f1: 0.8789  f1_dev: 0.6425\n",
      "[5/30] [479/594], loss: 0.1032, accyrasy: 0.9613, f1: 0.9219  f1_dev: 0.6237\n",
      "[5/30] [484/594], loss: 0.1059, accyrasy: 0.9601, f1: 0.926  f1_dev: 0.639\n",
      "[5/30] [489/594], loss: 0.1063, accyrasy: 0.9591, f1: 0.9159  f1_dev: 0.6515\n",
      "[5/30] [494/594], loss: 0.1119, accyrasy: 0.9616, f1: 0.9262  f1_dev: 0.6332\n",
      "[5/30] [499/594], loss: 0.1084, accyrasy: 0.9607, f1: 0.9292  f1_dev: 0.6427\n",
      "[5/30] [504/594], loss: 0.1153, accyrasy: 0.9599, f1: 0.9299  f1_dev: 0.6514\n",
      "[5/30] [509/594], loss: 0.1135, accyrasy: 0.9577, f1: 0.9261  f1_dev: 0.6593\n",
      "[5/30] [514/594], loss: 0.1159, accyrasy: 0.9615, f1: 0.9246  f1_dev: 0.6593\n",
      "[5/30] [519/594], loss: 0.1073, accyrasy: 0.96, f1: 0.9248  f1_dev: 0.6526\n",
      "[5/30] [524/594], loss: 0.1055, accyrasy: 0.9579, f1: 0.9238  f1_dev: 0.6626\n",
      "[5/30] [529/594], loss: 0.1038, accyrasy: 0.9616, f1: 0.926  f1_dev: 0.6691\n",
      "[5/30] [534/594], loss: 0.1014, accyrasy: 0.9616, f1: 0.9243  f1_dev: 0.6602\n",
      "[5/30] [539/594], loss: 0.1028, accyrasy: 0.9624, f1: 0.9243  f1_dev: 0.6631\n",
      "[5/30] [544/594], loss: 0.1045, accyrasy: 0.9601, f1: 0.9243  f1_dev: 0.673\n",
      "[5/30] [549/594], loss: 0.1013, accyrasy: 0.9615, f1: 0.9212  f1_dev: 0.6355\n",
      "[5/30] [554/594], loss: 0.1099, accyrasy: 0.9609, f1: 0.9321  f1_dev: 0.6356\n",
      "[5/30] [559/594], loss: 0.1094, accyrasy: 0.9549, f1: 0.9079  f1_dev: 0.6303\n",
      "[5/30] [564/594], loss: 0.107, accyrasy: 0.9626, f1: 0.9263  f1_dev: 0.6276\n",
      "[5/30] [569/594], loss: 0.0999, accyrasy: 0.9627, f1: 0.9331  f1_dev: 0.6343\n",
      "[5/30] [574/594], loss: 0.1032, accyrasy: 0.961, f1: 0.9194  f1_dev: 0.6741\n",
      "[5/30] [579/594], loss: 0.1049, accyrasy: 0.9624, f1: 0.9273  f1_dev: 0.6617\n",
      "[5/30] [584/594], loss: 0.1014, accyrasy: 0.9591, f1: 0.9171  f1_dev: 0.6493\n",
      "[5/30] [589/594], loss: 0.1048, accyrasy: 0.964, f1: 0.9341  f1_dev: 0.653\n",
      "[6/30] [4/594], loss: 0.099, accyrasy: 0.9603, f1: 0.9192  f1_dev: 0.6592\n",
      "[6/30] [9/594], loss: 0.1006, accyrasy: 0.9609, f1: 0.9315  f1_dev: 0.6814\n",
      "[6/30] [14/594], loss: 0.1075, accyrasy: 0.9584, f1: 0.9217  f1_dev: 0.6699\n",
      "[6/30] [19/594], loss: 0.1027, accyrasy: 0.9631, f1: 0.9316  f1_dev: 0.6587\n",
      "[6/30] [24/594], loss: 0.1109, accyrasy: 0.9597, f1: 0.9308  f1_dev: 0.6557\n",
      "[6/30] [29/594], loss: 0.1113, accyrasy: 0.9647, f1: 0.9338  f1_dev: 0.7079\n",
      "[6/30] [34/594], loss: 0.1126, accyrasy: 0.9561, f1: 0.9027  f1_dev: 0.6707\n",
      "[6/30] [39/594], loss: 0.1033, accyrasy: 0.9647, f1: 0.9325  f1_dev: 0.706\n",
      "[6/30] [44/594], loss: 0.1027, accyrasy: 0.9605, f1: 0.9277  f1_dev: 0.6884\n",
      "[6/30] [49/594], loss: 0.1062, accyrasy: 0.9629, f1: 0.9333  f1_dev: 0.6731\n",
      "[6/30] [54/594], loss: 0.1038, accyrasy: 0.9631, f1: 0.9332  f1_dev: 0.65\n",
      "[6/30] [59/594], loss: 0.1059, accyrasy: 0.9619, f1: 0.9307  f1_dev: 0.6505\n",
      "[6/30] [64/594], loss: 0.0998, accyrasy: 0.9634, f1: 0.9287  f1_dev: 0.6692\n",
      "[6/30] [69/594], loss: 0.0991, accyrasy: 0.9661, f1: 0.937  f1_dev: 0.6778\n",
      "[6/30] [74/594], loss: 0.1044, accyrasy: 0.959, f1: 0.9279  f1_dev: 0.6854\n",
      "[6/30] [79/594], loss: 0.1102, accyrasy: 0.9581, f1: 0.9238  f1_dev: 0.6609\n",
      "[6/30] [84/594], loss: 0.1075, accyrasy: 0.9617, f1: 0.9254  f1_dev: 0.6367\n",
      "[6/30] [89/594], loss: 0.104, accyrasy: 0.9626, f1: 0.9334  f1_dev: 0.6704\n",
      "[6/30] [94/594], loss: 0.107, accyrasy: 0.9543, f1: 0.9082  f1_dev: 0.6119\n",
      "[6/30] [99/594], loss: 0.118, accyrasy: 0.9607, f1: 0.9245  f1_dev: 0.6327\n",
      "[6/30] [104/594], loss: 0.1175, accyrasy: 0.9539, f1: 0.8941  f1_dev: 0.6485\n",
      "[6/30] [109/594], loss: 0.1049, accyrasy: 0.961, f1: 0.9191  f1_dev: 0.6361\n",
      "[6/30] [114/594], loss: 0.1138, accyrasy: 0.9586, f1: 0.918  f1_dev: 0.6113\n",
      "[6/30] [119/594], loss: 0.1045, accyrasy: 0.9655, f1: 0.9355  f1_dev: 0.6629\n",
      "[6/30] [124/594], loss: 0.1071, accyrasy: 0.9616, f1: 0.9277  f1_dev: 0.6513\n",
      "[6/30] [129/594], loss: 0.106, accyrasy: 0.9589, f1: 0.9228  f1_dev: 0.6629\n",
      "[6/30] [134/594], loss: 0.1004, accyrasy: 0.963, f1: 0.9238  f1_dev: 0.6381\n",
      "[6/30] [139/594], loss: 0.1032, accyrasy: 0.9603, f1: 0.9227  f1_dev: 0.6761\n",
      "[6/30] [144/594], loss: 0.1032, accyrasy: 0.9655, f1: 0.933  f1_dev: 0.6523\n",
      "[6/30] [149/594], loss: 0.1036, accyrasy: 0.9584, f1: 0.9149  f1_dev: 0.6359\n",
      "[6/30] [154/594], loss: 0.1096, accyrasy: 0.9574, f1: 0.9097  f1_dev: 0.6455\n",
      "[6/30] [159/594], loss: 0.1079, accyrasy: 0.9639, f1: 0.9322  f1_dev: 0.6464\n",
      "[6/30] [164/594], loss: 0.1021, accyrasy: 0.9605, f1: 0.9096  f1_dev: 0.6566\n",
      "[6/30] [169/594], loss: 0.1022, accyrasy: 0.9604, f1: 0.9285  f1_dev: 0.6351\n",
      "[6/30] [174/594], loss: 0.1043, accyrasy: 0.9593, f1: 0.9132  f1_dev: 0.6562\n",
      "[6/30] [179/594], loss: 0.1017, accyrasy: 0.9643, f1: 0.9352  f1_dev: 0.6545\n",
      "[6/30] [184/594], loss: 0.1042, accyrasy: 0.9579, f1: 0.9137  f1_dev: 0.649\n",
      "[6/30] [189/594], loss: 0.1009, accyrasy: 0.964, f1: 0.9313  f1_dev: 0.6345\n",
      "[6/30] [194/594], loss: 0.102, accyrasy: 0.9613, f1: 0.9268  f1_dev: 0.6659\n",
      "[6/30] [199/594], loss: 0.1015, accyrasy: 0.9657, f1: 0.9342  f1_dev: 0.6575\n",
      "[6/30] [204/594], loss: 0.1004, accyrasy: 0.9651, f1: 0.9291  f1_dev: 0.6659\n",
      "[6/30] [209/594], loss: 0.1065, accyrasy: 0.9643, f1: 0.9343  f1_dev: 0.6369\n",
      "[6/30] [214/594], loss: 0.0992, accyrasy: 0.9627, f1: 0.9309  f1_dev: 0.667\n",
      "[6/30] [219/594], loss: 0.1014, accyrasy: 0.957, f1: 0.9222  f1_dev: 0.6803\n",
      "[6/30] [224/594], loss: 0.1076, accyrasy: 0.9614, f1: 0.9317  f1_dev: 0.6643\n",
      "[6/30] [229/594], loss: 0.1041, accyrasy: 0.9615, f1: 0.9267  f1_dev: 0.6685\n",
      "[6/30] [234/594], loss: 0.1004, accyrasy: 0.9631, f1: 0.9362  f1_dev: 0.7088\n",
      "[6/30] [239/594], loss: 0.1036, accyrasy: 0.9617, f1: 0.9256  f1_dev: 0.6785\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[6/30] [244/594], loss: 0.1062, accyrasy: 0.955, f1: 0.9132  f1_dev: 0.668\n",
      "[6/30] [249/594], loss: 0.1067, accyrasy: 0.9624, f1: 0.9324  f1_dev: 0.6538\n",
      "[6/30] [254/594], loss: 0.1053, accyrasy: 0.9585, f1: 0.9249  f1_dev: 0.655\n",
      "[6/30] [259/594], loss: 0.0957, accyrasy: 0.9657, f1: 0.9341  f1_dev: 0.6713\n",
      "[6/30] [264/594], loss: 0.1002, accyrasy: 0.9599, f1: 0.9241  f1_dev: 0.6556\n",
      "[6/30] [269/594], loss: 0.1008, accyrasy: 0.9601, f1: 0.9173  f1_dev: 0.6612\n",
      "[6/30] [274/594], loss: 0.1064, accyrasy: 0.9607, f1: 0.9248  f1_dev: 0.6204\n",
      "[6/30] [279/594], loss: 0.1107, accyrasy: 0.956, f1: 0.915  f1_dev: 0.7075\n",
      "[6/30] [284/594], loss: 0.103, accyrasy: 0.9627, f1: 0.9274  f1_dev: 0.6749\n",
      "[6/30] [289/594], loss: 0.1051, accyrasy: 0.9611, f1: 0.9218  f1_dev: 0.6578\n",
      "[6/30] [294/594], loss: 0.1053, accyrasy: 0.9599, f1: 0.9262  f1_dev: 0.6849\n",
      "[6/30] [299/594], loss: 0.0954, accyrasy: 0.9653, f1: 0.9289  f1_dev: 0.685\n",
      "[6/30] [304/594], loss: 0.0977, accyrasy: 0.9625, f1: 0.9314  f1_dev: 0.6676\n",
      "[6/30] [309/594], loss: 0.0984, accyrasy: 0.9629, f1: 0.9285  f1_dev: 0.6768\n",
      "[6/30] [314/594], loss: 0.0982, accyrasy: 0.9675, f1: 0.9419  f1_dev: 0.6847\n",
      "[6/30] [319/594], loss: 0.1015, accyrasy: 0.9609, f1: 0.9278  f1_dev: 0.7284\n",
      "[6/30] [324/594], loss: 0.1024, accyrasy: 0.9595, f1: 0.9238  f1_dev: 0.6766\n",
      "[6/30] [329/594], loss: 0.1005, accyrasy: 0.9594, f1: 0.9236  f1_dev: 0.6581\n",
      "[6/30] [334/594], loss: 0.099, accyrasy: 0.9607, f1: 0.9247  f1_dev: 0.6522\n",
      "[6/30] [339/594], loss: 0.0995, accyrasy: 0.9659, f1: 0.9342  f1_dev: 0.6448\n",
      "[6/30] [344/594], loss: 0.108, accyrasy: 0.958, f1: 0.9174  f1_dev: 0.6577\n",
      "[6/30] [349/594], loss: 0.0995, accyrasy: 0.9647, f1: 0.9356  f1_dev: 0.6655\n",
      "[6/30] [354/594], loss: 0.104, accyrasy: 0.9594, f1: 0.9209  f1_dev: 0.6575\n",
      "[6/30] [359/594], loss: 0.0981, accyrasy: 0.9644, f1: 0.9274  f1_dev: 0.6634\n",
      "[6/30] [364/594], loss: 0.1033, accyrasy: 0.9629, f1: 0.9361  f1_dev: 0.6629\n",
      "[6/30] [369/594], loss: 0.1038, accyrasy: 0.9621, f1: 0.9276  f1_dev: 0.6706\n",
      "[6/30] [374/594], loss: 0.1077, accyrasy: 0.9595, f1: 0.9211  f1_dev: 0.6605\n",
      "[6/30] [379/594], loss: 0.1017, accyrasy: 0.9617, f1: 0.9324  f1_dev: 0.6722\n",
      "[6/30] [384/594], loss: 0.1075, accyrasy: 0.9569, f1: 0.9219  f1_dev: 0.6526\n",
      "[6/30] [389/594], loss: 0.1089, accyrasy: 0.9546, f1: 0.9061  f1_dev: 0.6692\n",
      "[6/30] [394/594], loss: 0.1051, accyrasy: 0.958, f1: 0.9225  f1_dev: 0.6562\n",
      "[6/30] [399/594], loss: 0.1043, accyrasy: 0.9593, f1: 0.9238  f1_dev: 0.6694\n",
      "[6/30] [404/594], loss: 0.1022, accyrasy: 0.9585, f1: 0.9172  f1_dev: 0.6816\n",
      "[6/30] [409/594], loss: 0.1024, accyrasy: 0.9664, f1: 0.9364  f1_dev: 0.7041\n",
      "[6/30] [414/594], loss: 0.1035, accyrasy: 0.9614, f1: 0.9255  f1_dev: 0.6929\n",
      "[6/30] [419/594], loss: 0.0987, accyrasy: 0.9634, f1: 0.9339  f1_dev: 0.6893\n",
      "[6/30] [424/594], loss: 0.0972, accyrasy: 0.9614, f1: 0.9228  f1_dev: 0.6517\n",
      "[6/30] [429/594], loss: 0.1127, accyrasy: 0.9593, f1: 0.914  f1_dev: 0.6324\n",
      "[6/30] [434/594], loss: 0.1058, accyrasy: 0.961, f1: 0.9241  f1_dev: 0.6452\n",
      "[6/30] [439/594], loss: 0.123, accyrasy: 0.9533, f1: 0.9151  f1_dev: 0.6908\n",
      "[6/30] [444/594], loss: 0.1116, accyrasy: 0.9576, f1: 0.9182  f1_dev: 0.683\n",
      "[6/30] [449/594], loss: 0.1046, accyrasy: 0.9646, f1: 0.9299  f1_dev: 0.6541\n",
      "[6/30] [454/594], loss: 0.1063, accyrasy: 0.957, f1: 0.9057  f1_dev: 0.66\n",
      "[6/30] [459/594], loss: 0.1022, accyrasy: 0.9635, f1: 0.9282  f1_dev: 0.6723\n",
      "[6/30] [464/594], loss: 0.1017, accyrasy: 0.9644, f1: 0.9382  f1_dev: 0.6745\n",
      "[6/30] [469/594], loss: 0.1008, accyrasy: 0.9627, f1: 0.9231  f1_dev: 0.6471\n",
      "[6/30] [474/594], loss: 0.0975, accyrasy: 0.9619, f1: 0.9292  f1_dev: 0.6746\n",
      "[6/30] [479/594], loss: 0.1004, accyrasy: 0.9634, f1: 0.9325  f1_dev: 0.6526\n",
      "[6/30] [484/594], loss: 0.1019, accyrasy: 0.9623, f1: 0.933  f1_dev: 0.6467\n",
      "[6/30] [489/594], loss: 0.1051, accyrasy: 0.9627, f1: 0.9313  f1_dev: 0.6694\n",
      "[6/30] [494/594], loss: 0.103, accyrasy: 0.9624, f1: 0.9251  f1_dev: 0.7066\n",
      "[6/30] [499/594], loss: 0.1129, accyrasy: 0.9664, f1: 0.9362  f1_dev: 0.7906\n",
      "[6/30] [504/594], loss: 0.11, accyrasy: 0.9583, f1: 0.919  f1_dev: 0.7849\n",
      "[6/30] [509/594], loss: 0.1074, accyrasy: 0.9629, f1: 0.9261  f1_dev: 0.6763\n",
      "[6/30] [514/594], loss: 0.1001, accyrasy: 0.9631, f1: 0.9354  f1_dev: 0.6597\n",
      "[6/30] [519/594], loss: 0.1091, accyrasy: 0.9593, f1: 0.9181  f1_dev: 0.6528\n",
      "[6/30] [524/594], loss: 0.1235, accyrasy: 0.9526, f1: 0.9045  f1_dev: 0.6851\n",
      "[6/30] [529/594], loss: 0.1071, accyrasy: 0.963, f1: 0.9316  f1_dev: 0.6731\n",
      "[6/30] [534/594], loss: 0.1027, accyrasy: 0.9644, f1: 0.9323  f1_dev: 0.665\n",
      "[6/30] [539/594], loss: 0.1046, accyrasy: 0.9565, f1: 0.9186  f1_dev: 0.6453\n",
      "[6/30] [544/594], loss: 0.1024, accyrasy: 0.9615, f1: 0.9304  f1_dev: 0.6706\n",
      "[6/30] [549/594], loss: 0.0952, accyrasy: 0.9617, f1: 0.92  f1_dev: 0.6645\n",
      "[6/30] [554/594], loss: 0.0905, accyrasy: 0.9673, f1: 0.9376  f1_dev: 0.666\n",
      "[6/30] [559/594], loss: 0.0986, accyrasy: 0.962, f1: 0.9325  f1_dev: 0.6772\n",
      "[6/30] [564/594], loss: 0.0959, accyrasy: 0.9641, f1: 0.9348  f1_dev: 0.6714\n",
      "[6/30] [569/594], loss: 0.1025, accyrasy: 0.9606, f1: 0.9308  f1_dev: 0.6763\n",
      "[6/30] [574/594], loss: 0.0955, accyrasy: 0.9676, f1: 0.9405  f1_dev: 0.6652\n",
      "[6/30] [579/594], loss: 0.0979, accyrasy: 0.9637, f1: 0.93  f1_dev: 0.6682\n",
      "[6/30] [584/594], loss: 0.1002, accyrasy: 0.9654, f1: 0.9362  f1_dev: 0.6688\n",
      "[6/30] [589/594], loss: 0.1088, accyrasy: 0.9616, f1: 0.9277  f1_dev: 0.652\n",
      "[7/30] [4/594], loss: 0.0996, accyrasy: 0.9646, f1: 0.9298  f1_dev: 0.6734\n",
      "[7/30] [9/594], loss: 0.1001, accyrasy: 0.9635, f1: 0.9336  f1_dev: 0.6637\n",
      "[7/30] [14/594], loss: 0.0971, accyrasy: 0.9625, f1: 0.9293  f1_dev: 0.6652\n",
      "[7/30] [19/594], loss: 0.0969, accyrasy: 0.9655, f1: 0.9315  f1_dev: 0.6857\n",
      "[7/30] [24/594], loss: 0.0971, accyrasy: 0.9639, f1: 0.9324  f1_dev: 0.6675\n",
      "[7/30] [29/594], loss: 0.0993, accyrasy: 0.9643, f1: 0.9349  f1_dev: 0.6455\n",
      "[7/30] [34/594], loss: 0.0995, accyrasy: 0.9647, f1: 0.9343  f1_dev: 0.66\n",
      "[7/30] [39/594], loss: 0.0996, accyrasy: 0.9647, f1: 0.9356  f1_dev: 0.6587\n",
      "[7/30] [44/594], loss: 0.1084, accyrasy: 0.9569, f1: 0.9193  f1_dev: 0.6571\n",
      "[7/30] [49/594], loss: 0.102, accyrasy: 0.9554, f1: 0.9236  f1_dev: 0.6749\n",
      "[7/30] [54/594], loss: 0.1077, accyrasy: 0.9644, f1: 0.9301  f1_dev: 0.6891\n",
      "[7/30] [59/594], loss: 0.1054, accyrasy: 0.9619, f1: 0.9341  f1_dev: 0.6889\n",
      "[7/30] [64/594], loss: 0.1066, accyrasy: 0.9604, f1: 0.9215  f1_dev: 0.6996\n",
      "[7/30] [69/594], loss: 0.0992, accyrasy: 0.9627, f1: 0.9278  f1_dev: 0.7296\n",
      "[7/30] [74/594], loss: 0.0995, accyrasy: 0.9601, f1: 0.9269  f1_dev: 0.7078\n",
      "[7/30] [79/594], loss: 0.1076, accyrasy: 0.955, f1: 0.9009  f1_dev: 0.6539\n",
      "[7/30] [84/594], loss: 0.1031, accyrasy: 0.9644, f1: 0.9321  f1_dev: 0.6714\n",
      "[7/30] [89/594], loss: 0.0973, accyrasy: 0.9617, f1: 0.9296  f1_dev: 0.6614\n",
      "[7/30] [94/594], loss: 0.0956, accyrasy: 0.9673, f1: 0.9349  f1_dev: 0.6773\n",
      "[7/30] [99/594], loss: 0.0954, accyrasy: 0.9651, f1: 0.9352  f1_dev: 0.7054\n",
      "[7/30] [104/594], loss: 0.1027, accyrasy: 0.9626, f1: 0.9244  f1_dev: 0.6485\n",
      "[7/30] [109/594], loss: 0.0996, accyrasy: 0.965, f1: 0.9366  f1_dev: 0.6497\n",
      "[7/30] [114/594], loss: 0.103, accyrasy: 0.9606, f1: 0.9181  f1_dev: 0.6643\n",
      "[7/30] [119/594], loss: 0.0975, accyrasy: 0.9626, f1: 0.9286  f1_dev: 0.6943\n",
      "[7/30] [124/594], loss: 0.0989, accyrasy: 0.9607, f1: 0.924  f1_dev: 0.6575\n",
      "[7/30] [129/594], loss: 0.0961, accyrasy: 0.9613, f1: 0.923  f1_dev: 0.6518\n",
      "[7/30] [134/594], loss: 0.0976, accyrasy: 0.9647, f1: 0.9379  f1_dev: 0.6557\n",
      "[7/30] [139/594], loss: 0.1001, accyrasy: 0.9626, f1: 0.9359  f1_dev: 0.6853\n",
      "[7/30] [144/594], loss: 0.095, accyrasy: 0.9665, f1: 0.9389  f1_dev: 0.6829\n",
      "[7/30] [149/594], loss: 0.0946, accyrasy: 0.9634, f1: 0.9279  f1_dev: 0.6987\n",
      "[7/30] [154/594], loss: 0.0968, accyrasy: 0.9666, f1: 0.9337  f1_dev: 0.6775\n",
      "[7/30] [159/594], loss: 0.0932, accyrasy: 0.9631, f1: 0.9215  f1_dev: 0.6435\n",
      "[7/30] [164/594], loss: 0.1025, accyrasy: 0.9625, f1: 0.9239  f1_dev: 0.69\n",
      "[7/30] [169/594], loss: 0.1082, accyrasy: 0.9561, f1: 0.9069  f1_dev: 0.6483\n",
      "[7/30] [174/594], loss: 0.1007, accyrasy: 0.9655, f1: 0.9347  f1_dev: 0.6688\n",
      "[7/30] [179/594], loss: 0.1002, accyrasy: 0.9649, f1: 0.9357  f1_dev: 0.6888\n",
      "[7/30] [184/594], loss: 0.1081, accyrasy: 0.9577, f1: 0.9205  f1_dev: 0.7152\n",
      "[7/30] [189/594], loss: 0.1048, accyrasy: 0.9593, f1: 0.9152  f1_dev: 0.7274\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[7/30] [194/594], loss: 0.1021, accyrasy: 0.9614, f1: 0.9325  f1_dev: 0.6868\n",
      "[7/30] [199/594], loss: 0.1043, accyrasy: 0.9646, f1: 0.9374  f1_dev: 0.6609\n",
      "[7/30] [204/594], loss: 0.1156, accyrasy: 0.9591, f1: 0.9219  f1_dev: 0.691\n",
      "[7/30] [209/594], loss: 0.1101, accyrasy: 0.9604, f1: 0.9289  f1_dev: 0.6637\n",
      "[7/30] [214/594], loss: 0.1054, accyrasy: 0.9559, f1: 0.9139  f1_dev: 0.6984\n",
      "[7/30] [219/594], loss: 0.1005, accyrasy: 0.96, f1: 0.9262  f1_dev: 0.6574\n",
      "[7/30] [224/594], loss: 0.1012, accyrasy: 0.9604, f1: 0.9254  f1_dev: 0.6929\n",
      "[7/30] [229/594], loss: 0.1011, accyrasy: 0.963, f1: 0.9333  f1_dev: 0.6842\n",
      "[7/30] [234/594], loss: 0.1037, accyrasy: 0.958, f1: 0.9197  f1_dev: 0.6945\n",
      "[7/30] [239/594], loss: 0.1045, accyrasy: 0.9619, f1: 0.9334  f1_dev: 0.6978\n",
      "[7/30] [244/594], loss: 0.1154, accyrasy: 0.9509, f1: 0.91  f1_dev: 0.6973\n",
      "[7/30] [249/594], loss: 0.1052, accyrasy: 0.9629, f1: 0.9198  f1_dev: 0.6423\n",
      "[7/30] [254/594], loss: 0.1074, accyrasy: 0.9614, f1: 0.9276  f1_dev: 0.6733\n",
      "[7/30] [259/594], loss: 0.1052, accyrasy: 0.9569, f1: 0.9225  f1_dev: 0.6827\n",
      "[7/30] [264/594], loss: 0.1048, accyrasy: 0.9639, f1: 0.9323  f1_dev: 0.6721\n",
      "[7/30] [269/594], loss: 0.0999, accyrasy: 0.9626, f1: 0.9337  f1_dev: 0.7565\n",
      "[7/30] [274/594], loss: 0.1025, accyrasy: 0.9626, f1: 0.9312  f1_dev: 0.6694\n",
      "[7/30] [279/594], loss: 0.096, accyrasy: 0.9654, f1: 0.935  f1_dev: 0.6666\n",
      "[7/30] [284/594], loss: 0.0995, accyrasy: 0.964, f1: 0.9289  f1_dev: 0.6839\n",
      "[7/30] [289/594], loss: 0.097, accyrasy: 0.9647, f1: 0.9357  f1_dev: 0.6679\n",
      "[7/30] [294/594], loss: 0.0997, accyrasy: 0.964, f1: 0.9343  f1_dev: 0.668\n",
      "[7/30] [299/594], loss: 0.0992, accyrasy: 0.966, f1: 0.9353  f1_dev: 0.6951\n",
      "[7/30] [304/594], loss: 0.1007, accyrasy: 0.9631, f1: 0.9255  f1_dev: 0.6732\n",
      "[7/30] [309/594], loss: 0.0971, accyrasy: 0.9689, f1: 0.9402  f1_dev: 0.7091\n",
      "[7/30] [314/594], loss: 0.0952, accyrasy: 0.9607, f1: 0.9265  f1_dev: 0.7035\n",
      "[7/30] [319/594], loss: 0.0989, accyrasy: 0.9603, f1: 0.9207  f1_dev: 0.6358\n",
      "[7/30] [324/594], loss: 0.1025, accyrasy: 0.9571, f1: 0.9215  f1_dev: 0.6206\n",
      "[7/30] [329/594], loss: 0.1082, accyrasy: 0.9641, f1: 0.9294  f1_dev: 0.6683\n",
      "[7/30] [334/594], loss: 0.1056, accyrasy: 0.9623, f1: 0.9299  f1_dev: 0.6392\n",
      "[7/30] [339/594], loss: 0.1029, accyrasy: 0.9623, f1: 0.9305  f1_dev: 0.6759\n",
      "[7/30] [344/594], loss: 0.1037, accyrasy: 0.9621, f1: 0.9244  f1_dev: 0.6817\n",
      "[7/30] [349/594], loss: 0.1079, accyrasy: 0.9563, f1: 0.9146  f1_dev: 0.6761\n",
      "[7/30] [354/594], loss: 0.1037, accyrasy: 0.9623, f1: 0.9318  f1_dev: 0.6643\n",
      "[7/30] [359/594], loss: 0.1025, accyrasy: 0.9596, f1: 0.9262  f1_dev: 0.6733\n",
      "[7/30] [364/594], loss: 0.1055, accyrasy: 0.9616, f1: 0.9224  f1_dev: 0.6859\n",
      "[7/30] [369/594], loss: 0.1008, accyrasy: 0.9637, f1: 0.9345  f1_dev: 0.7048\n",
      "[7/30] [374/594], loss: 0.1044, accyrasy: 0.9629, f1: 0.9358  f1_dev: 0.6805\n",
      "[7/30] [379/594], loss: 0.0961, accyrasy: 0.9637, f1: 0.9284  f1_dev: 0.6392\n",
      "[7/30] [384/594], loss: 0.0963, accyrasy: 0.9659, f1: 0.9308  f1_dev: 0.6569\n",
      "[7/30] [389/594], loss: 0.0931, accyrasy: 0.9653, f1: 0.9253  f1_dev: 0.6638\n",
      "[7/30] [394/594], loss: 0.0996, accyrasy: 0.9621, f1: 0.9328  f1_dev: 0.6957\n",
      "[7/30] [399/594], loss: 0.0968, accyrasy: 0.9621, f1: 0.925  f1_dev: 0.694\n",
      "[7/30] [404/594], loss: 0.1, accyrasy: 0.9635, f1: 0.925  f1_dev: 0.696\n",
      "[7/30] [409/594], loss: 0.0988, accyrasy: 0.9649, f1: 0.9211  f1_dev: 0.807\n",
      "[7/30] [414/594], loss: 0.0964, accyrasy: 0.9654, f1: 0.9251  f1_dev: 0.8132\n",
      "[7/30] [419/594], loss: 0.0952, accyrasy: 0.9635, f1: 0.9286  f1_dev: 0.6984\n",
      "[7/30] [424/594], loss: 0.1047, accyrasy: 0.9481, f1: 0.8976  f1_dev: 0.7999\n",
      "[7/30] [429/594], loss: 0.101, accyrasy: 0.9593, f1: 0.917  f1_dev: 0.6988\n",
      "[7/30] [434/594], loss: 0.0962, accyrasy: 0.9645, f1: 0.9284  f1_dev: 0.6984\n",
      "[7/30] [439/594], loss: 0.0924, accyrasy: 0.9684, f1: 0.9389  f1_dev: 0.6872\n",
      "[7/30] [444/594], loss: 0.0983, accyrasy: 0.9625, f1: 0.927  f1_dev: 0.7273\n",
      "[7/30] [449/594], loss: 0.1018, accyrasy: 0.9629, f1: 0.9299  f1_dev: 0.7254\n",
      "[7/30] [454/594], loss: 0.0952, accyrasy: 0.9663, f1: 0.9367  f1_dev: 0.7015\n",
      "[7/30] [459/594], loss: 0.0934, accyrasy: 0.9681, f1: 0.9388  f1_dev: 0.6912\n",
      "[7/30] [464/594], loss: 0.0972, accyrasy: 0.9627, f1: 0.9314  f1_dev: 0.7003\n",
      "[7/30] [469/594], loss: 0.0945, accyrasy: 0.9625, f1: 0.9303  f1_dev: 0.6922\n",
      "[7/30] [474/594], loss: 0.1025, accyrasy: 0.9631, f1: 0.9278  f1_dev: 0.6918\n",
      "[7/30] [479/594], loss: 0.101, accyrasy: 0.9605, f1: 0.9241  f1_dev: 0.682\n",
      "[7/30] [484/594], loss: 0.1001, accyrasy: 0.9555, f1: 0.906  f1_dev: 0.6692\n",
      "[7/30] [489/594], loss: 0.098, accyrasy: 0.9636, f1: 0.9253  f1_dev: 0.6839\n",
      "[7/30] [494/594], loss: 0.0958, accyrasy: 0.9649, f1: 0.9309  f1_dev: 0.6554\n",
      "[7/30] [499/594], loss: 0.1005, accyrasy: 0.9626, f1: 0.9329  f1_dev: 0.6699\n",
      "[7/30] [504/594], loss: 0.0993, accyrasy: 0.9654, f1: 0.936  f1_dev: 0.68\n",
      "[7/30] [509/594], loss: 0.101, accyrasy: 0.9649, f1: 0.9261  f1_dev: 0.6784\n",
      "[7/30] [514/594], loss: 0.0982, accyrasy: 0.966, f1: 0.9388  f1_dev: 0.7144\n",
      "[7/30] [519/594], loss: 0.0954, accyrasy: 0.9616, f1: 0.9283  f1_dev: 0.7239\n",
      "[7/30] [524/594], loss: 0.0957, accyrasy: 0.964, f1: 0.9259  f1_dev: 0.6849\n",
      "[7/30] [529/594], loss: 0.1012, accyrasy: 0.9657, f1: 0.9351  f1_dev: 0.6831\n",
      "[7/30] [534/594], loss: 0.0969, accyrasy: 0.9624, f1: 0.9322  f1_dev: 0.7012\n",
      "[7/30] [539/594], loss: 0.0983, accyrasy: 0.9587, f1: 0.9168  f1_dev: 0.6979\n",
      "[7/30] [544/594], loss: 0.0975, accyrasy: 0.9631, f1: 0.924  f1_dev: 0.7124\n",
      "[7/30] [549/594], loss: 0.0998, accyrasy: 0.9645, f1: 0.938  f1_dev: 0.7079\n",
      "[7/30] [554/594], loss: 0.0989, accyrasy: 0.9639, f1: 0.9339  f1_dev: 0.676\n",
      "[7/30] [559/594], loss: 0.0919, accyrasy: 0.965, f1: 0.9369  f1_dev: 0.6735\n",
      "[7/30] [564/594], loss: 0.0962, accyrasy: 0.9617, f1: 0.9192  f1_dev: 0.7227\n",
      "[7/30] [569/594], loss: 0.1015, accyrasy: 0.9561, f1: 0.916  f1_dev: 0.8028\n",
      "[7/30] [574/594], loss: 0.1027, accyrasy: 0.9587, f1: 0.9231  f1_dev: 0.7336\n",
      "[7/30] [579/594], loss: 0.0986, accyrasy: 0.9637, f1: 0.9292  f1_dev: 0.8652\n",
      "[7/30] [584/594], loss: 0.0996, accyrasy: 0.958, f1: 0.9195  f1_dev: 0.7039\n",
      "[7/30] [589/594], loss: 0.1084, accyrasy: 0.9621, f1: 0.9307  f1_dev: 0.674\n",
      "[8/30] [4/594], loss: 0.1084, accyrasy: 0.9596, f1: 0.9273  f1_dev: 0.6983\n",
      "[8/30] [9/594], loss: 0.1092, accyrasy: 0.9569, f1: 0.9126  f1_dev: 0.6691\n",
      "[8/30] [14/594], loss: 0.103, accyrasy: 0.9609, f1: 0.9314  f1_dev: 0.7076\n",
      "[8/30] [19/594], loss: 0.1027, accyrasy: 0.9625, f1: 0.9266  f1_dev: 0.6843\n",
      "[8/30] [24/594], loss: 0.1009, accyrasy: 0.965, f1: 0.929  f1_dev: 0.6913\n",
      "[8/30] [29/594], loss: 0.1025, accyrasy: 0.9636, f1: 0.9281  f1_dev: 0.6655\n",
      "[8/30] [34/594], loss: 0.0922, accyrasy: 0.9633, f1: 0.928  f1_dev: 0.6523\n",
      "[8/30] [39/594], loss: 0.096, accyrasy: 0.9679, f1: 0.9395  f1_dev: 0.6651\n",
      "[8/30] [44/594], loss: 0.1037, accyrasy: 0.9563, f1: 0.9099  f1_dev: 0.672\n",
      "[8/30] [49/594], loss: 0.0938, accyrasy: 0.9688, f1: 0.9418  f1_dev: 0.6982\n",
      "[8/30] [54/594], loss: 0.0984, accyrasy: 0.9639, f1: 0.9255  f1_dev: 0.7172\n",
      "[8/30] [59/594], loss: 0.0956, accyrasy: 0.963, f1: 0.9245  f1_dev: 0.736\n",
      "[8/30] [64/594], loss: 0.0901, accyrasy: 0.9641, f1: 0.9342  f1_dev: 0.7298\n",
      "[8/30] [69/594], loss: 0.0951, accyrasy: 0.9644, f1: 0.9341  f1_dev: 0.6634\n",
      "[8/30] [74/594], loss: 0.1034, accyrasy: 0.9604, f1: 0.9287  f1_dev: 0.7056\n",
      "[8/30] [79/594], loss: 0.1003, accyrasy: 0.9634, f1: 0.9347  f1_dev: 0.7347\n",
      "[8/30] [84/594], loss: 0.0949, accyrasy: 0.961, f1: 0.9275  f1_dev: 0.7017\n",
      "[8/30] [89/594], loss: 0.0973, accyrasy: 0.9643, f1: 0.9315  f1_dev: 0.6851\n",
      "[8/30] [94/594], loss: 0.0981, accyrasy: 0.9649, f1: 0.9283  f1_dev: 0.6773\n",
      "[8/30] [99/594], loss: 0.0954, accyrasy: 0.9657, f1: 0.9327  f1_dev: 0.7107\n",
      "[8/30] [104/594], loss: 0.0918, accyrasy: 0.9627, f1: 0.9276  f1_dev: 0.695\n",
      "[8/30] [109/594], loss: 0.0963, accyrasy: 0.9617, f1: 0.9232  f1_dev: 0.7323\n",
      "[8/30] [114/594], loss: 0.0939, accyrasy: 0.9664, f1: 0.9367  f1_dev: 0.7497\n",
      "[8/30] [119/594], loss: 0.0944, accyrasy: 0.9667, f1: 0.9342  f1_dev: 0.6897\n",
      "[8/30] [124/594], loss: 0.0923, accyrasy: 0.9654, f1: 0.9385  f1_dev: 0.7567\n",
      "[8/30] [129/594], loss: 0.099, accyrasy: 0.9571, f1: 0.9116  f1_dev: 0.7416\n",
      "[8/30] [134/594], loss: 0.0947, accyrasy: 0.9669, f1: 0.9378  f1_dev: 0.7171\n",
      "[8/30] [139/594], loss: 0.0912, accyrasy: 0.9665, f1: 0.938  f1_dev: 0.7197\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[8/30] [144/594], loss: 0.0915, accyrasy: 0.966, f1: 0.9262  f1_dev: 0.7416\n",
      "[8/30] [149/594], loss: 0.097, accyrasy: 0.9634, f1: 0.9279  f1_dev: 0.7423\n",
      "[8/30] [154/594], loss: 0.0936, accyrasy: 0.9675, f1: 0.9303  f1_dev: 0.7559\n",
      "[8/30] [159/594], loss: 0.0967, accyrasy: 0.962, f1: 0.9336  f1_dev: 0.7291\n",
      "[8/30] [164/594], loss: 0.1054, accyrasy: 0.9617, f1: 0.9303  f1_dev: 0.7434\n",
      "[8/30] [169/594], loss: 0.0991, accyrasy: 0.9681, f1: 0.9407  f1_dev: 0.6945\n",
      "[8/30] [174/594], loss: 0.0976, accyrasy: 0.9613, f1: 0.9252  f1_dev: 0.6938\n",
      "[8/30] [179/594], loss: 0.0962, accyrasy: 0.9603, f1: 0.924  f1_dev: 0.6753\n",
      "[8/30] [184/594], loss: 0.095, accyrasy: 0.9647, f1: 0.9272  f1_dev: 0.6677\n",
      "[8/30] [189/594], loss: 0.0992, accyrasy: 0.963, f1: 0.9246  f1_dev: 0.6839\n",
      "[8/30] [194/594], loss: 0.0983, accyrasy: 0.967, f1: 0.9363  f1_dev: 0.7065\n",
      "[8/30] [199/594], loss: 0.098, accyrasy: 0.9625, f1: 0.9269  f1_dev: 0.6908\n",
      "[8/30] [204/594], loss: 0.0925, accyrasy: 0.9604, f1: 0.9254  f1_dev: 0.7196\n",
      "[8/30] [209/594], loss: 0.0937, accyrasy: 0.9639, f1: 0.9299  f1_dev: 0.8135\n",
      "[8/30] [214/594], loss: 0.0973, accyrasy: 0.9639, f1: 0.9297  f1_dev: 0.6726\n",
      "[8/30] [219/594], loss: 0.0955, accyrasy: 0.9604, f1: 0.9172  f1_dev: 0.7351\n",
      "[8/30] [224/594], loss: 0.0989, accyrasy: 0.9646, f1: 0.9372  f1_dev: 0.8666\n",
      "[8/30] [229/594], loss: 0.1037, accyrasy: 0.9621, f1: 0.929  f1_dev: 0.7878\n",
      "[8/30] [234/594], loss: 0.0994, accyrasy: 0.9621, f1: 0.9309  f1_dev: 0.7039\n",
      "[8/30] [239/594], loss: 0.0959, accyrasy: 0.9655, f1: 0.9335  f1_dev: 0.7101\n",
      "[8/30] [244/594], loss: 0.1022, accyrasy: 0.9581, f1: 0.9238  f1_dev: 0.6606\n",
      "[8/30] [249/594], loss: 0.1096, accyrasy: 0.9664, f1: 0.9372  f1_dev: 0.6758\n",
      "[8/30] [254/594], loss: 0.111, accyrasy: 0.9567, f1: 0.9184  f1_dev: 0.8054\n",
      "[8/30] [259/594], loss: 0.1054, accyrasy: 0.9573, f1: 0.9152  f1_dev: 0.6835\n",
      "[8/30] [264/594], loss: 0.0961, accyrasy: 0.9659, f1: 0.9346  f1_dev: 0.7092\n",
      "[8/30] [269/594], loss: 0.1018, accyrasy: 0.9596, f1: 0.9214  f1_dev: 0.6812\n",
      "[8/30] [274/594], loss: 0.1069, accyrasy: 0.9567, f1: 0.9136  f1_dev: 0.7411\n",
      "[8/30] [279/594], loss: 0.1148, accyrasy: 0.9599, f1: 0.9229  f1_dev: 0.7708\n",
      "[8/30] [284/594], loss: 0.1057, accyrasy: 0.9627, f1: 0.9306  f1_dev: 0.7208\n",
      "[8/30] [289/594], loss: 0.0988, accyrasy: 0.9669, f1: 0.933  f1_dev: 0.7922\n",
      "[8/30] [294/594], loss: 0.1024, accyrasy: 0.9656, f1: 0.9355  f1_dev: 0.8623\n",
      "[8/30] [299/594], loss: 0.112, accyrasy: 0.957, f1: 0.8971  f1_dev: 0.7611\n",
      "[8/30] [304/594], loss: 0.1119, accyrasy: 0.9605, f1: 0.9177  f1_dev: 0.7088\n",
      "[8/30] [309/594], loss: 0.1136, accyrasy: 0.9564, f1: 0.9269  f1_dev: 0.6876\n",
      "[8/30] [314/594], loss: 0.0986, accyrasy: 0.967, f1: 0.9343  f1_dev: 0.6692\n",
      "[8/30] [319/594], loss: 0.0945, accyrasy: 0.9655, f1: 0.9339  f1_dev: 0.6764\n",
      "[8/30] [324/594], loss: 0.0979, accyrasy: 0.9609, f1: 0.9212  f1_dev: 0.7125\n",
      "[8/30] [329/594], loss: 0.095, accyrasy: 0.9646, f1: 0.9292  f1_dev: 0.6928\n",
      "[8/30] [334/594], loss: 0.0928, accyrasy: 0.9649, f1: 0.9294  f1_dev: 0.6903\n",
      "[8/30] [339/594], loss: 0.0945, accyrasy: 0.9651, f1: 0.9388  f1_dev: 0.7078\n",
      "[8/30] [344/594], loss: 0.0919, accyrasy: 0.9609, f1: 0.9135  f1_dev: 0.7045\n",
      "[8/30] [349/594], loss: 0.0936, accyrasy: 0.9664, f1: 0.9376  f1_dev: 0.7093\n",
      "[8/30] [354/594], loss: 0.0979, accyrasy: 0.9609, f1: 0.9207  f1_dev: 0.722\n",
      "[8/30] [359/594], loss: 0.0949, accyrasy: 0.958, f1: 0.921  f1_dev: 0.651\n",
      "[8/30] [364/594], loss: 0.0949, accyrasy: 0.9671, f1: 0.9352  f1_dev: 0.6952\n",
      "[8/30] [369/594], loss: 0.0956, accyrasy: 0.9616, f1: 0.9264  f1_dev: 0.7033\n",
      "[8/30] [374/594], loss: 0.0933, accyrasy: 0.9663, f1: 0.9354  f1_dev: 0.7301\n",
      "[8/30] [379/594], loss: 0.0914, accyrasy: 0.964, f1: 0.929  f1_dev: 0.676\n",
      "[8/30] [384/594], loss: 0.0887, accyrasy: 0.9659, f1: 0.9345  f1_dev: 0.7172\n",
      "[8/30] [389/594], loss: 0.095, accyrasy: 0.9633, f1: 0.931  f1_dev: 0.7743\n",
      "[8/30] [394/594], loss: 0.0903, accyrasy: 0.9647, f1: 0.9342  f1_dev: 0.7429\n",
      "[8/30] [399/594], loss: 0.0951, accyrasy: 0.9677, f1: 0.9383  f1_dev: 0.7565\n",
      "[8/30] [404/594], loss: 0.0953, accyrasy: 0.9615, f1: 0.9215  f1_dev: 0.7722\n",
      "[8/30] [409/594], loss: 0.0928, accyrasy: 0.9647, f1: 0.9369  f1_dev: 0.739\n",
      "[8/30] [414/594], loss: 0.0924, accyrasy: 0.9667, f1: 0.937  f1_dev: 0.716\n",
      "[8/30] [419/594], loss: 0.0901, accyrasy: 0.9677, f1: 0.9347  f1_dev: 0.7266\n",
      "[8/30] [424/594], loss: 0.0937, accyrasy: 0.9625, f1: 0.9357  f1_dev: 0.7113\n",
      "[8/30] [429/594], loss: 0.0912, accyrasy: 0.963, f1: 0.9296  f1_dev: 0.7206\n",
      "[8/30] [434/594], loss: 0.0946, accyrasy: 0.9663, f1: 0.937  f1_dev: 0.6519\n",
      "[8/30] [439/594], loss: 0.097, accyrasy: 0.9655, f1: 0.9303  f1_dev: 0.7283\n",
      "[8/30] [444/594], loss: 0.0895, accyrasy: 0.9689, f1: 0.9383  f1_dev: 0.8557\n",
      "[8/30] [449/594], loss: 0.0949, accyrasy: 0.9615, f1: 0.9334  f1_dev: 0.6739\n",
      "[8/30] [454/594], loss: 0.1003, accyrasy: 0.9624, f1: 0.9236  f1_dev: 0.7001\n",
      "[8/30] [459/594], loss: 0.1032, accyrasy: 0.9621, f1: 0.9291  f1_dev: 0.7335\n",
      "[8/30] [464/594], loss: 0.096, accyrasy: 0.9617, f1: 0.9259  f1_dev: 0.7935\n",
      "[8/30] [469/594], loss: 0.094, accyrasy: 0.9664, f1: 0.9391  f1_dev: 0.785\n",
      "[8/30] [474/594], loss: 0.0985, accyrasy: 0.9599, f1: 0.922  f1_dev: 0.7489\n",
      "[8/30] [479/594], loss: 0.098, accyrasy: 0.9649, f1: 0.9374  f1_dev: 0.7439\n",
      "[8/30] [484/594], loss: 0.1, accyrasy: 0.9633, f1: 0.9297  f1_dev: 0.7465\n",
      "[8/30] [489/594], loss: 0.097, accyrasy: 0.9671, f1: 0.9354  f1_dev: 0.705\n",
      "[8/30] [494/594], loss: 0.0976, accyrasy: 0.9596, f1: 0.9234  f1_dev: 0.6956\n",
      "[8/30] [499/594], loss: 0.1013, accyrasy: 0.9583, f1: 0.9169  f1_dev: 0.745\n",
      "[8/30] [504/594], loss: 0.0994, accyrasy: 0.9623, f1: 0.9281  f1_dev: 0.7741\n",
      "[8/30] [509/594], loss: 0.093, accyrasy: 0.9657, f1: 0.9328  f1_dev: 0.689\n",
      "[8/30] [514/594], loss: 0.0915, accyrasy: 0.9626, f1: 0.9325  f1_dev: 0.7212\n",
      "[8/30] [519/594], loss: 0.0927, accyrasy: 0.9659, f1: 0.939  f1_dev: 0.7147\n",
      "[8/30] [524/594], loss: 0.0865, accyrasy: 0.9627, f1: 0.9285  f1_dev: 0.7183\n",
      "[8/30] [529/594], loss: 0.0957, accyrasy: 0.9656, f1: 0.9358  f1_dev: 0.7334\n",
      "[8/30] [534/594], loss: 0.0912, accyrasy: 0.9683, f1: 0.9421  f1_dev: 0.7118\n",
      "[8/30] [539/594], loss: 0.0937, accyrasy: 0.9659, f1: 0.9352  f1_dev: 0.6888\n",
      "[8/30] [544/594], loss: 0.096, accyrasy: 0.9575, f1: 0.921  f1_dev: 0.7491\n",
      "[8/30] [549/594], loss: 0.0967, accyrasy: 0.9667, f1: 0.9362  f1_dev: 0.7345\n",
      "[8/30] [554/594], loss: 0.0945, accyrasy: 0.962, f1: 0.9204  f1_dev: 0.7051\n",
      "[8/30] [559/594], loss: 0.0962, accyrasy: 0.9623, f1: 0.936  f1_dev: 0.7857\n",
      "[8/30] [564/594], loss: 0.1006, accyrasy: 0.9607, f1: 0.9351  f1_dev: 0.7816\n",
      "[8/30] [569/594], loss: 0.0971, accyrasy: 0.9635, f1: 0.9263  f1_dev: 0.7515\n",
      "[8/30] [574/594], loss: 0.1013, accyrasy: 0.9647, f1: 0.9321  f1_dev: 0.7886\n",
      "[8/30] [579/594], loss: 0.1067, accyrasy: 0.9561, f1: 0.9203  f1_dev: 0.6358\n",
      "[8/30] [584/594], loss: 0.1386, accyrasy: 0.9531, f1: 0.9073  f1_dev: 0.6163\n",
      "[8/30] [589/594], loss: 0.1575, accyrasy: 0.9599, f1: 0.9288  f1_dev: 0.7724\n",
      "[9/30] [4/594], loss: 0.1099, accyrasy: 0.9613, f1: 0.9247  f1_dev: 0.7076\n",
      "[9/30] [9/594], loss: 0.1053, accyrasy: 0.9614, f1: 0.9291  f1_dev: 0.737\n",
      "[9/30] [14/594], loss: 0.1066, accyrasy: 0.9553, f1: 0.9156  f1_dev: 0.6816\n",
      "[9/30] [19/594], loss: 0.1174, accyrasy: 0.9616, f1: 0.9182  f1_dev: 0.6725\n",
      "[9/30] [24/594], loss: 0.1125, accyrasy: 0.9577, f1: 0.9229  f1_dev: 0.6756\n",
      "[9/30] [29/594], loss: 0.1048, accyrasy: 0.9613, f1: 0.9261  f1_dev: 0.6675\n",
      "[9/30] [34/594], loss: 0.1032, accyrasy: 0.9627, f1: 0.9169  f1_dev: 0.6511\n",
      "[9/30] [39/594], loss: 0.0997, accyrasy: 0.9619, f1: 0.9196  f1_dev: 0.6543\n",
      "[9/30] [44/594], loss: 0.1008, accyrasy: 0.9645, f1: 0.9402  f1_dev: 0.6657\n",
      "[9/30] [49/594], loss: 0.0984, accyrasy: 0.9624, f1: 0.9284  f1_dev: 0.6488\n",
      "[9/30] [54/594], loss: 0.0971, accyrasy: 0.9637, f1: 0.9257  f1_dev: 0.6821\n",
      "[9/30] [59/594], loss: 0.1052, accyrasy: 0.9525, f1: 0.9033  f1_dev: 0.6613\n",
      "[9/30] [64/594], loss: 0.0959, accyrasy: 0.9671, f1: 0.9318  f1_dev: 0.6682\n",
      "[9/30] [69/594], loss: 0.096, accyrasy: 0.9603, f1: 0.9279  f1_dev: 0.6881\n",
      "[9/30] [74/594], loss: 0.0947, accyrasy: 0.9643, f1: 0.9356  f1_dev: 0.6676\n",
      "[9/30] [79/594], loss: 0.0992, accyrasy: 0.9585, f1: 0.9227  f1_dev: 0.6827\n",
      "[9/30] [84/594], loss: 0.093, accyrasy: 0.962, f1: 0.9219  f1_dev: 0.7151\n",
      "[9/30] [89/594], loss: 0.0937, accyrasy: 0.9636, f1: 0.9282  f1_dev: 0.6786\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[9/30] [94/594], loss: 0.1013, accyrasy: 0.9626, f1: 0.9333  f1_dev: 0.6878\n",
      "[9/30] [99/594], loss: 0.0991, accyrasy: 0.9633, f1: 0.9321  f1_dev: 0.711\n",
      "[9/30] [104/594], loss: 0.0959, accyrasy: 0.9609, f1: 0.9106  f1_dev: 0.6807\n",
      "[9/30] [109/594], loss: 0.0965, accyrasy: 0.9614, f1: 0.9345  f1_dev: 0.6767\n",
      "[9/30] [114/594], loss: 0.0937, accyrasy: 0.9679, f1: 0.9392  f1_dev: 0.6768\n",
      "[9/30] [119/594], loss: 0.0915, accyrasy: 0.9663, f1: 0.9355  f1_dev: 0.6829\n",
      "[9/30] [124/594], loss: 0.0955, accyrasy: 0.9657, f1: 0.933  f1_dev: 0.7381\n",
      "[9/30] [129/594], loss: 0.1045, accyrasy: 0.9539, f1: 0.9011  f1_dev: 0.7205\n",
      "[9/30] [134/594], loss: 0.1006, accyrasy: 0.9646, f1: 0.9277  f1_dev: 0.6947\n",
      "[9/30] [139/594], loss: 0.0956, accyrasy: 0.963, f1: 0.9198  f1_dev: 0.8262\n",
      "[9/30] [144/594], loss: 0.0897, accyrasy: 0.9715, f1: 0.94  f1_dev: 0.812\n",
      "[9/30] [149/594], loss: 0.0975, accyrasy: 0.9606, f1: 0.9204  f1_dev: 0.7147\n",
      "[9/30] [154/594], loss: 0.0926, accyrasy: 0.9653, f1: 0.9296  f1_dev: 0.7019\n",
      "[9/30] [159/594], loss: 0.0957, accyrasy: 0.9653, f1: 0.9331  f1_dev: 0.6427\n",
      "[9/30] [164/594], loss: 0.0999, accyrasy: 0.9655, f1: 0.9267  f1_dev: 0.676\n",
      "[9/30] [169/594], loss: 0.0969, accyrasy: 0.9624, f1: 0.9224  f1_dev: 0.6798\n",
      "[9/30] [174/594], loss: 0.1082, accyrasy: 0.9641, f1: 0.9318  f1_dev: 0.7322\n",
      "[9/30] [179/594], loss: 0.1074, accyrasy: 0.9643, f1: 0.9331  f1_dev: 0.6705\n",
      "[9/30] [184/594], loss: 0.102, accyrasy: 0.9624, f1: 0.923  f1_dev: 0.7265\n",
      "[9/30] [189/594], loss: 0.1009, accyrasy: 0.9646, f1: 0.9272  f1_dev: 0.7621\n",
      "[9/30] [194/594], loss: 0.0996, accyrasy: 0.9613, f1: 0.9302  f1_dev: 0.7379\n",
      "[9/30] [199/594], loss: 0.0931, accyrasy: 0.9639, f1: 0.9273  f1_dev: 0.723\n",
      "[9/30] [204/594], loss: 0.0942, accyrasy: 0.9614, f1: 0.9131  f1_dev: 0.6969\n",
      "[9/30] [209/594], loss: 0.0957, accyrasy: 0.9609, f1: 0.918  f1_dev: 0.7345\n",
      "[9/30] [214/594], loss: 0.0965, accyrasy: 0.9664, f1: 0.9294  f1_dev: 0.8703\n",
      "[9/30] [219/594], loss: 0.0984, accyrasy: 0.9631, f1: 0.9314  f1_dev: 0.8054\n",
      "[9/30] [224/594], loss: 0.1033, accyrasy: 0.9601, f1: 0.91  f1_dev: 0.7388\n",
      "[9/30] [229/594], loss: 0.0994, accyrasy: 0.9645, f1: 0.9383  f1_dev: 0.7204\n",
      "[9/30] [234/594], loss: 0.0978, accyrasy: 0.9623, f1: 0.9206  f1_dev: 0.8292\n",
      "[9/30] [239/594], loss: 0.0948, accyrasy: 0.9646, f1: 0.9314  f1_dev: 0.8373\n",
      "[9/30] [244/594], loss: 0.0976, accyrasy: 0.9641, f1: 0.9323  f1_dev: 0.7223\n",
      "[9/30] [249/594], loss: 0.0923, accyrasy: 0.9689, f1: 0.9404  f1_dev: 0.69\n",
      "[9/30] [254/594], loss: 0.1003, accyrasy: 0.9603, f1: 0.9241  f1_dev: 0.8003\n",
      "[9/30] [259/594], loss: 0.0977, accyrasy: 0.9615, f1: 0.9284  f1_dev: 0.7053\n",
      "[9/30] [264/594], loss: 0.0966, accyrasy: 0.9626, f1: 0.9251  f1_dev: 0.6896\n",
      "[9/30] [269/594], loss: 0.1035, accyrasy: 0.9644, f1: 0.9334  f1_dev: 0.6792\n",
      "[9/30] [274/594], loss: 0.0989, accyrasy: 0.9646, f1: 0.9351  f1_dev: 0.7262\n",
      "[9/30] [279/594], loss: 0.0924, accyrasy: 0.964, f1: 0.9305  f1_dev: 0.775\n",
      "[9/30] [284/594], loss: 0.0928, accyrasy: 0.9651, f1: 0.9351  f1_dev: 0.7253\n",
      "[9/30] [289/594], loss: 0.0954, accyrasy: 0.9653, f1: 0.9351  f1_dev: 0.7368\n",
      "[9/30] [294/594], loss: 0.1, accyrasy: 0.9616, f1: 0.934  f1_dev: 0.7216\n",
      "[9/30] [299/594], loss: 0.0989, accyrasy: 0.9649, f1: 0.9267  f1_dev: 0.8174\n",
      "[9/30] [304/594], loss: 0.1012, accyrasy: 0.9639, f1: 0.9355  f1_dev: 0.7243\n",
      "[9/30] [309/594], loss: 0.1003, accyrasy: 0.9589, f1: 0.9149  f1_dev: 0.6884\n",
      "[9/30] [314/594], loss: 0.1017, accyrasy: 0.956, f1: 0.9016  f1_dev: 0.7174\n",
      "[9/30] [319/594], loss: 0.094, accyrasy: 0.9671, f1: 0.9409  f1_dev: 0.7155\n",
      "[9/30] [324/594], loss: 0.0925, accyrasy: 0.9654, f1: 0.9345  f1_dev: 0.753\n",
      "[9/30] [329/594], loss: 0.0905, accyrasy: 0.9636, f1: 0.9304  f1_dev: 0.7356\n",
      "[9/30] [334/594], loss: 0.0976, accyrasy: 0.968, f1: 0.9391  f1_dev: 0.7241\n",
      "[9/30] [339/594], loss: 0.091, accyrasy: 0.9664, f1: 0.9306  f1_dev: 0.843\n",
      "[9/30] [344/594], loss: 0.0917, accyrasy: 0.9676, f1: 0.942  f1_dev: 0.7379\n",
      "[9/30] [349/594], loss: 0.0915, accyrasy: 0.9666, f1: 0.9389  f1_dev: 0.7369\n",
      "[9/30] [354/594], loss: 0.0937, accyrasy: 0.9636, f1: 0.9358  f1_dev: 0.6851\n",
      "[9/30] [359/594], loss: 0.0941, accyrasy: 0.9639, f1: 0.9315  f1_dev: 0.8367\n",
      "[9/30] [364/594], loss: 0.0945, accyrasy: 0.9617, f1: 0.927  f1_dev: 0.7565\n",
      "[9/30] [369/594], loss: 0.0925, accyrasy: 0.9644, f1: 0.9355  f1_dev: 0.6884\n",
      "[9/30] [374/594], loss: 0.0981, accyrasy: 0.9603, f1: 0.9236  f1_dev: 0.8562\n",
      "[9/30] [379/594], loss: 0.0974, accyrasy: 0.9651, f1: 0.9336  f1_dev: 0.7682\n",
      "[9/30] [384/594], loss: 0.094, accyrasy: 0.9663, f1: 0.9355  f1_dev: 0.7757\n",
      "[9/30] [389/594], loss: 0.0893, accyrasy: 0.9624, f1: 0.9287  f1_dev: 0.7314\n",
      "[9/30] [394/594], loss: 0.0911, accyrasy: 0.9644, f1: 0.9326  f1_dev: 0.7127\n",
      "[9/30] [399/594], loss: 0.0975, accyrasy: 0.964, f1: 0.9276  f1_dev: 0.8167\n",
      "[9/30] [404/594], loss: 0.0956, accyrasy: 0.9631, f1: 0.9278  f1_dev: 0.6887\n",
      "[9/30] [409/594], loss: 0.0945, accyrasy: 0.9639, f1: 0.9323  f1_dev: 0.7049\n",
      "[9/30] [414/594], loss: 0.0965, accyrasy: 0.962, f1: 0.9209  f1_dev: 0.8494\n",
      "[9/30] [419/594], loss: 0.0939, accyrasy: 0.9663, f1: 0.9388  f1_dev: 0.7184\n",
      "[9/30] [424/594], loss: 0.0928, accyrasy: 0.9624, f1: 0.9273  f1_dev: 0.7483\n",
      "[9/30] [429/594], loss: 0.09, accyrasy: 0.9683, f1: 0.9395  f1_dev: 0.68\n",
      "[9/30] [434/594], loss: 0.096, accyrasy: 0.9597, f1: 0.9177  f1_dev: 0.8669\n",
      "[9/30] [439/594], loss: 0.096, accyrasy: 0.9643, f1: 0.9298  f1_dev: 0.8674\n",
      "[9/30] [444/594], loss: 0.0943, accyrasy: 0.9656, f1: 0.9377  f1_dev: 0.7464\n",
      "[9/30] [449/594], loss: 0.0946, accyrasy: 0.9637, f1: 0.9322  f1_dev: 0.7073\n",
      "[9/30] [454/594], loss: 0.0937, accyrasy: 0.9634, f1: 0.9267  f1_dev: 0.8192\n",
      "[9/30] [459/594], loss: 0.0904, accyrasy: 0.9654, f1: 0.9357  f1_dev: 0.841\n",
      "[9/30] [464/594], loss: 0.0929, accyrasy: 0.9657, f1: 0.9378  f1_dev: 0.7785\n",
      "[9/30] [469/594], loss: 0.0944, accyrasy: 0.9616, f1: 0.9148  f1_dev: 0.7384\n",
      "[9/30] [474/594], loss: 0.0937, accyrasy: 0.9655, f1: 0.9245  f1_dev: 0.7376\n",
      "[9/30] [479/594], loss: 0.0958, accyrasy: 0.9597, f1: 0.9164  f1_dev: 0.7093\n",
      "[9/30] [484/594], loss: 0.0931, accyrasy: 0.964, f1: 0.9309  f1_dev: 0.7416\n",
      "[9/30] [489/594], loss: 0.1016, accyrasy: 0.9606, f1: 0.9209  f1_dev: 0.8398\n",
      "[9/30] [494/594], loss: 0.0921, accyrasy: 0.9681, f1: 0.945  f1_dev: 0.7294\n",
      "[9/30] [499/594], loss: 0.0888, accyrasy: 0.9637, f1: 0.9336  f1_dev: 0.8468\n",
      "[9/30] [504/594], loss: 0.0953, accyrasy: 0.962, f1: 0.9321  f1_dev: 0.8319\n",
      "[9/30] [509/594], loss: 0.0977, accyrasy: 0.9635, f1: 0.9328  f1_dev: 0.705\n",
      "[9/30] [514/594], loss: 0.0956, accyrasy: 0.9633, f1: 0.9341  f1_dev: 0.7548\n",
      "[9/30] [519/594], loss: 0.0983, accyrasy: 0.9616, f1: 0.9184  f1_dev: 0.7576\n",
      "[9/30] [524/594], loss: 0.1023, accyrasy: 0.9596, f1: 0.9188  f1_dev: 0.8687\n",
      "[9/30] [529/594], loss: 0.0998, accyrasy: 0.9575, f1: 0.9266  f1_dev: 0.7189\n",
      "[9/30] [534/594], loss: 0.1049, accyrasy: 0.9616, f1: 0.9278  f1_dev: 0.7445\n",
      "[9/30] [539/594], loss: 0.1006, accyrasy: 0.9673, f1: 0.9289  f1_dev: 0.6779\n",
      "[9/30] [544/594], loss: 0.0947, accyrasy: 0.9651, f1: 0.9274  f1_dev: 0.749\n",
      "[9/30] [549/594], loss: 0.0938, accyrasy: 0.9641, f1: 0.9322  f1_dev: 0.7486\n",
      "[9/30] [554/594], loss: 0.1007, accyrasy: 0.9566, f1: 0.9186  f1_dev: 0.7925\n",
      "[9/30] [559/594], loss: 0.1108, accyrasy: 0.9616, f1: 0.9188  f1_dev: 0.6599\n",
      "[9/30] [564/594], loss: 0.1081, accyrasy: 0.9653, f1: 0.9299  f1_dev: 0.6811\n",
      "[9/30] [569/594], loss: 0.1015, accyrasy: 0.9601, f1: 0.9171  f1_dev: 0.7112\n",
      "[9/30] [574/594], loss: 0.1018, accyrasy: 0.9637, f1: 0.9307  f1_dev: 0.6241\n",
      "[9/30] [579/594], loss: 0.1049, accyrasy: 0.9559, f1: 0.9016  f1_dev: 0.6682\n",
      "[9/30] [584/594], loss: 0.1037, accyrasy: 0.9667, f1: 0.9344  f1_dev: 0.8193\n",
      "[9/30] [589/594], loss: 0.0985, accyrasy: 0.9631, f1: 0.9317  f1_dev: 0.7829\n",
      "[10/30] [4/594], loss: 0.0926, accyrasy: 0.9653, f1: 0.9371  f1_dev: 0.7275\n",
      "[10/30] [9/594], loss: 0.0929, accyrasy: 0.9667, f1: 0.9355  f1_dev: 0.7201\n",
      "[10/30] [14/594], loss: 0.0905, accyrasy: 0.9689, f1: 0.9357  f1_dev: 0.773\n",
      "[10/30] [19/594], loss: 0.0877, accyrasy: 0.9634, f1: 0.9211  f1_dev: 0.763\n",
      "[10/30] [24/594], loss: 0.0942, accyrasy: 0.9635, f1: 0.93  f1_dev: 0.7247\n",
      "[10/30] [29/594], loss: 0.0915, accyrasy: 0.9675, f1: 0.9376  f1_dev: 0.6988\n",
      "[10/30] [34/594], loss: 0.09, accyrasy: 0.9659, f1: 0.9363  f1_dev: 0.721\n",
      "[10/30] [39/594], loss: 0.0923, accyrasy: 0.965, f1: 0.9286  f1_dev: 0.7375\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[10/30] [44/594], loss: 0.0977, accyrasy: 0.9584, f1: 0.9252  f1_dev: 0.7506\n",
      "[10/30] [49/594], loss: 0.0928, accyrasy: 0.9664, f1: 0.9368  f1_dev: 0.72\n",
      "[10/30] [54/594], loss: 0.0936, accyrasy: 0.964, f1: 0.9396  f1_dev: 0.7326\n",
      "[10/30] [59/594], loss: 0.0893, accyrasy: 0.9689, f1: 0.9361  f1_dev: 0.7392\n",
      "[10/30] [64/594], loss: 0.0903, accyrasy: 0.9634, f1: 0.9254  f1_dev: 0.7408\n",
      "[10/30] [69/594], loss: 0.0968, accyrasy: 0.9683, f1: 0.9359  f1_dev: 0.7423\n",
      "[10/30] [74/594], loss: 0.0908, accyrasy: 0.9644, f1: 0.9342  f1_dev: 0.7408\n",
      "[10/30] [79/594], loss: 0.0926, accyrasy: 0.9656, f1: 0.9368  f1_dev: 0.7053\n",
      "[10/30] [84/594], loss: 0.0964, accyrasy: 0.9661, f1: 0.9344  f1_dev: 0.7507\n",
      "[10/30] [89/594], loss: 0.0969, accyrasy: 0.9635, f1: 0.9288  f1_dev: 0.7028\n",
      "[10/30] [94/594], loss: 0.0957, accyrasy: 0.9647, f1: 0.9327  f1_dev: 0.7276\n",
      "[10/30] [99/594], loss: 0.0963, accyrasy: 0.9616, f1: 0.9301  f1_dev: 0.8338\n",
      "[10/30] [104/594], loss: 0.1082, accyrasy: 0.966, f1: 0.9325  f1_dev: 0.7805\n",
      "[10/30] [109/594], loss: 0.094, accyrasy: 0.9659, f1: 0.93  f1_dev: 0.731\n",
      "[10/30] [114/594], loss: 0.0918, accyrasy: 0.9656, f1: 0.9307  f1_dev: 0.6804\n",
      "[10/30] [119/594], loss: 0.092, accyrasy: 0.9637, f1: 0.941  f1_dev: 0.6939\n",
      "[10/30] [124/594], loss: 0.0957, accyrasy: 0.9639, f1: 0.9266  f1_dev: 0.7094\n",
      "[10/30] [129/594], loss: 0.0927, accyrasy: 0.9646, f1: 0.9316  f1_dev: 0.8499\n",
      "[10/30] [134/594], loss: 0.0955, accyrasy: 0.9641, f1: 0.9292  f1_dev: 0.8368\n",
      "[10/30] [139/594], loss: 0.0985, accyrasy: 0.9639, f1: 0.9332  f1_dev: 0.7449\n",
      "[10/30] [144/594], loss: 0.1017, accyrasy: 0.9579, f1: 0.9196  f1_dev: 0.8674\n",
      "[10/30] [149/594], loss: 0.0923, accyrasy: 0.9684, f1: 0.9322  f1_dev: 0.7459\n",
      "[10/30] [154/594], loss: 0.1059, accyrasy: 0.9493, f1: 0.895  f1_dev: 0.6968\n",
      "[10/30] [159/594], loss: 0.1019, accyrasy: 0.9601, f1: 0.9267  f1_dev: 0.7208\n",
      "[10/30] [164/594], loss: 0.0984, accyrasy: 0.9651, f1: 0.9223  f1_dev: 0.7594\n",
      "[10/30] [169/594], loss: 0.0952, accyrasy: 0.9631, f1: 0.9348  f1_dev: 0.7409\n",
      "[10/30] [174/594], loss: 0.0898, accyrasy: 0.9684, f1: 0.9371  f1_dev: 0.7825\n",
      "[10/30] [179/594], loss: 0.0961, accyrasy: 0.9673, f1: 0.9361  f1_dev: 0.8412\n",
      "[10/30] [184/594], loss: 0.0979, accyrasy: 0.963, f1: 0.9295  f1_dev: 0.8699\n",
      "[10/30] [189/594], loss: 0.099, accyrasy: 0.9605, f1: 0.9254  f1_dev: 0.7371\n",
      "[10/30] [194/594], loss: 0.0963, accyrasy: 0.9646, f1: 0.9329  f1_dev: 0.7412\n",
      "[10/30] [199/594], loss: 0.0959, accyrasy: 0.9587, f1: 0.9176  f1_dev: 0.7418\n",
      "[10/30] [204/594], loss: 0.0942, accyrasy: 0.9591, f1: 0.9212  f1_dev: 0.7106\n",
      "[10/30] [209/594], loss: 0.1018, accyrasy: 0.9557, f1: 0.9154  f1_dev: 0.7453\n",
      "[10/30] [214/594], loss: 0.097, accyrasy: 0.9655, f1: 0.9343  f1_dev: 0.8503\n",
      "[10/30] [219/594], loss: 0.098, accyrasy: 0.9621, f1: 0.9283  f1_dev: 0.6969\n",
      "[10/30] [224/594], loss: 0.0914, accyrasy: 0.9673, f1: 0.9282  f1_dev: 0.7117\n",
      "[10/30] [229/594], loss: 0.0957, accyrasy: 0.9591, f1: 0.9177  f1_dev: 0.7377\n",
      "[10/30] [234/594], loss: 0.0912, accyrasy: 0.9654, f1: 0.9302  f1_dev: 0.7586\n",
      "[10/30] [239/594], loss: 0.0875, accyrasy: 0.9664, f1: 0.9366  f1_dev: 0.763\n",
      "[10/30] [244/594], loss: 0.0877, accyrasy: 0.9617, f1: 0.924  f1_dev: 0.7898\n",
      "[10/30] [249/594], loss: 0.0849, accyrasy: 0.9663, f1: 0.934  f1_dev: 0.8596\n",
      "[10/30] [254/594], loss: 0.0905, accyrasy: 0.9667, f1: 0.9422  f1_dev: 0.8685\n",
      "[10/30] [259/594], loss: 0.0885, accyrasy: 0.9654, f1: 0.9408  f1_dev: 0.7941\n",
      "[10/30] [264/594], loss: 0.0931, accyrasy: 0.9606, f1: 0.933  f1_dev: 0.7285\n",
      "[10/30] [269/594], loss: 0.0957, accyrasy: 0.9643, f1: 0.9334  f1_dev: 0.7327\n",
      "[10/30] [274/594], loss: 0.0997, accyrasy: 0.9546, f1: 0.9148  f1_dev: 0.7354\n",
      "[10/30] [279/594], loss: 0.0959, accyrasy: 0.96, f1: 0.9288  f1_dev: 0.8503\n",
      "[10/30] [284/594], loss: 0.0935, accyrasy: 0.9601, f1: 0.9136  f1_dev: 0.8264\n",
      "[10/30] [289/594], loss: 0.0956, accyrasy: 0.969, f1: 0.9408  f1_dev: 0.6552\n",
      "[10/30] [294/594], loss: 0.1042, accyrasy: 0.9645, f1: 0.9284  f1_dev: 0.7297\n",
      "[10/30] [299/594], loss: 0.1063, accyrasy: 0.9637, f1: 0.9357  f1_dev: 0.7708\n",
      "[10/30] [304/594], loss: 0.1037, accyrasy: 0.9625, f1: 0.9252  f1_dev: 0.7673\n",
      "[10/30] [309/594], loss: 0.0985, accyrasy: 0.9669, f1: 0.9372  f1_dev: 0.7504\n",
      "[10/30] [314/594], loss: 0.0968, accyrasy: 0.9653, f1: 0.9352  f1_dev: 0.7258\n",
      "[10/30] [319/594], loss: 0.0964, accyrasy: 0.9644, f1: 0.931  f1_dev: 0.8126\n",
      "[10/30] [324/594], loss: 0.0934, accyrasy: 0.9639, f1: 0.9328  f1_dev: 0.8132\n",
      "[10/30] [329/594], loss: 0.0911, accyrasy: 0.9665, f1: 0.9304  f1_dev: 0.7037\n",
      "[10/30] [334/594], loss: 0.0987, accyrasy: 0.964, f1: 0.9299  f1_dev: 0.7199\n",
      "[10/30] [339/594], loss: 0.0912, accyrasy: 0.9659, f1: 0.9302  f1_dev: 0.7645\n",
      "[10/30] [344/594], loss: 0.0903, accyrasy: 0.9677, f1: 0.9375  f1_dev: 0.826\n",
      "[10/30] [349/594], loss: 0.098, accyrasy: 0.9655, f1: 0.9349  f1_dev: 0.7486\n",
      "[10/30] [354/594], loss: 0.0955, accyrasy: 0.9655, f1: 0.9297  f1_dev: 0.8601\n",
      "[10/30] [359/594], loss: 0.1003, accyrasy: 0.9654, f1: 0.9343  f1_dev: 0.6887\n",
      "[10/30] [364/594], loss: 0.0962, accyrasy: 0.9631, f1: 0.9176  f1_dev: 0.6644\n",
      "[10/30] [369/594], loss: 0.0964, accyrasy: 0.9633, f1: 0.929  f1_dev: 0.8057\n",
      "[10/30] [374/594], loss: 0.0899, accyrasy: 0.9673, f1: 0.937  f1_dev: 0.8573\n",
      "[10/30] [379/594], loss: 0.0929, accyrasy: 0.9665, f1: 0.9312  f1_dev: 0.8309\n",
      "[10/30] [384/594], loss: 0.0913, accyrasy: 0.9671, f1: 0.9389  f1_dev: 0.8302\n",
      "[10/30] [389/594], loss: 0.0935, accyrasy: 0.9657, f1: 0.9341  f1_dev: 0.759\n",
      "[10/30] [394/594], loss: 0.0929, accyrasy: 0.9637, f1: 0.9304  f1_dev: 0.7511\n",
      "[10/30] [399/594], loss: 0.0903, accyrasy: 0.9653, f1: 0.9333  f1_dev: 0.7413\n",
      "[10/30] [404/594], loss: 0.0948, accyrasy: 0.9635, f1: 0.9269  f1_dev: 0.7538\n",
      "[10/30] [409/594], loss: 0.0892, accyrasy: 0.9661, f1: 0.9337  f1_dev: 0.7668\n",
      "[10/30] [414/594], loss: 0.0934, accyrasy: 0.9675, f1: 0.9323  f1_dev: 0.7926\n",
      "[10/30] [419/594], loss: 0.0917, accyrasy: 0.966, f1: 0.9368  f1_dev: 0.8691\n",
      "[10/30] [424/594], loss: 0.0876, accyrasy: 0.967, f1: 0.9358  f1_dev: 0.7417\n",
      "[10/30] [429/594], loss: 0.0905, accyrasy: 0.9615, f1: 0.9255  f1_dev: 0.7437\n",
      "[10/30] [434/594], loss: 0.1003, accyrasy: 0.9605, f1: 0.9239  f1_dev: 0.7443\n",
      "[10/30] [439/594], loss: 0.0905, accyrasy: 0.9657, f1: 0.9341  f1_dev: 0.8563\n",
      "[10/30] [444/594], loss: 0.0956, accyrasy: 0.9607, f1: 0.9239  f1_dev: 0.736\n",
      "[10/30] [449/594], loss: 0.0987, accyrasy: 0.9673, f1: 0.9307  f1_dev: 0.7459\n",
      "[10/30] [454/594], loss: 0.0911, accyrasy: 0.9698, f1: 0.9367  f1_dev: 0.6861\n",
      "[10/30] [459/594], loss: 0.1055, accyrasy: 0.9559, f1: 0.9169  f1_dev: 0.7218\n",
      "[10/30] [464/594], loss: 0.1101, accyrasy: 0.9596, f1: 0.9256  f1_dev: 0.7758\n",
      "[10/30] [469/594], loss: 0.1031, accyrasy: 0.9656, f1: 0.9346  f1_dev: 0.8681\n",
      "[10/30] [474/594], loss: 0.1002, accyrasy: 0.9639, f1: 0.9295  f1_dev: 0.8631\n",
      "[10/30] [479/594], loss: 0.0945, accyrasy: 0.9651, f1: 0.9344  f1_dev: 0.7441\n",
      "[10/30] [484/594], loss: 0.0977, accyrasy: 0.9624, f1: 0.9274  f1_dev: 0.8436\n",
      "[10/30] [489/594], loss: 0.0983, accyrasy: 0.9577, f1: 0.9261  f1_dev: 0.7354\n",
      "[10/30] [494/594], loss: 0.0957, accyrasy: 0.9666, f1: 0.9341  f1_dev: 0.7433\n",
      "[10/30] [499/594], loss: 0.0999, accyrasy: 0.9664, f1: 0.9329  f1_dev: 0.7545\n",
      "[10/30] [504/594], loss: 0.0929, accyrasy: 0.9623, f1: 0.9293  f1_dev: 0.8484\n",
      "[10/30] [509/594], loss: 0.0955, accyrasy: 0.965, f1: 0.935  f1_dev: 0.8232\n",
      "[10/30] [514/594], loss: 0.0926, accyrasy: 0.9646, f1: 0.9364  f1_dev: 0.8305\n",
      "[10/30] [519/594], loss: 0.0874, accyrasy: 0.968, f1: 0.9388  f1_dev: 0.7606\n",
      "[10/30] [524/594], loss: 0.0885, accyrasy: 0.9709, f1: 0.9448  f1_dev: 0.7547\n",
      "[10/30] [529/594], loss: 0.09, accyrasy: 0.965, f1: 0.9306  f1_dev: 0.7759\n",
      "[10/30] [534/594], loss: 0.0893, accyrasy: 0.965, f1: 0.9323  f1_dev: 0.851\n",
      "[10/30] [539/594], loss: 0.0887, accyrasy: 0.965, f1: 0.9249  f1_dev: 0.8175\n",
      "[10/30] [544/594], loss: 0.091, accyrasy: 0.9645, f1: 0.9305  f1_dev: 0.7161\n",
      "[10/30] [549/594], loss: 0.0913, accyrasy: 0.9666, f1: 0.9335  f1_dev: 0.7468\n",
      "[10/30] [554/594], loss: 0.0878, accyrasy: 0.9676, f1: 0.9346  f1_dev: 0.7701\n",
      "[10/30] [559/594], loss: 0.0899, accyrasy: 0.9685, f1: 0.9334  f1_dev: 0.7378\n",
      "[10/30] [564/594], loss: 0.0904, accyrasy: 0.9651, f1: 0.9286  f1_dev: 0.6976\n",
      "[10/30] [569/594], loss: 0.1018, accyrasy: 0.9631, f1: 0.9287  f1_dev: 0.7448\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[10/30] [574/594], loss: 0.0929, accyrasy: 0.965, f1: 0.9295  f1_dev: 0.7113\n",
      "[10/30] [579/594], loss: 0.0987, accyrasy: 0.9615, f1: 0.9189  f1_dev: 0.8077\n",
      "[10/30] [584/594], loss: 0.0941, accyrasy: 0.9657, f1: 0.9289  f1_dev: 0.7799\n",
      "[10/30] [589/594], loss: 0.09, accyrasy: 0.964, f1: 0.9301  f1_dev: 0.7869\n",
      "[11/30] [4/594], loss: 0.0882, accyrasy: 0.9675, f1: 0.9374  f1_dev: 0.7755\n",
      "[11/30] [9/594], loss: 0.0942, accyrasy: 0.9595, f1: 0.9228  f1_dev: 0.7694\n",
      "[11/30] [14/594], loss: 0.0872, accyrasy: 0.9666, f1: 0.9361  f1_dev: 0.7651\n",
      "[11/30] [19/594], loss: 0.0959, accyrasy: 0.9601, f1: 0.9205  f1_dev: 0.7234\n",
      "[11/30] [24/594], loss: 0.0979, accyrasy: 0.9665, f1: 0.9311  f1_dev: 0.7498\n",
      "[11/30] [29/594], loss: 0.0902, accyrasy: 0.9626, f1: 0.9238  f1_dev: 0.7062\n",
      "[11/30] [34/594], loss: 0.0995, accyrasy: 0.9615, f1: 0.9265  f1_dev: 0.8688\n",
      "[11/30] [39/594], loss: 0.0981, accyrasy: 0.967, f1: 0.9365  f1_dev: 0.8458\n",
      "[11/30] [44/594], loss: 0.0977, accyrasy: 0.966, f1: 0.9303  f1_dev: 0.7504\n",
      "[11/30] [49/594], loss: 0.0949, accyrasy: 0.9651, f1: 0.9358  f1_dev: 0.7137\n",
      "[11/30] [54/594], loss: 0.0887, accyrasy: 0.9674, f1: 0.937  f1_dev: 0.7331\n",
      "[11/30] [59/594], loss: 0.0937, accyrasy: 0.9669, f1: 0.938  f1_dev: 0.741\n",
      "[11/30] [64/594], loss: 0.0963, accyrasy: 0.9646, f1: 0.9311  f1_dev: 0.7309\n",
      "[11/30] [69/594], loss: 0.0913, accyrasy: 0.9625, f1: 0.9268  f1_dev: 0.7435\n",
      "[11/30] [74/594], loss: 0.0945, accyrasy: 0.9653, f1: 0.9351  f1_dev: 0.7337\n",
      "[11/30] [79/594], loss: 0.0902, accyrasy: 0.962, f1: 0.9228  f1_dev: 0.7153\n",
      "[11/30] [84/594], loss: 0.0946, accyrasy: 0.9633, f1: 0.9293  f1_dev: 0.6855\n",
      "[11/30] [89/594], loss: 0.1065, accyrasy: 0.9609, f1: 0.9272  f1_dev: 0.758\n",
      "[11/30] [94/594], loss: 0.0965, accyrasy: 0.9619, f1: 0.9119  f1_dev: 0.7358\n",
      "[11/30] [99/594], loss: 0.094, accyrasy: 0.9635, f1: 0.9276  f1_dev: 0.7787\n",
      "[11/30] [104/594], loss: 0.0969, accyrasy: 0.9614, f1: 0.9225  f1_dev: 0.7789\n",
      "[11/30] [109/594], loss: 0.0925, accyrasy: 0.9631, f1: 0.9294  f1_dev: 0.7604\n",
      "[11/30] [114/594], loss: 0.0915, accyrasy: 0.9661, f1: 0.9311  f1_dev: 0.7267\n",
      "[11/30] [119/594], loss: 0.0959, accyrasy: 0.9634, f1: 0.9286  f1_dev: 0.8477\n",
      "[11/30] [124/594], loss: 0.0954, accyrasy: 0.9677, f1: 0.9383  f1_dev: 0.8477\n",
      "[11/30] [129/594], loss: 0.0964, accyrasy: 0.9716, f1: 0.9412  f1_dev: 0.7111\n",
      "[11/30] [134/594], loss: 0.0894, accyrasy: 0.9684, f1: 0.9355  f1_dev: 0.7532\n",
      "[11/30] [139/594], loss: 0.0893, accyrasy: 0.9625, f1: 0.9325  f1_dev: 0.7813\n",
      "[11/30] [144/594], loss: 0.0943, accyrasy: 0.9655, f1: 0.927  f1_dev: 0.7642\n",
      "[11/30] [149/594], loss: 0.0936, accyrasy: 0.9621, f1: 0.9282  f1_dev: 0.7761\n",
      "[11/30] [154/594], loss: 0.0946, accyrasy: 0.9636, f1: 0.9227  f1_dev: 0.731\n",
      "[11/30] [159/594], loss: 0.0978, accyrasy: 0.9616, f1: 0.9173  f1_dev: 0.7414\n",
      "[11/30] [164/594], loss: 0.1013, accyrasy: 0.9576, f1: 0.9233  f1_dev: 0.6802\n",
      "[11/30] [169/594], loss: 0.1015, accyrasy: 0.9627, f1: 0.9243  f1_dev: 0.6748\n",
      "[11/30] [174/594], loss: 0.1114, accyrasy: 0.9563, f1: 0.9036  f1_dev: 0.7337\n",
      "[11/30] [179/594], loss: 0.1059, accyrasy: 0.9644, f1: 0.9337  f1_dev: 0.7007\n",
      "[11/30] [184/594], loss: 0.1018, accyrasy: 0.965, f1: 0.9373  f1_dev: 0.7429\n",
      "[11/30] [189/594], loss: 0.106, accyrasy: 0.9583, f1: 0.9241  f1_dev: 0.871\n",
      "[11/30] [194/594], loss: 0.0989, accyrasy: 0.9675, f1: 0.9379  f1_dev: 0.7482\n",
      "[11/30] [199/594], loss: 0.0938, accyrasy: 0.9635, f1: 0.931  f1_dev: 0.7441\n",
      "[11/30] [204/594], loss: 0.0926, accyrasy: 0.9686, f1: 0.9412  f1_dev: 0.871\n",
      "[11/30] [209/594], loss: 0.0957, accyrasy: 0.9684, f1: 0.9389  f1_dev: 0.8559\n",
      "[11/30] [214/594], loss: 0.0936, accyrasy: 0.9649, f1: 0.9332  f1_dev: 0.8551\n",
      "[11/30] [219/594], loss: 0.098, accyrasy: 0.9655, f1: 0.9348  f1_dev: 0.8442\n",
      "[11/30] [224/594], loss: 0.0962, accyrasy: 0.9633, f1: 0.9293  f1_dev: 0.8385\n",
      "[11/30] [229/594], loss: 0.0966, accyrasy: 0.9673, f1: 0.9368  f1_dev: 0.7289\n",
      "[11/30] [234/594], loss: 0.0915, accyrasy: 0.9657, f1: 0.9338  f1_dev: 0.6898\n",
      "[11/30] [239/594], loss: 0.0964, accyrasy: 0.9649, f1: 0.9369  f1_dev: 0.8504\n",
      "[11/30] [244/594], loss: 0.0893, accyrasy: 0.9689, f1: 0.9409  f1_dev: 0.836\n",
      "[11/30] [249/594], loss: 0.0889, accyrasy: 0.9643, f1: 0.9291  f1_dev: 0.8675\n",
      "[11/30] [254/594], loss: 0.0929, accyrasy: 0.9625, f1: 0.9272  f1_dev: 0.7847\n",
      "[11/30] [259/594], loss: 0.0949, accyrasy: 0.9634, f1: 0.9265  f1_dev: 0.7788\n",
      "[11/30] [264/594], loss: 0.0928, accyrasy: 0.9649, f1: 0.9326  f1_dev: 0.7812\n",
      "[11/30] [269/594], loss: 0.0937, accyrasy: 0.9663, f1: 0.9379  f1_dev: 0.7718\n",
      "[11/30] [274/594], loss: 0.0895, accyrasy: 0.9683, f1: 0.9395  f1_dev: 0.8611\n",
      "[11/30] [279/594], loss: 0.0927, accyrasy: 0.9617, f1: 0.9246  f1_dev: 0.7376\n",
      "[11/30] [284/594], loss: 0.0906, accyrasy: 0.9684, f1: 0.9361  f1_dev: 0.737\n",
      "[11/30] [289/594], loss: 0.093, accyrasy: 0.9675, f1: 0.9434  f1_dev: 0.7354\n",
      "[11/30] [294/594], loss: 0.0879, accyrasy: 0.9683, f1: 0.937  f1_dev: 0.7438\n",
      "[11/30] [299/594], loss: 0.0899, accyrasy: 0.9665, f1: 0.9349  f1_dev: 0.8309\n",
      "[11/30] [304/594], loss: 0.092, accyrasy: 0.9647, f1: 0.9297  f1_dev: 0.8531\n",
      "[11/30] [309/594], loss: 0.0933, accyrasy: 0.965, f1: 0.9282  f1_dev: 0.8528\n",
      "[11/30] [314/594], loss: 0.0978, accyrasy: 0.9631, f1: 0.9264  f1_dev: 0.691\n",
      "[11/30] [319/594], loss: 0.0903, accyrasy: 0.9696, f1: 0.9394  f1_dev: 0.7402\n",
      "[11/30] [324/594], loss: 0.1044, accyrasy: 0.9576, f1: 0.9183  f1_dev: 0.8449\n",
      "[11/30] [329/594], loss: 0.1025, accyrasy: 0.9663, f1: 0.9349  f1_dev: 0.835\n",
      "[11/30] [334/594], loss: 0.0934, accyrasy: 0.9614, f1: 0.9283  f1_dev: 0.7293\n",
      "[11/30] [339/594], loss: 0.0928, accyrasy: 0.9676, f1: 0.9316  f1_dev: 0.7238\n",
      "[11/30] [344/594], loss: 0.0917, accyrasy: 0.963, f1: 0.9274  f1_dev: 0.7063\n",
      "[11/30] [349/594], loss: 0.0939, accyrasy: 0.9647, f1: 0.9366  f1_dev: 0.8062\n",
      "[11/30] [354/594], loss: 0.0961, accyrasy: 0.9604, f1: 0.9184  f1_dev: 0.799\n",
      "[11/30] [359/594], loss: 0.0914, accyrasy: 0.9616, f1: 0.9221  f1_dev: 0.7262\n",
      "[11/30] [364/594], loss: 0.0927, accyrasy: 0.9659, f1: 0.9316  f1_dev: 0.7507\n",
      "[11/30] [369/594], loss: 0.091, accyrasy: 0.9605, f1: 0.9218  f1_dev: 0.7529\n",
      "[11/30] [374/594], loss: 0.093, accyrasy: 0.9621, f1: 0.9263  f1_dev: 0.7645\n",
      "[11/30] [379/594], loss: 0.0898, accyrasy: 0.966, f1: 0.938  f1_dev: 0.8636\n",
      "[11/30] [384/594], loss: 0.0922, accyrasy: 0.9654, f1: 0.9403  f1_dev: 0.8546\n",
      "[11/30] [389/594], loss: 0.0909, accyrasy: 0.9689, f1: 0.9406  f1_dev: 0.7064\n",
      "[11/30] [394/594], loss: 0.0916, accyrasy: 0.9646, f1: 0.9359  f1_dev: 0.7208\n",
      "[11/30] [399/594], loss: 0.0904, accyrasy: 0.9643, f1: 0.9335  f1_dev: 0.7952\n",
      "[11/30] [404/594], loss: 0.0897, accyrasy: 0.9675, f1: 0.9398  f1_dev: 0.7811\n",
      "[11/30] [409/594], loss: 0.0908, accyrasy: 0.9673, f1: 0.9391  f1_dev: 0.8324\n",
      "[11/30] [414/594], loss: 0.0989, accyrasy: 0.9654, f1: 0.9364  f1_dev: 0.7286\n",
      "[11/30] [419/594], loss: 0.0918, accyrasy: 0.9666, f1: 0.9371  f1_dev: 0.657\n",
      "[11/30] [424/594], loss: 0.0954, accyrasy: 0.9635, f1: 0.9323  f1_dev: 0.7369\n",
      "[11/30] [429/594], loss: 0.0947, accyrasy: 0.9604, f1: 0.9172  f1_dev: 0.8527\n",
      "[11/30] [434/594], loss: 0.0983, accyrasy: 0.9645, f1: 0.9323  f1_dev: 0.7033\n",
      "[11/30] [439/594], loss: 0.103, accyrasy: 0.96, f1: 0.9247  f1_dev: 0.7448\n",
      "[11/30] [444/594], loss: 0.1027, accyrasy: 0.9594, f1: 0.9207  f1_dev: 0.8519\n",
      "[11/30] [449/594], loss: 0.0898, accyrasy: 0.9675, f1: 0.9416  f1_dev: 0.7504\n",
      "[11/30] [454/594], loss: 0.0918, accyrasy: 0.9637, f1: 0.9235  f1_dev: 0.7403\n",
      "[11/30] [459/594], loss: 0.0895, accyrasy: 0.9671, f1: 0.9363  f1_dev: 0.7616\n",
      "[11/30] [464/594], loss: 0.0886, accyrasy: 0.9676, f1: 0.9389  f1_dev: 0.7913\n",
      "[11/30] [469/594], loss: 0.0907, accyrasy: 0.9616, f1: 0.9231  f1_dev: 0.7395\n",
      "[11/30] [474/594], loss: 0.0926, accyrasy: 0.9655, f1: 0.9315  f1_dev: 0.859\n",
      "[11/30] [479/594], loss: 0.0972, accyrasy: 0.9609, f1: 0.9174  f1_dev: 0.8729\n",
      "[11/30] [484/594], loss: 0.0897, accyrasy: 0.9649, f1: 0.9346  f1_dev: 0.8676\n",
      "[11/30] [489/594], loss: 0.0924, accyrasy: 0.9643, f1: 0.9305  f1_dev: 0.7797\n",
      "[11/30] [494/594], loss: 0.0901, accyrasy: 0.9609, f1: 0.9262  f1_dev: 0.7421\n",
      "[11/30] [499/594], loss: 0.0865, accyrasy: 0.9666, f1: 0.9366  f1_dev: 0.8633\n",
      "[11/30] [504/594], loss: 0.0916, accyrasy: 0.9669, f1: 0.9364  f1_dev: 0.8611\n",
      "[11/30] [509/594], loss: 0.0929, accyrasy: 0.9643, f1: 0.9249  f1_dev: 0.8434\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[11/30] [514/594], loss: 0.0916, accyrasy: 0.9644, f1: 0.9343  f1_dev: 0.7382\n",
      "[11/30] [519/594], loss: 0.0905, accyrasy: 0.9657, f1: 0.9356  f1_dev: 0.783\n",
      "[11/30] [524/594], loss: 0.0943, accyrasy: 0.9544, f1: 0.9048  f1_dev: 0.7466\n",
      "[11/30] [529/594], loss: 0.0919, accyrasy: 0.9634, f1: 0.935  f1_dev: 0.7537\n",
      "[11/30] [534/594], loss: 0.0909, accyrasy: 0.9655, f1: 0.9352  f1_dev: 0.7821\n",
      "[11/30] [539/594], loss: 0.0945, accyrasy: 0.9634, f1: 0.9309  f1_dev: 0.7155\n",
      "[11/30] [544/594], loss: 0.0952, accyrasy: 0.9624, f1: 0.9267  f1_dev: 0.7613\n",
      "[11/30] [549/594], loss: 0.1032, accyrasy: 0.9599, f1: 0.9228  f1_dev: 0.7647\n",
      "[11/30] [554/594], loss: 0.0994, accyrasy: 0.9625, f1: 0.9228  f1_dev: 0.8722\n",
      "[11/30] [559/594], loss: 0.0881, accyrasy: 0.9676, f1: 0.9392  f1_dev: 0.8705\n",
      "[11/30] [564/594], loss: 0.0873, accyrasy: 0.9696, f1: 0.9384  f1_dev: 0.8501\n",
      "[11/30] [569/594], loss: 0.091, accyrasy: 0.9639, f1: 0.9263  f1_dev: 0.8471\n",
      "[11/30] [574/594], loss: 0.0946, accyrasy: 0.9617, f1: 0.9267  f1_dev: 0.8129\n",
      "[11/30] [579/594], loss: 0.0947, accyrasy: 0.9655, f1: 0.9319  f1_dev: 0.7676\n",
      "[11/30] [584/594], loss: 0.103, accyrasy: 0.9621, f1: 0.9217  f1_dev: 0.7248\n",
      "[11/30] [589/594], loss: 0.0934, accyrasy: 0.9641, f1: 0.9257  f1_dev: 0.7229\n",
      "[12/30] [4/594], loss: 0.0932, accyrasy: 0.9667, f1: 0.9358  f1_dev: 0.8649\n",
      "[12/30] [9/594], loss: 0.0969, accyrasy: 0.9603, f1: 0.912  f1_dev: 0.7645\n",
      "[12/30] [14/594], loss: 0.0898, accyrasy: 0.9683, f1: 0.9368  f1_dev: 0.7568\n",
      "[12/30] [19/594], loss: 0.0946, accyrasy: 0.9606, f1: 0.9196  f1_dev: 0.7591\n",
      "[12/30] [24/594], loss: 0.0934, accyrasy: 0.9674, f1: 0.9282  f1_dev: 0.694\n",
      "[12/30] [29/594], loss: 0.0945, accyrasy: 0.9633, f1: 0.9275  f1_dev: 0.7416\n",
      "[12/30] [34/594], loss: 0.0958, accyrasy: 0.9617, f1: 0.9223  f1_dev: 0.8529\n",
      "[12/30] [39/594], loss: 0.0945, accyrasy: 0.9636, f1: 0.9268  f1_dev: 0.7532\n",
      "[12/30] [44/594], loss: 0.0936, accyrasy: 0.9663, f1: 0.9312  f1_dev: 0.7627\n",
      "[12/30] [49/594], loss: 0.0872, accyrasy: 0.9663, f1: 0.9356  f1_dev: 0.8053\n",
      "[12/30] [54/594], loss: 0.0908, accyrasy: 0.9647, f1: 0.9351  f1_dev: 0.8172\n",
      "[12/30] [59/594], loss: 0.0934, accyrasy: 0.9615, f1: 0.9189  f1_dev: 0.7595\n",
      "[12/30] [64/594], loss: 0.0929, accyrasy: 0.9625, f1: 0.9321  f1_dev: 0.8067\n",
      "[12/30] [69/594], loss: 0.0987, accyrasy: 0.9605, f1: 0.931  f1_dev: 0.7667\n",
      "[12/30] [74/594], loss: 0.0989, accyrasy: 0.966, f1: 0.931  f1_dev: 0.7351\n",
      "[12/30] [79/594], loss: 0.0989, accyrasy: 0.9669, f1: 0.9255  f1_dev: 0.6751\n",
      "[12/30] [84/594], loss: 0.097, accyrasy: 0.9635, f1: 0.9254  f1_dev: 0.731\n",
      "[12/30] [89/594], loss: 0.0972, accyrasy: 0.9663, f1: 0.9303  f1_dev: 0.7129\n",
      "[12/30] [94/594], loss: 0.0988, accyrasy: 0.9621, f1: 0.9315  f1_dev: 0.7644\n",
      "[12/30] [99/594], loss: 0.0931, accyrasy: 0.965, f1: 0.9341  f1_dev: 0.7438\n",
      "[12/30] [104/594], loss: 0.0899, accyrasy: 0.9653, f1: 0.9366  f1_dev: 0.7462\n",
      "[12/30] [109/594], loss: 0.0855, accyrasy: 0.9679, f1: 0.9397  f1_dev: 0.8412\n",
      "[12/30] [114/594], loss: 0.0898, accyrasy: 0.9654, f1: 0.9295  f1_dev: 0.7382\n",
      "[12/30] [119/594], loss: 0.0945, accyrasy: 0.9647, f1: 0.9352  f1_dev: 0.7269\n",
      "[12/30] [124/594], loss: 0.092, accyrasy: 0.9607, f1: 0.9262  f1_dev: 0.7212\n",
      "[12/30] [129/594], loss: 0.0935, accyrasy: 0.9624, f1: 0.9229  f1_dev: 0.847\n",
      "[12/30] [134/594], loss: 0.0915, accyrasy: 0.9647, f1: 0.9301  f1_dev: 0.8631\n",
      "[12/30] [139/594], loss: 0.0904, accyrasy: 0.9684, f1: 0.9426  f1_dev: 0.8277\n",
      "[12/30] [144/594], loss: 0.0868, accyrasy: 0.965, f1: 0.9321  f1_dev: 0.8116\n",
      "[12/30] [149/594], loss: 0.0895, accyrasy: 0.9683, f1: 0.9364  f1_dev: 0.8488\n",
      "[12/30] [154/594], loss: 0.0884, accyrasy: 0.9639, f1: 0.9225  f1_dev: 0.8434\n",
      "[12/30] [159/594], loss: 0.0889, accyrasy: 0.968, f1: 0.9375  f1_dev: 0.8052\n",
      "[12/30] [164/594], loss: 0.087, accyrasy: 0.9633, f1: 0.9277  f1_dev: 0.748\n",
      "[12/30] [169/594], loss: 0.0904, accyrasy: 0.9667, f1: 0.9345  f1_dev: 0.7606\n",
      "[12/30] [174/594], loss: 0.093, accyrasy: 0.9617, f1: 0.9297  f1_dev: 0.7808\n",
      "[12/30] [179/594], loss: 0.0889, accyrasy: 0.9671, f1: 0.9436  f1_dev: 0.7384\n",
      "[12/30] [184/594], loss: 0.0904, accyrasy: 0.9673, f1: 0.9361  f1_dev: 0.7507\n",
      "[12/30] [189/594], loss: 0.094, accyrasy: 0.9607, f1: 0.919  f1_dev: 0.7109\n",
      "[12/30] [194/594], loss: 0.093, accyrasy: 0.965, f1: 0.9353  f1_dev: 0.743\n",
      "[12/30] [199/594], loss: 0.0927, accyrasy: 0.9657, f1: 0.9294  f1_dev: 0.8089\n",
      "[12/30] [204/594], loss: 0.095, accyrasy: 0.9631, f1: 0.927  f1_dev: 0.7466\n",
      "[12/30] [209/594], loss: 0.0945, accyrasy: 0.9671, f1: 0.9339  f1_dev: 0.7681\n",
      "[12/30] [214/594], loss: 0.0879, accyrasy: 0.9669, f1: 0.932  f1_dev: 0.7793\n",
      "[12/30] [219/594], loss: 0.0887, accyrasy: 0.9663, f1: 0.9338  f1_dev: 0.7477\n",
      "[12/30] [224/594], loss: 0.0946, accyrasy: 0.9576, f1: 0.9101  f1_dev: 0.6923\n",
      "[12/30] [229/594], loss: 0.0919, accyrasy: 0.971, f1: 0.9431  f1_dev: 0.732\n",
      "[12/30] [234/594], loss: 0.0925, accyrasy: 0.9634, f1: 0.931  f1_dev: 0.6898\n",
      "[12/30] [239/594], loss: 0.0903, accyrasy: 0.9664, f1: 0.931  f1_dev: 0.771\n",
      "[12/30] [244/594], loss: 0.0941, accyrasy: 0.9649, f1: 0.9333  f1_dev: 0.8408\n",
      "[12/30] [249/594], loss: 0.0934, accyrasy: 0.9641, f1: 0.9365  f1_dev: 0.805\n",
      "[12/30] [254/594], loss: 0.0923, accyrasy: 0.9671, f1: 0.9321  f1_dev: 0.8345\n",
      "[12/30] [259/594], loss: 0.0878, accyrasy: 0.969, f1: 0.942  f1_dev: 0.7645\n",
      "[12/30] [264/594], loss: 0.0932, accyrasy: 0.9617, f1: 0.9186  f1_dev: 0.8743\n",
      "[12/30] [269/594], loss: 0.094, accyrasy: 0.965, f1: 0.9279  f1_dev: 0.868\n",
      "[12/30] [274/594], loss: 0.0885, accyrasy: 0.9653, f1: 0.9363  f1_dev: 0.853\n",
      "[12/30] [279/594], loss: 0.0897, accyrasy: 0.9635, f1: 0.9197  f1_dev: 0.7798\n",
      "[12/30] [284/594], loss: 0.0865, accyrasy: 0.969, f1: 0.9398  f1_dev: 0.7769\n",
      "[12/30] [289/594], loss: 0.0854, accyrasy: 0.9711, f1: 0.9477  f1_dev: 0.7671\n",
      "[12/30] [294/594], loss: 0.0865, accyrasy: 0.9674, f1: 0.9289  f1_dev: 0.6983\n",
      "[12/30] [299/594], loss: 0.0874, accyrasy: 0.9684, f1: 0.939  f1_dev: 0.7874\n",
      "[12/30] [304/594], loss: 0.087, accyrasy: 0.9673, f1: 0.9316  f1_dev: 0.7992\n",
      "[12/30] [309/594], loss: 0.0887, accyrasy: 0.9646, f1: 0.9274  f1_dev: 0.7814\n",
      "[12/30] [314/594], loss: 0.0859, accyrasy: 0.9663, f1: 0.9337  f1_dev: 0.7948\n",
      "[12/30] [319/594], loss: 0.0904, accyrasy: 0.9641, f1: 0.9285  f1_dev: 0.7947\n",
      "[12/30] [324/594], loss: 0.0894, accyrasy: 0.9653, f1: 0.9348  f1_dev: 0.7388\n",
      "[12/30] [329/594], loss: 0.087, accyrasy: 0.9663, f1: 0.9277  f1_dev: 0.7388\n",
      "[12/30] [334/594], loss: 0.0907, accyrasy: 0.9653, f1: 0.9288  f1_dev: 0.748\n",
      "[12/30] [339/594], loss: 0.0915, accyrasy: 0.9624, f1: 0.932  f1_dev: 0.8539\n",
      "[12/30] [344/594], loss: 0.0917, accyrasy: 0.965, f1: 0.9321  f1_dev: 0.7976\n",
      "[12/30] [349/594], loss: 0.0905, accyrasy: 0.9673, f1: 0.9357  f1_dev: 0.769\n",
      "[12/30] [354/594], loss: 0.0933, accyrasy: 0.9624, f1: 0.9224  f1_dev: 0.8467\n",
      "[12/30] [359/594], loss: 0.094, accyrasy: 0.9634, f1: 0.9339  f1_dev: 0.7863\n",
      "[12/30] [364/594], loss: 0.0952, accyrasy: 0.9666, f1: 0.938  f1_dev: 0.7623\n",
      "[12/30] [369/594], loss: 0.0881, accyrasy: 0.9677, f1: 0.935  f1_dev: 0.712\n",
      "[12/30] [374/594], loss: 0.087, accyrasy: 0.9637, f1: 0.9283  f1_dev: 0.7417\n",
      "[12/30] [379/594], loss: 0.0892, accyrasy: 0.9627, f1: 0.9295  f1_dev: 0.719\n",
      "[12/30] [384/594], loss: 0.0914, accyrasy: 0.9674, f1: 0.9351  f1_dev: 0.7618\n",
      "[12/30] [389/594], loss: 0.0885, accyrasy: 0.9656, f1: 0.9379  f1_dev: 0.7899\n",
      "[12/30] [394/594], loss: 0.0902, accyrasy: 0.9621, f1: 0.9332  f1_dev: 0.785\n",
      "[12/30] [399/594], loss: 0.0908, accyrasy: 0.9671, f1: 0.9391  f1_dev: 0.798\n",
      "[12/30] [404/594], loss: 0.0911, accyrasy: 0.9654, f1: 0.9304  f1_dev: 0.8065\n",
      "[12/30] [409/594], loss: 0.0958, accyrasy: 0.9676, f1: 0.9385  f1_dev: 0.8678\n",
      "[12/30] [414/594], loss: 0.0965, accyrasy: 0.9617, f1: 0.9309  f1_dev: 0.8856\n",
      "[12/30] [419/594], loss: 0.1057, accyrasy: 0.9556, f1: 0.9197  f1_dev: 0.6734\n",
      "[12/30] [424/594], loss: 0.1126, accyrasy: 0.961, f1: 0.925  f1_dev: 0.7806\n",
      "[12/30] [429/594], loss: 0.1106, accyrasy: 0.9549, f1: 0.9147  f1_dev: 0.6932\n",
      "[12/30] [434/594], loss: 0.1013, accyrasy: 0.9633, f1: 0.9324  f1_dev: 0.7685\n",
      "[12/30] [439/594], loss: 0.0958, accyrasy: 0.9656, f1: 0.9304  f1_dev: 0.7502\n",
      "[12/30] [444/594], loss: 0.0927, accyrasy: 0.9659, f1: 0.9348  f1_dev: 0.811\n",
      "[12/30] [449/594], loss: 0.0955, accyrasy: 0.9646, f1: 0.928  f1_dev: 0.8128\n",
      "[12/30] [454/594], loss: 0.0906, accyrasy: 0.9653, f1: 0.9321  f1_dev: 0.7497\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[12/30] [459/594], loss: 0.0868, accyrasy: 0.968, f1: 0.9345  f1_dev: 0.8596\n",
      "[12/30] [464/594], loss: 0.0829, accyrasy: 0.9688, f1: 0.9404  f1_dev: 0.8698\n",
      "[12/30] [469/594], loss: 0.0923, accyrasy: 0.9657, f1: 0.9354  f1_dev: 0.8677\n",
      "[12/30] [474/594], loss: 0.0925, accyrasy: 0.9659, f1: 0.9335  f1_dev: 0.8761\n",
      "[12/30] [479/594], loss: 0.0931, accyrasy: 0.9613, f1: 0.9276  f1_dev: 0.7643\n",
      "[12/30] [484/594], loss: 0.0908, accyrasy: 0.96, f1: 0.9303  f1_dev: 0.738\n",
      "[12/30] [489/594], loss: 0.0916, accyrasy: 0.9665, f1: 0.9281  f1_dev: 0.7301\n",
      "[12/30] [494/594], loss: 0.0907, accyrasy: 0.9669, f1: 0.9346  f1_dev: 0.7397\n",
      "[12/30] [499/594], loss: 0.0912, accyrasy: 0.9699, f1: 0.9349  f1_dev: 0.8659\n",
      "[12/30] [504/594], loss: 0.0913, accyrasy: 0.9651, f1: 0.9296  f1_dev: 0.8367\n",
      "[12/30] [509/594], loss: 0.0921, accyrasy: 0.9675, f1: 0.9433  f1_dev: 0.7912\n",
      "[12/30] [514/594], loss: 0.0926, accyrasy: 0.9645, f1: 0.9313  f1_dev: 0.7651\n",
      "[12/30] [519/594], loss: 0.0957, accyrasy: 0.9647, f1: 0.9338  f1_dev: 0.7616\n",
      "[12/30] [524/594], loss: 0.0902, accyrasy: 0.9666, f1: 0.9324  f1_dev: 0.8618\n",
      "[12/30] [529/594], loss: 0.0951, accyrasy: 0.9655, f1: 0.9363  f1_dev: 0.8663\n",
      "[12/30] [534/594], loss: 0.0921, accyrasy: 0.9696, f1: 0.9377  f1_dev: 0.8566\n",
      "[12/30] [539/594], loss: 0.0886, accyrasy: 0.9669, f1: 0.9368  f1_dev: 0.7706\n",
      "[12/30] [544/594], loss: 0.0917, accyrasy: 0.9676, f1: 0.9402  f1_dev: 0.8418\n",
      "[12/30] [549/594], loss: 0.0884, accyrasy: 0.9667, f1: 0.9371  f1_dev: 0.8279\n",
      "[12/30] [554/594], loss: 0.0915, accyrasy: 0.9643, f1: 0.9251  f1_dev: 0.7479\n",
      "[12/30] [559/594], loss: 0.0931, accyrasy: 0.9611, f1: 0.9291  f1_dev: 0.7499\n",
      "[12/30] [564/594], loss: 0.0966, accyrasy: 0.9613, f1: 0.9312  f1_dev: 0.7228\n",
      "[12/30] [569/594], loss: 0.0921, accyrasy: 0.969, f1: 0.9398  f1_dev: 0.8052\n",
      "[12/30] [574/594], loss: 0.0978, accyrasy: 0.9656, f1: 0.936  f1_dev: 0.7502\n",
      "[12/30] [579/594], loss: 0.0874, accyrasy: 0.9691, f1: 0.9343  f1_dev: 0.7932\n",
      "[12/30] [584/594], loss: 0.0904, accyrasy: 0.9621, f1: 0.9247  f1_dev: 0.7884\n",
      "[12/30] [589/594], loss: 0.0866, accyrasy: 0.9685, f1: 0.9347  f1_dev: 0.7848\n",
      "[13/30] [4/594], loss: 0.0929, accyrasy: 0.9575, f1: 0.9196  f1_dev: 0.8082\n",
      "[13/30] [9/594], loss: 0.0991, accyrasy: 0.9653, f1: 0.9319  f1_dev: 0.7346\n",
      "[13/30] [14/594], loss: 0.0952, accyrasy: 0.9655, f1: 0.9344  f1_dev: 0.6917\n",
      "[13/30] [19/594], loss: 0.0902, accyrasy: 0.9677, f1: 0.9335  f1_dev: 0.718\n",
      "[13/30] [24/594], loss: 0.0871, accyrasy: 0.9686, f1: 0.9302  f1_dev: 0.7705\n",
      "[13/30] [29/594], loss: 0.0911, accyrasy: 0.9685, f1: 0.9432  f1_dev: 0.7344\n",
      "[13/30] [34/594], loss: 0.0892, accyrasy: 0.9653, f1: 0.931  f1_dev: 0.8442\n",
      "[13/30] [39/594], loss: 0.0936, accyrasy: 0.9631, f1: 0.9315  f1_dev: 0.7078\n",
      "[13/30] [44/594], loss: 0.0961, accyrasy: 0.9579, f1: 0.9229  f1_dev: 0.7798\n",
      "[13/30] [49/594], loss: 0.0928, accyrasy: 0.9684, f1: 0.9427  f1_dev: 0.817\n",
      "[13/30] [54/594], loss: 0.1127, accyrasy: 0.9637, f1: 0.9263  f1_dev: 0.5378\n",
      "[13/30] [59/594], loss: 0.1354, accyrasy: 0.9373, f1: 0.8598  f1_dev: 0.6561\n",
      "[13/30] [64/594], loss: 0.1195, accyrasy: 0.9603, f1: 0.9103  f1_dev: 0.7797\n",
      "[13/30] [69/594], loss: 0.1094, accyrasy: 0.9626, f1: 0.9307  f1_dev: 0.7626\n",
      "[13/30] [74/594], loss: 0.1015, accyrasy: 0.9627, f1: 0.9305  f1_dev: 0.7513\n",
      "[13/30] [79/594], loss: 0.0923, accyrasy: 0.9646, f1: 0.9288  f1_dev: 0.7134\n",
      "[13/30] [84/594], loss: 0.0984, accyrasy: 0.9685, f1: 0.939  f1_dev: 0.7068\n",
      "[13/30] [89/594], loss: 0.0966, accyrasy: 0.9616, f1: 0.9261  f1_dev: 0.8174\n",
      "[13/30] [94/594], loss: 0.0943, accyrasy: 0.9597, f1: 0.9149  f1_dev: 0.7287\n",
      "[13/30] [99/594], loss: 0.0884, accyrasy: 0.9679, f1: 0.941  f1_dev: 0.7022\n",
      "[13/30] [104/594], loss: 0.0941, accyrasy: 0.9645, f1: 0.9298  f1_dev: 0.7431\n",
      "[13/30] [109/594], loss: 0.0959, accyrasy: 0.9574, f1: 0.9118  f1_dev: 0.7441\n",
      "[13/30] [114/594], loss: 0.0938, accyrasy: 0.9685, f1: 0.9361  f1_dev: 0.7512\n",
      "[13/30] [119/594], loss: 0.0947, accyrasy: 0.9643, f1: 0.931  f1_dev: 0.7674\n",
      "[13/30] [124/594], loss: 0.0917, accyrasy: 0.967, f1: 0.9348  f1_dev: 0.7729\n",
      "[13/30] [129/594], loss: 0.0906, accyrasy: 0.9595, f1: 0.9209  f1_dev: 0.7842\n",
      "[13/30] [134/594], loss: 0.0889, accyrasy: 0.9701, f1: 0.9456  f1_dev: 0.7785\n",
      "[13/30] [139/594], loss: 0.0869, accyrasy: 0.9646, f1: 0.9235  f1_dev: 0.7779\n",
      "[13/30] [144/594], loss: 0.0855, accyrasy: 0.9686, f1: 0.9401  f1_dev: 0.8265\n",
      "[13/30] [149/594], loss: 0.0872, accyrasy: 0.9686, f1: 0.9411  f1_dev: 0.752\n",
      "[13/30] [154/594], loss: 0.0903, accyrasy: 0.9673, f1: 0.9406  f1_dev: 0.8372\n",
      "[13/30] [159/594], loss: 0.0893, accyrasy: 0.9694, f1: 0.9429  f1_dev: 0.7311\n",
      "[13/30] [164/594], loss: 0.0924, accyrasy: 0.9669, f1: 0.9382  f1_dev: 0.863\n",
      "[13/30] [169/594], loss: 0.0931, accyrasy: 0.9617, f1: 0.9249  f1_dev: 0.7195\n",
      "[13/30] [174/594], loss: 0.0945, accyrasy: 0.9653, f1: 0.9374  f1_dev: 0.8178\n",
      "[13/30] [179/594], loss: 0.0922, accyrasy: 0.9609, f1: 0.9244  f1_dev: 0.7903\n",
      "[13/30] [184/594], loss: 0.0942, accyrasy: 0.9661, f1: 0.9417  f1_dev: 0.8466\n",
      "[13/30] [189/594], loss: 0.0888, accyrasy: 0.9688, f1: 0.9367  f1_dev: 0.7779\n",
      "[13/30] [194/594], loss: 0.0879, accyrasy: 0.9596, f1: 0.9264  f1_dev: 0.7849\n",
      "[13/30] [199/594], loss: 0.0971, accyrasy: 0.9535, f1: 0.9075  f1_dev: 0.787\n",
      "[13/30] [204/594], loss: 0.0884, accyrasy: 0.9665, f1: 0.9368  f1_dev: 0.7757\n",
      "[13/30] [209/594], loss: 0.0896, accyrasy: 0.9696, f1: 0.945  f1_dev: 0.7406\n",
      "[13/30] [214/594], loss: 0.0849, accyrasy: 0.9673, f1: 0.9362  f1_dev: 0.7556\n",
      "[13/30] [219/594], loss: 0.087, accyrasy: 0.9629, f1: 0.9307  f1_dev: 0.8523\n",
      "[13/30] [224/594], loss: 0.0914, accyrasy: 0.9634, f1: 0.9326  f1_dev: 0.7674\n",
      "[13/30] [229/594], loss: 0.0901, accyrasy: 0.9666, f1: 0.9355  f1_dev: 0.717\n",
      "[13/30] [234/594], loss: 0.0948, accyrasy: 0.9633, f1: 0.9267  f1_dev: 0.7652\n",
      "[13/30] [239/594], loss: 0.0896, accyrasy: 0.9688, f1: 0.9384  f1_dev: 0.7555\n",
      "[13/30] [244/594], loss: 0.0915, accyrasy: 0.9666, f1: 0.9397  f1_dev: 0.7586\n",
      "[13/30] [249/594], loss: 0.0947, accyrasy: 0.9694, f1: 0.9294  f1_dev: 0.8137\n",
      "[13/30] [254/594], loss: 0.1038, accyrasy: 0.9586, f1: 0.9104  f1_dev: 0.8454\n",
      "[13/30] [259/594], loss: 0.0909, accyrasy: 0.9644, f1: 0.9245  f1_dev: 0.7374\n",
      "[13/30] [264/594], loss: 0.0921, accyrasy: 0.968, f1: 0.9352  f1_dev: 0.8384\n",
      "[13/30] [269/594], loss: 0.0852, accyrasy: 0.9676, f1: 0.9365  f1_dev: 0.838\n",
      "[13/30] [274/594], loss: 0.0901, accyrasy: 0.9689, f1: 0.9412  f1_dev: 0.84\n",
      "[13/30] [279/594], loss: 0.095, accyrasy: 0.9624, f1: 0.9366  f1_dev: 0.8664\n",
      "[13/30] [284/594], loss: 0.0923, accyrasy: 0.9651, f1: 0.9273  f1_dev: 0.8687\n",
      "[13/30] [289/594], loss: 0.0865, accyrasy: 0.9681, f1: 0.9407  f1_dev: 0.7692\n",
      "[13/30] [294/594], loss: 0.0874, accyrasy: 0.9639, f1: 0.9292  f1_dev: 0.8516\n",
      "[13/30] [299/594], loss: 0.0858, accyrasy: 0.9699, f1: 0.9421  f1_dev: 0.8633\n",
      "[13/30] [304/594], loss: 0.0899, accyrasy: 0.9611, f1: 0.9254  f1_dev: 0.7464\n",
      "[13/30] [309/594], loss: 0.0888, accyrasy: 0.9661, f1: 0.9341  f1_dev: 0.7569\n",
      "[13/30] [314/594], loss: 0.0904, accyrasy: 0.9659, f1: 0.928  f1_dev: 0.778\n",
      "[13/30] [319/594], loss: 0.0889, accyrasy: 0.9671, f1: 0.9335  f1_dev: 0.7751\n",
      "[13/30] [324/594], loss: 0.0923, accyrasy: 0.9605, f1: 0.9212  f1_dev: 0.8774\n",
      "[13/30] [329/594], loss: 0.0869, accyrasy: 0.9691, f1: 0.9423  f1_dev: 0.8783\n",
      "[13/30] [334/594], loss: 0.0873, accyrasy: 0.9695, f1: 0.9449  f1_dev: 0.8784\n",
      "[13/30] [339/594], loss: 0.0954, accyrasy: 0.9593, f1: 0.9239  f1_dev: 0.8743\n",
      "[13/30] [344/594], loss: 0.0897, accyrasy: 0.9636, f1: 0.9308  f1_dev: 0.8588\n",
      "[13/30] [349/594], loss: 0.0887, accyrasy: 0.9654, f1: 0.9306  f1_dev: 0.8595\n",
      "[13/30] [354/594], loss: 0.0904, accyrasy: 0.9677, f1: 0.9387  f1_dev: 0.8582\n",
      "[13/30] [359/594], loss: 0.0894, accyrasy: 0.9679, f1: 0.9399  f1_dev: 0.8756\n",
      "[13/30] [364/594], loss: 0.0916, accyrasy: 0.9691, f1: 0.9364  f1_dev: 0.7436\n",
      "[13/30] [369/594], loss: 0.0889, accyrasy: 0.9647, f1: 0.9282  f1_dev: 0.7741\n",
      "[13/30] [374/594], loss: 0.0911, accyrasy: 0.9676, f1: 0.9362  f1_dev: 0.8711\n",
      "[13/30] [379/594], loss: 0.0915, accyrasy: 0.9702, f1: 0.9442  f1_dev: 0.8528\n",
      "[13/30] [384/594], loss: 0.0939, accyrasy: 0.9624, f1: 0.9351  f1_dev: 0.8647\n",
      "[13/30] [389/594], loss: 0.0923, accyrasy: 0.9684, f1: 0.9416  f1_dev: 0.8675\n",
      "[13/30] [394/594], loss: 0.0909, accyrasy: 0.9657, f1: 0.9342  f1_dev: 0.8625\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[13/30] [399/594], loss: 0.0871, accyrasy: 0.965, f1: 0.9366  f1_dev: 0.8446\n",
      "[13/30] [404/594], loss: 0.0913, accyrasy: 0.964, f1: 0.9247  f1_dev: 0.7591\n",
      "[13/30] [409/594], loss: 0.0936, accyrasy: 0.9637, f1: 0.9276  f1_dev: 0.8672\n",
      "[13/30] [414/594], loss: 0.0915, accyrasy: 0.9646, f1: 0.9313  f1_dev: 0.8622\n",
      "[13/30] [419/594], loss: 0.0896, accyrasy: 0.9663, f1: 0.9319  f1_dev: 0.8724\n",
      "[13/30] [424/594], loss: 0.0952, accyrasy: 0.9621, f1: 0.922  f1_dev: 0.8699\n",
      "[13/30] [429/594], loss: 0.0969, accyrasy: 0.9646, f1: 0.932  f1_dev: 0.8754\n",
      "[13/30] [434/594], loss: 0.0962, accyrasy: 0.9667, f1: 0.929  f1_dev: 0.8199\n",
      "[13/30] [439/594], loss: 0.0921, accyrasy: 0.9639, f1: 0.9271  f1_dev: 0.7134\n",
      "[13/30] [444/594], loss: 0.0945, accyrasy: 0.9649, f1: 0.9337  f1_dev: 0.8357\n",
      "[13/30] [449/594], loss: 0.089, accyrasy: 0.9639, f1: 0.9257  f1_dev: 0.8664\n",
      "[13/30] [454/594], loss: 0.0877, accyrasy: 0.9646, f1: 0.9394  f1_dev: 0.8632\n",
      "[13/30] [459/594], loss: 0.0897, accyrasy: 0.9692, f1: 0.937  f1_dev: 0.7373\n",
      "[13/30] [464/594], loss: 0.0896, accyrasy: 0.9654, f1: 0.9268  f1_dev: 0.8646\n",
      "[13/30] [469/594], loss: 0.0902, accyrasy: 0.9667, f1: 0.9328  f1_dev: 0.8774\n",
      "[13/30] [474/594], loss: 0.0942, accyrasy: 0.9651, f1: 0.938  f1_dev: 0.8534\n",
      "[13/30] [479/594], loss: 0.0959, accyrasy: 0.9657, f1: 0.9388  f1_dev: 0.8462\n",
      "[13/30] [484/594], loss: 0.0964, accyrasy: 0.9606, f1: 0.9173  f1_dev: 0.828\n",
      "[13/30] [489/594], loss: 0.1044, accyrasy: 0.9649, f1: 0.9301  f1_dev: 0.8732\n",
      "[13/30] [494/594], loss: 0.1058, accyrasy: 0.9636, f1: 0.929  f1_dev: 0.864\n",
      "[13/30] [499/594], loss: 0.0992, accyrasy: 0.962, f1: 0.9264  f1_dev: 0.7671\n",
      "[13/30] [504/594], loss: 0.0942, accyrasy: 0.9664, f1: 0.933  f1_dev: 0.8602\n",
      "[13/30] [509/594], loss: 0.0921, accyrasy: 0.966, f1: 0.9288  f1_dev: 0.7744\n",
      "[13/30] [514/594], loss: 0.0913, accyrasy: 0.9626, f1: 0.9211  f1_dev: 0.8665\n",
      "[13/30] [519/594], loss: 0.0893, accyrasy: 0.9651, f1: 0.9263  f1_dev: 0.7564\n",
      "[13/30] [524/594], loss: 0.0898, accyrasy: 0.966, f1: 0.9343  f1_dev: 0.8553\n",
      "[13/30] [529/594], loss: 0.0971, accyrasy: 0.9616, f1: 0.9163  f1_dev: 0.855\n",
      "[13/30] [534/594], loss: 0.0946, accyrasy: 0.9629, f1: 0.9223  f1_dev: 0.895\n",
      "[13/30] [539/594], loss: 0.0913, accyrasy: 0.9641, f1: 0.933  f1_dev: 0.7861\n",
      "[13/30] [544/594], loss: 0.0907, accyrasy: 0.9637, f1: 0.9194  f1_dev: 0.765\n",
      "[13/30] [549/594], loss: 0.0907, accyrasy: 0.9635, f1: 0.9264  f1_dev: 0.877\n",
      "[13/30] [554/594], loss: 0.0883, accyrasy: 0.967, f1: 0.9398  f1_dev: 0.8723\n",
      "[13/30] [559/594], loss: 0.09, accyrasy: 0.9663, f1: 0.9357  f1_dev: 0.8617\n",
      "[13/30] [564/594], loss: 0.0919, accyrasy: 0.9661, f1: 0.938  f1_dev: 0.757\n",
      "[13/30] [569/594], loss: 0.0988, accyrasy: 0.9581, f1: 0.907  f1_dev: 0.871\n",
      "[13/30] [574/594], loss: 0.0925, accyrasy: 0.9704, f1: 0.9353  f1_dev: 0.8366\n",
      "[13/30] [579/594], loss: 0.0877, accyrasy: 0.9661, f1: 0.9377  f1_dev: 0.7664\n",
      "[13/30] [584/594], loss: 0.092, accyrasy: 0.9676, f1: 0.9417  f1_dev: 0.7566\n",
      "[13/30] [589/594], loss: 0.0903, accyrasy: 0.9649, f1: 0.9381  f1_dev: 0.8634\n",
      "[14/30] [4/594], loss: 0.0954, accyrasy: 0.9626, f1: 0.9213  f1_dev: 0.7507\n",
      "[14/30] [9/594], loss: 0.0907, accyrasy: 0.9656, f1: 0.9308  f1_dev: 0.7093\n",
      "[14/30] [14/594], loss: 0.0919, accyrasy: 0.9669, f1: 0.9305  f1_dev: 0.709\n",
      "[14/30] [19/594], loss: 0.0897, accyrasy: 0.9656, f1: 0.9347  f1_dev: 0.743\n",
      "[14/30] [24/594], loss: 0.0883, accyrasy: 0.9637, f1: 0.9233  f1_dev: 0.8743\n",
      "[14/30] [29/594], loss: 0.0915, accyrasy: 0.9699, f1: 0.9408  f1_dev: 0.7175\n",
      "[14/30] [34/594], loss: 0.0924, accyrasy: 0.9666, f1: 0.9371  f1_dev: 0.7692\n",
      "[14/30] [39/594], loss: 0.0925, accyrasy: 0.9634, f1: 0.916  f1_dev: 0.7674\n",
      "[14/30] [44/594], loss: 0.0895, accyrasy: 0.9633, f1: 0.9263  f1_dev: 0.7744\n",
      "[14/30] [49/594], loss: 0.0881, accyrasy: 0.9644, f1: 0.9364  f1_dev: 0.8957\n",
      "[14/30] [54/594], loss: 0.0882, accyrasy: 0.9655, f1: 0.9324  f1_dev: 0.797\n",
      "[14/30] [59/594], loss: 0.09, accyrasy: 0.9655, f1: 0.9356  f1_dev: 0.7466\n",
      "[14/30] [64/594], loss: 0.0872, accyrasy: 0.9649, f1: 0.9261  f1_dev: 0.787\n",
      "[14/30] [69/594], loss: 0.0883, accyrasy: 0.9671, f1: 0.9346  f1_dev: 0.7025\n",
      "[14/30] [74/594], loss: 0.0896, accyrasy: 0.9667, f1: 0.9366  f1_dev: 0.7365\n",
      "[14/30] [79/594], loss: 0.0925, accyrasy: 0.9641, f1: 0.9358  f1_dev: 0.872\n",
      "[14/30] [84/594], loss: 0.0913, accyrasy: 0.9659, f1: 0.9336  f1_dev: 0.8667\n",
      "[14/30] [89/594], loss: 0.0887, accyrasy: 0.9673, f1: 0.9397  f1_dev: 0.7774\n",
      "[14/30] [94/594], loss: 0.0945, accyrasy: 0.9665, f1: 0.9358  f1_dev: 0.7196\n",
      "[14/30] [99/594], loss: 0.0896, accyrasy: 0.9654, f1: 0.9389  f1_dev: 0.7644\n",
      "[14/30] [104/594], loss: 0.0929, accyrasy: 0.9683, f1: 0.9366  f1_dev: 0.7634\n",
      "[14/30] [109/594], loss: 0.0964, accyrasy: 0.9645, f1: 0.9277  f1_dev: 0.7482\n",
      "[14/30] [114/594], loss: 0.0896, accyrasy: 0.9677, f1: 0.9332  f1_dev: 0.7295\n",
      "[14/30] [119/594], loss: 0.1009, accyrasy: 0.9541, f1: 0.9032  f1_dev: 0.7294\n",
      "[14/30] [124/594], loss: 0.0897, accyrasy: 0.9651, f1: 0.9311  f1_dev: 0.8586\n",
      "[14/30] [129/594], loss: 0.0903, accyrasy: 0.9649, f1: 0.933  f1_dev: 0.8298\n",
      "[14/30] [134/594], loss: 0.0887, accyrasy: 0.9683, f1: 0.9332  f1_dev: 0.855\n",
      "[14/30] [139/594], loss: 0.0878, accyrasy: 0.9659, f1: 0.9337  f1_dev: 0.8555\n",
      "[14/30] [144/594], loss: 0.0913, accyrasy: 0.9675, f1: 0.9373  f1_dev: 0.7706\n",
      "[14/30] [149/594], loss: 0.0842, accyrasy: 0.9665, f1: 0.9338  f1_dev: 0.8809\n",
      "[14/30] [154/594], loss: 0.0961, accyrasy: 0.9584, f1: 0.9105  f1_dev: 0.812\n",
      "[14/30] [159/594], loss: 0.088, accyrasy: 0.9625, f1: 0.9246  f1_dev: 0.8973\n",
      "[14/30] [164/594], loss: 0.0883, accyrasy: 0.9684, f1: 0.9396  f1_dev: 0.7971\n",
      "[14/30] [169/594], loss: 0.0939, accyrasy: 0.9605, f1: 0.9198  f1_dev: 0.8735\n",
      "[14/30] [174/594], loss: 0.0883, accyrasy: 0.9664, f1: 0.9345  f1_dev: 0.8692\n",
      "[14/30] [179/594], loss: 0.0893, accyrasy: 0.9677, f1: 0.9383  f1_dev: 0.8101\n",
      "[14/30] [184/594], loss: 0.0945, accyrasy: 0.9577, f1: 0.9169  f1_dev: 0.8435\n",
      "[14/30] [189/594], loss: 0.0889, accyrasy: 0.9654, f1: 0.9319  f1_dev: 0.8699\n",
      "[14/30] [194/594], loss: 0.084, accyrasy: 0.9671, f1: 0.9384  f1_dev: 0.7206\n",
      "[14/30] [199/594], loss: 0.0918, accyrasy: 0.9643, f1: 0.9321  f1_dev: 0.8029\n",
      "[14/30] [204/594], loss: 0.0931, accyrasy: 0.9633, f1: 0.914  f1_dev: 0.8475\n",
      "[14/30] [209/594], loss: 0.087, accyrasy: 0.9633, f1: 0.9223  f1_dev: 0.8574\n",
      "[14/30] [214/594], loss: 0.0879, accyrasy: 0.9647, f1: 0.9344  f1_dev: 0.7428\n",
      "[14/30] [219/594], loss: 0.0917, accyrasy: 0.9699, f1: 0.9422  f1_dev: 0.8468\n",
      "[14/30] [224/594], loss: 0.0899, accyrasy: 0.9641, f1: 0.9288  f1_dev: 0.8602\n",
      "[14/30] [229/594], loss: 0.0911, accyrasy: 0.9629, f1: 0.9314  f1_dev: 0.8712\n",
      "[14/30] [234/594], loss: 0.0886, accyrasy: 0.963, f1: 0.9316  f1_dev: 0.8967\n",
      "[14/30] [239/594], loss: 0.0899, accyrasy: 0.965, f1: 0.937  f1_dev: 0.7983\n",
      "[14/30] [244/594], loss: 0.0874, accyrasy: 0.9643, f1: 0.9308  f1_dev: 0.7824\n",
      "[14/30] [249/594], loss: 0.0866, accyrasy: 0.9711, f1: 0.9406  f1_dev: 0.7944\n",
      "[14/30] [254/594], loss: 0.0913, accyrasy: 0.9609, f1: 0.9297  f1_dev: 0.8727\n",
      "[14/30] [259/594], loss: 0.0859, accyrasy: 0.9684, f1: 0.9352  f1_dev: 0.8423\n",
      "[14/30] [264/594], loss: 0.0885, accyrasy: 0.9671, f1: 0.9346  f1_dev: 0.7472\n",
      "[14/30] [269/594], loss: 0.0835, accyrasy: 0.9706, f1: 0.9383  f1_dev: 0.7625\n",
      "[14/30] [274/594], loss: 0.089, accyrasy: 0.9634, f1: 0.9251  f1_dev: 0.7936\n",
      "[14/30] [279/594], loss: 0.0856, accyrasy: 0.9661, f1: 0.9347  f1_dev: 0.8135\n",
      "[14/30] [284/594], loss: 0.0868, accyrasy: 0.9686, f1: 0.9374  f1_dev: 0.79\n",
      "[14/30] [289/594], loss: 0.0866, accyrasy: 0.9645, f1: 0.9289  f1_dev: 0.784\n",
      "[14/30] [294/594], loss: 0.0844, accyrasy: 0.9665, f1: 0.9288  f1_dev: 0.7636\n",
      "[14/30] [299/594], loss: 0.0886, accyrasy: 0.9634, f1: 0.9336  f1_dev: 0.7722\n",
      "[14/30] [304/594], loss: 0.0905, accyrasy: 0.9623, f1: 0.9208  f1_dev: 0.8609\n",
      "[14/30] [309/594], loss: 0.0952, accyrasy: 0.9631, f1: 0.9242  f1_dev: 0.7776\n",
      "[14/30] [314/594], loss: 0.1, accyrasy: 0.9515, f1: 0.8977  f1_dev: 0.7569\n",
      "[14/30] [319/594], loss: 0.1031, accyrasy: 0.9601, f1: 0.9293  f1_dev: 0.7267\n",
      "[14/30] [324/594], loss: 0.099, accyrasy: 0.9629, f1: 0.9334  f1_dev: 0.7426\n",
      "[14/30] [329/594], loss: 0.0999, accyrasy: 0.9626, f1: 0.9292  f1_dev: 0.8555\n",
      "[14/30] [334/594], loss: 0.0939, accyrasy: 0.9634, f1: 0.9327  f1_dev: 0.7839\n",
      "[14/30] [339/594], loss: 0.0881, accyrasy: 0.9636, f1: 0.9336  f1_dev: 0.7708\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[14/30] [344/594], loss: 0.0972, accyrasy: 0.9595, f1: 0.9267  f1_dev: 0.7586\n",
      "[14/30] [349/594], loss: 0.0924, accyrasy: 0.9674, f1: 0.9391  f1_dev: 0.7715\n",
      "[14/30] [354/594], loss: 0.0942, accyrasy: 0.9676, f1: 0.9365  f1_dev: 0.751\n",
      "[14/30] [359/594], loss: 0.0911, accyrasy: 0.9649, f1: 0.9291  f1_dev: 0.7308\n",
      "[14/30] [364/594], loss: 0.0866, accyrasy: 0.9684, f1: 0.9381  f1_dev: 0.7608\n",
      "[14/30] [369/594], loss: 0.0888, accyrasy: 0.9666, f1: 0.9358  f1_dev: 0.8684\n",
      "[14/30] [374/594], loss: 0.0964, accyrasy: 0.9649, f1: 0.9296  f1_dev: 0.8764\n",
      "[14/30] [379/594], loss: 0.09, accyrasy: 0.9676, f1: 0.9353  f1_dev: 0.8757\n",
      "[14/30] [384/594], loss: 0.0918, accyrasy: 0.9673, f1: 0.9375  f1_dev: 0.8961\n",
      "[14/30] [389/594], loss: 0.088, accyrasy: 0.9659, f1: 0.9316  f1_dev: 0.8898\n",
      "[14/30] [394/594], loss: 0.0878, accyrasy: 0.9667, f1: 0.9364  f1_dev: 0.8975\n",
      "[14/30] [399/594], loss: 0.09, accyrasy: 0.9641, f1: 0.935  f1_dev: 0.8791\n",
      "[14/30] [404/594], loss: 0.0882, accyrasy: 0.9684, f1: 0.9424  f1_dev: 0.8743\n",
      "[14/30] [409/594], loss: 0.0853, accyrasy: 0.9664, f1: 0.9324  f1_dev: 0.8769\n",
      "[14/30] [414/594], loss: 0.0928, accyrasy: 0.963, f1: 0.9258  f1_dev: 0.8646\n",
      "[14/30] [419/594], loss: 0.0913, accyrasy: 0.964, f1: 0.934  f1_dev: 0.7725\n",
      "[14/30] [424/594], loss: 0.0852, accyrasy: 0.9673, f1: 0.9381  f1_dev: 0.8555\n",
      "[14/30] [429/594], loss: 0.0896, accyrasy: 0.9626, f1: 0.922  f1_dev: 0.8671\n",
      "[14/30] [434/594], loss: 0.0889, accyrasy: 0.9645, f1: 0.9241  f1_dev: 0.8734\n",
      "[14/30] [439/594], loss: 0.0898, accyrasy: 0.964, f1: 0.9319  f1_dev: 0.8916\n",
      "[14/30] [444/594], loss: 0.0934, accyrasy: 0.9634, f1: 0.9317  f1_dev: 0.8703\n",
      "[14/30] [449/594], loss: 0.0941, accyrasy: 0.9656, f1: 0.9338  f1_dev: 0.7465\n",
      "[14/30] [454/594], loss: 0.09, accyrasy: 0.9663, f1: 0.934  f1_dev: 0.8597\n",
      "[14/30] [459/594], loss: 0.0863, accyrasy: 0.9686, f1: 0.9385  f1_dev: 0.8709\n",
      "[14/30] [464/594], loss: 0.0889, accyrasy: 0.9676, f1: 0.9384  f1_dev: 0.7387\n",
      "[14/30] [469/594], loss: 0.092, accyrasy: 0.9683, f1: 0.9403  f1_dev: 0.799\n",
      "[14/30] [474/594], loss: 0.0972, accyrasy: 0.9625, f1: 0.9235  f1_dev: 0.8645\n",
      "[14/30] [479/594], loss: 0.0866, accyrasy: 0.9681, f1: 0.942  f1_dev: 0.8688\n",
      "[14/30] [484/594], loss: 0.0897, accyrasy: 0.9653, f1: 0.9301  f1_dev: 0.8413\n",
      "[14/30] [489/594], loss: 0.0863, accyrasy: 0.9673, f1: 0.9172  f1_dev: 0.7771\n",
      "[14/30] [494/594], loss: 0.0896, accyrasy: 0.9688, f1: 0.9401  f1_dev: 0.8887\n",
      "[14/30] [499/594], loss: 0.0923, accyrasy: 0.9679, f1: 0.9342  f1_dev: 0.8578\n",
      "[14/30] [504/594], loss: 0.0882, accyrasy: 0.965, f1: 0.9268  f1_dev: 0.8759\n",
      "[14/30] [509/594], loss: 0.0903, accyrasy: 0.963, f1: 0.9317  f1_dev: 0.8557\n",
      "[14/30] [514/594], loss: 0.0933, accyrasy: 0.9644, f1: 0.9258  f1_dev: 0.8488\n",
      "[14/30] [519/594], loss: 0.0911, accyrasy: 0.9603, f1: 0.916  f1_dev: 0.8228\n",
      "[14/30] [524/594], loss: 0.0947, accyrasy: 0.9654, f1: 0.9304  f1_dev: 0.7543\n",
      "[14/30] [529/594], loss: 0.0945, accyrasy: 0.962, f1: 0.9207  f1_dev: 0.8659\n",
      "[14/30] [534/594], loss: 0.0927, accyrasy: 0.9584, f1: 0.9211  f1_dev: 0.7959\n",
      "[14/30] [539/594], loss: 0.0954, accyrasy: 0.9649, f1: 0.9298  f1_dev: 0.7685\n",
      "[14/30] [544/594], loss: 0.1299, accyrasy: 0.9514, f1: 0.9117  f1_dev: 0.7033\n",
      "[14/30] [549/594], loss: 0.1405, accyrasy: 0.9573, f1: 0.9218  f1_dev: 0.6349\n",
      "[14/30] [554/594], loss: 0.129, accyrasy: 0.9571, f1: 0.9216  f1_dev: 0.6216\n",
      "[14/30] [559/594], loss: 0.1149, accyrasy: 0.9614, f1: 0.9235  f1_dev: 0.6835\n",
      "[14/30] [564/594], loss: 0.1052, accyrasy: 0.966, f1: 0.9345  f1_dev: 0.7403\n",
      "[14/30] [569/594], loss: 0.1018, accyrasy: 0.9639, f1: 0.9295  f1_dev: 0.8306\n",
      "[14/30] [574/594], loss: 0.0984, accyrasy: 0.9667, f1: 0.9352  f1_dev: 0.7681\n",
      "[14/30] [579/594], loss: 0.0922, accyrasy: 0.9641, f1: 0.9378  f1_dev: 0.6873\n",
      "[14/30] [584/594], loss: 0.096, accyrasy: 0.9629, f1: 0.9233  f1_dev: 0.6943\n",
      "[14/30] [589/594], loss: 0.1004, accyrasy: 0.9647, f1: 0.9327  f1_dev: 0.7054\n",
      "[15/30] [4/594], loss: 0.094, accyrasy: 0.9665, f1: 0.9382  f1_dev: 0.7579\n",
      "[15/30] [9/594], loss: 0.0881, accyrasy: 0.9667, f1: 0.9402  f1_dev: 0.7198\n",
      "[15/30] [14/594], loss: 0.0915, accyrasy: 0.9661, f1: 0.9319  f1_dev: 0.7179\n",
      "[15/30] [19/594], loss: 0.0892, accyrasy: 0.9665, f1: 0.9363  f1_dev: 0.7491\n",
      "[15/30] [24/594], loss: 0.0878, accyrasy: 0.9669, f1: 0.9393  f1_dev: 0.812\n",
      "[15/30] [29/594], loss: 0.0868, accyrasy: 0.966, f1: 0.9311  f1_dev: 0.733\n",
      "[15/30] [34/594], loss: 0.0877, accyrasy: 0.9673, f1: 0.9354  f1_dev: 0.763\n",
      "[15/30] [39/594], loss: 0.0862, accyrasy: 0.9718, f1: 0.9448  f1_dev: 0.8244\n",
      "[15/30] [44/594], loss: 0.0907, accyrasy: 0.9669, f1: 0.9352  f1_dev: 0.7529\n",
      "[15/30] [49/594], loss: 0.0936, accyrasy: 0.9696, f1: 0.9459  f1_dev: 0.7387\n",
      "[15/30] [54/594], loss: 0.0879, accyrasy: 0.9692, f1: 0.9376  f1_dev: 0.7328\n",
      "[15/30] [59/594], loss: 0.0894, accyrasy: 0.9661, f1: 0.9327  f1_dev: 0.7426\n",
      "[15/30] [64/594], loss: 0.0884, accyrasy: 0.9645, f1: 0.9266  f1_dev: 0.7583\n",
      "[15/30] [69/594], loss: 0.0869, accyrasy: 0.9706, f1: 0.9424  f1_dev: 0.7832\n",
      "[15/30] [74/594], loss: 0.089, accyrasy: 0.9664, f1: 0.9354  f1_dev: 0.7801\n",
      "[15/30] [79/594], loss: 0.0896, accyrasy: 0.9623, f1: 0.927  f1_dev: 0.8372\n",
      "[15/30] [84/594], loss: 0.087, accyrasy: 0.9663, f1: 0.9359  f1_dev: 0.8366\n",
      "[15/30] [89/594], loss: 0.0881, accyrasy: 0.9686, f1: 0.939  f1_dev: 0.772\n",
      "[15/30] [94/594], loss: 0.09, accyrasy: 0.9647, f1: 0.9308  f1_dev: 0.8475\n",
      "[15/30] [99/594], loss: 0.0916, accyrasy: 0.965, f1: 0.9272  f1_dev: 0.8412\n",
      "[15/30] [104/594], loss: 0.0846, accyrasy: 0.9688, f1: 0.9369  f1_dev: 0.8265\n",
      "[15/30] [109/594], loss: 0.0873, accyrasy: 0.9681, f1: 0.9435  f1_dev: 0.8353\n",
      "[15/30] [114/594], loss: 0.085, accyrasy: 0.9661, f1: 0.9365  f1_dev: 0.8125\n",
      "[15/30] [119/594], loss: 0.0947, accyrasy: 0.9565, f1: 0.9193  f1_dev: 0.8581\n",
      "[15/30] [124/594], loss: 0.0892, accyrasy: 0.9696, f1: 0.9425  f1_dev: 0.8457\n",
      "[15/30] [129/594], loss: 0.0853, accyrasy: 0.9708, f1: 0.9455  f1_dev: 0.7649\n",
      "[15/30] [134/594], loss: 0.0892, accyrasy: 0.9651, f1: 0.9339  f1_dev: 0.7801\n",
      "[15/30] [139/594], loss: 0.085, accyrasy: 0.9688, f1: 0.9338  f1_dev: 0.8677\n",
      "[15/30] [144/594], loss: 0.0885, accyrasy: 0.9686, f1: 0.937  f1_dev: 0.8753\n",
      "[15/30] [149/594], loss: 0.0899, accyrasy: 0.9609, f1: 0.9234  f1_dev: 0.8284\n",
      "[15/30] [154/594], loss: 0.0888, accyrasy: 0.969, f1: 0.9335  f1_dev: 0.8236\n",
      "[15/30] [159/594], loss: 0.0858, accyrasy: 0.9649, f1: 0.9341  f1_dev: 0.8291\n",
      "[15/30] [164/594], loss: 0.0977, accyrasy: 0.9613, f1: 0.9325  f1_dev: 0.782\n",
      "[15/30] [169/594], loss: 0.096, accyrasy: 0.9625, f1: 0.9152  f1_dev: 0.83\n",
      "[15/30] [174/594], loss: 0.0922, accyrasy: 0.9663, f1: 0.9336  f1_dev: 0.8442\n",
      "[15/30] [179/594], loss: 0.0883, accyrasy: 0.9675, f1: 0.9365  f1_dev: 0.8151\n",
      "[15/30] [184/594], loss: 0.0867, accyrasy: 0.9686, f1: 0.9368  f1_dev: 0.6959\n",
      "[15/30] [189/594], loss: 0.0874, accyrasy: 0.9671, f1: 0.9304  f1_dev: 0.682\n",
      "[15/30] [194/594], loss: 0.0873, accyrasy: 0.9666, f1: 0.9402  f1_dev: 0.8622\n",
      "[15/30] [199/594], loss: 0.0914, accyrasy: 0.9611, f1: 0.9215  f1_dev: 0.8654\n",
      "[15/30] [204/594], loss: 0.09, accyrasy: 0.9673, f1: 0.9308  f1_dev: 0.8352\n",
      "[15/30] [209/594], loss: 0.0861, accyrasy: 0.968, f1: 0.9406  f1_dev: 0.8268\n",
      "[15/30] [214/594], loss: 0.0882, accyrasy: 0.9679, f1: 0.9407  f1_dev: 0.8548\n",
      "[15/30] [219/594], loss: 0.0869, accyrasy: 0.9689, f1: 0.9403  f1_dev: 0.8546\n",
      "[15/30] [224/594], loss: 0.0884, accyrasy: 0.9641, f1: 0.9297  f1_dev: 0.8659\n",
      "[15/30] [229/594], loss: 0.092, accyrasy: 0.9706, f1: 0.9366  f1_dev: 0.7776\n",
      "[15/30] [234/594], loss: 0.0919, accyrasy: 0.9631, f1: 0.9265  f1_dev: 0.7416\n",
      "[15/30] [239/594], loss: 0.0951, accyrasy: 0.9661, f1: 0.9291  f1_dev: 0.7967\n",
      "[15/30] [244/594], loss: 0.0919, accyrasy: 0.964, f1: 0.9286  f1_dev: 0.8708\n",
      "[15/30] [249/594], loss: 0.0906, accyrasy: 0.9674, f1: 0.9362  f1_dev: 0.899\n",
      "[15/30] [254/594], loss: 0.09, accyrasy: 0.965, f1: 0.9327  f1_dev: 0.887\n",
      "[15/30] [259/594], loss: 0.0908, accyrasy: 0.9679, f1: 0.9399  f1_dev: 0.7858\n",
      "[15/30] [264/594], loss: 0.0905, accyrasy: 0.9663, f1: 0.9329  f1_dev: 0.8576\n",
      "[15/30] [269/594], loss: 0.0958, accyrasy: 0.9537, f1: 0.9092  f1_dev: 0.7697\n",
      "[15/30] [274/594], loss: 0.0917, accyrasy: 0.9644, f1: 0.9238  f1_dev: 0.8614\n",
      "[15/30] [279/594], loss: 0.0895, accyrasy: 0.9667, f1: 0.932  f1_dev: 0.8458\n",
      "[15/30] [284/594], loss: 0.0849, accyrasy: 0.9701, f1: 0.9452  f1_dev: 0.8412\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[15/30] [289/594], loss: 0.0932, accyrasy: 0.9623, f1: 0.9276  f1_dev: 0.8427\n",
      "[15/30] [294/594], loss: 0.0915, accyrasy: 0.9606, f1: 0.9275  f1_dev: 0.7781\n",
      "[15/30] [299/594], loss: 0.0911, accyrasy: 0.9666, f1: 0.932  f1_dev: 0.7673\n",
      "[15/30] [304/594], loss: 0.0935, accyrasy: 0.9656, f1: 0.9318  f1_dev: 0.7729\n",
      "[15/30] [309/594], loss: 0.096, accyrasy: 0.9595, f1: 0.9233  f1_dev: 0.8299\n",
      "[15/30] [314/594], loss: 0.0947, accyrasy: 0.9617, f1: 0.9219  f1_dev: 0.856\n",
      "[15/30] [319/594], loss: 0.0883, accyrasy: 0.9681, f1: 0.93  f1_dev: 0.8539\n",
      "[15/30] [324/594], loss: 0.0845, accyrasy: 0.9669, f1: 0.9287  f1_dev: 0.7753\n",
      "[15/30] [329/594], loss: 0.0879, accyrasy: 0.965, f1: 0.9302  f1_dev: 0.8523\n",
      "[15/30] [334/594], loss: 0.0915, accyrasy: 0.9655, f1: 0.9239  f1_dev: 0.8362\n",
      "[15/30] [339/594], loss: 0.0927, accyrasy: 0.964, f1: 0.9171  f1_dev: 0.8332\n",
      "[15/30] [344/594], loss: 0.088, accyrasy: 0.9677, f1: 0.9293  f1_dev: 0.8658\n",
      "[15/30] [349/594], loss: 0.0891, accyrasy: 0.9647, f1: 0.9245  f1_dev: 0.8751\n",
      "[15/30] [354/594], loss: 0.0875, accyrasy: 0.9673, f1: 0.934  f1_dev: 0.8709\n",
      "[15/30] [359/594], loss: 0.0906, accyrasy: 0.9675, f1: 0.9363  f1_dev: 0.8448\n",
      "[15/30] [364/594], loss: 0.0902, accyrasy: 0.9667, f1: 0.9363  f1_dev: 0.8672\n",
      "[15/30] [369/594], loss: 0.0924, accyrasy: 0.9669, f1: 0.9318  f1_dev: 0.8449\n",
      "[15/30] [374/594], loss: 0.0901, accyrasy: 0.966, f1: 0.9321  f1_dev: 0.7534\n",
      "[15/30] [379/594], loss: 0.0853, accyrasy: 0.9701, f1: 0.9402  f1_dev: 0.8375\n",
      "[15/30] [384/594], loss: 0.0852, accyrasy: 0.9695, f1: 0.9471  f1_dev: 0.8685\n",
      "[15/30] [389/594], loss: 0.0899, accyrasy: 0.9617, f1: 0.9268  f1_dev: 0.8006\n",
      "[15/30] [394/594], loss: 0.0852, accyrasy: 0.9656, f1: 0.9317  f1_dev: 0.745\n",
      "[15/30] [399/594], loss: 0.0862, accyrasy: 0.9649, f1: 0.9286  f1_dev: 0.7868\n",
      "[15/30] [404/594], loss: 0.094, accyrasy: 0.9681, f1: 0.9305  f1_dev: 0.8048\n",
      "[15/30] [409/594], loss: 0.1003, accyrasy: 0.9599, f1: 0.9139  f1_dev: 0.8363\n",
      "[15/30] [414/594], loss: 0.0953, accyrasy: 0.964, f1: 0.9261  f1_dev: 0.7815\n",
      "[15/30] [419/594], loss: 0.0908, accyrasy: 0.969, f1: 0.938  f1_dev: 0.8582\n",
      "[15/30] [424/594], loss: 0.0908, accyrasy: 0.9646, f1: 0.9299  f1_dev: 0.8728\n",
      "[15/30] [429/594], loss: 0.0897, accyrasy: 0.9644, f1: 0.9242  f1_dev: 0.7297\n",
      "[15/30] [434/594], loss: 0.0873, accyrasy: 0.9689, f1: 0.9421  f1_dev: 0.861\n",
      "[15/30] [439/594], loss: 0.0868, accyrasy: 0.9627, f1: 0.926  f1_dev: 0.7579\n",
      "[15/30] [444/594], loss: 0.0844, accyrasy: 0.9674, f1: 0.9334  f1_dev: 0.8519\n",
      "[15/30] [449/594], loss: 0.0911, accyrasy: 0.9643, f1: 0.9299  f1_dev: 0.8728\n",
      "[15/30] [454/594], loss: 0.0893, accyrasy: 0.9665, f1: 0.9364  f1_dev: 0.8408\n",
      "[15/30] [459/594], loss: 0.0834, accyrasy: 0.9688, f1: 0.9415  f1_dev: 0.7796\n",
      "[15/30] [464/594], loss: 0.0901, accyrasy: 0.9624, f1: 0.9239  f1_dev: 0.7676\n",
      "[15/30] [469/594], loss: 0.0862, accyrasy: 0.9699, f1: 0.9385  f1_dev: 0.8633\n",
      "[15/30] [474/594], loss: 0.0879, accyrasy: 0.968, f1: 0.9364  f1_dev: 0.8762\n",
      "[15/30] [479/594], loss: 0.0836, accyrasy: 0.9676, f1: 0.9366  f1_dev: 0.8812\n",
      "[15/30] [484/594], loss: 0.0969, accyrasy: 0.9647, f1: 0.9362  f1_dev: 0.874\n",
      "[15/30] [489/594], loss: 0.0857, accyrasy: 0.9647, f1: 0.9295  f1_dev: 0.8839\n",
      "[15/30] [494/594], loss: 0.0867, accyrasy: 0.9702, f1: 0.9445  f1_dev: 0.8859\n",
      "[15/30] [499/594], loss: 0.0881, accyrasy: 0.9635, f1: 0.9333  f1_dev: 0.7644\n",
      "[15/30] [504/594], loss: 0.0864, accyrasy: 0.966, f1: 0.9302  f1_dev: 0.7468\n",
      "[15/30] [509/594], loss: 0.0916, accyrasy: 0.9606, f1: 0.9202  f1_dev: 0.8744\n",
      "[15/30] [514/594], loss: 0.093, accyrasy: 0.9644, f1: 0.9352  f1_dev: 0.8597\n",
      "[15/30] [519/594], loss: 0.0884, accyrasy: 0.9634, f1: 0.9294  f1_dev: 0.8726\n",
      "[15/30] [524/594], loss: 0.0853, accyrasy: 0.9686, f1: 0.9391  f1_dev: 0.7747\n",
      "[15/30] [529/594], loss: 0.091, accyrasy: 0.9686, f1: 0.9332  f1_dev: 0.9033\n",
      "[15/30] [534/594], loss: 0.15, accyrasy: 0.9295, f1: 0.8787  f1_dev: 0.4086\n",
      "[15/30] [539/594], loss: 0.1875, accyrasy: 0.9569, f1: 0.9179  f1_dev: 0.708\n",
      "[15/30] [544/594], loss: 0.1285, accyrasy: 0.9566, f1: 0.9197  f1_dev: 0.7322\n",
      "[15/30] [549/594], loss: 0.1172, accyrasy: 0.9567, f1: 0.9184  f1_dev: 0.6932\n",
      "[15/30] [554/594], loss: 0.108, accyrasy: 0.962, f1: 0.9255  f1_dev: 0.6899\n",
      "[15/30] [559/594], loss: 0.1016, accyrasy: 0.958, f1: 0.9199  f1_dev: 0.748\n",
      "[15/30] [564/594], loss: 0.1089, accyrasy: 0.9637, f1: 0.9329  f1_dev: 0.7235\n",
      "[15/30] [569/594], loss: 0.1082, accyrasy: 0.9609, f1: 0.9241  f1_dev: 0.6579\n",
      "[15/30] [574/594], loss: 0.1016, accyrasy: 0.962, f1: 0.9192  f1_dev: 0.6592\n",
      "[15/30] [579/594], loss: 0.108, accyrasy: 0.955, f1: 0.9167  f1_dev: 0.6715\n",
      "[15/30] [584/594], loss: 0.1079, accyrasy: 0.959, f1: 0.9212  f1_dev: 0.632\n",
      "[15/30] [589/594], loss: 0.1129, accyrasy: 0.9597, f1: 0.9207  f1_dev: 0.6261\n",
      "[16/30] [4/594], loss: 0.1066, accyrasy: 0.9614, f1: 0.9219  f1_dev: 0.7554\n",
      "[16/30] [9/594], loss: 0.1005, accyrasy: 0.9626, f1: 0.9242  f1_dev: 0.7318\n",
      "[16/30] [14/594], loss: 0.0986, accyrasy: 0.9633, f1: 0.928  f1_dev: 0.6992\n",
      "[16/30] [19/594], loss: 0.095, accyrasy: 0.9653, f1: 0.9345  f1_dev: 0.7501\n",
      "[16/30] [24/594], loss: 0.095, accyrasy: 0.9624, f1: 0.9284  f1_dev: 0.74\n",
      "[16/30] [29/594], loss: 0.0922, accyrasy: 0.9657, f1: 0.9338  f1_dev: 0.7132\n",
      "[16/30] [34/594], loss: 0.0871, accyrasy: 0.9653, f1: 0.9309  f1_dev: 0.758\n",
      "[16/30] [39/594], loss: 0.0897, accyrasy: 0.9647, f1: 0.9274  f1_dev: 0.7323\n",
      "[16/30] [44/594], loss: 0.0943, accyrasy: 0.9637, f1: 0.929  f1_dev: 0.7122\n",
      "[16/30] [49/594], loss: 0.0889, accyrasy: 0.9684, f1: 0.938  f1_dev: 0.7305\n",
      "[16/30] [54/594], loss: 0.0906, accyrasy: 0.9656, f1: 0.9311  f1_dev: 0.7575\n",
      "[16/30] [59/594], loss: 0.0905, accyrasy: 0.9654, f1: 0.9277  f1_dev: 0.7577\n",
      "[16/30] [64/594], loss: 0.0894, accyrasy: 0.9683, f1: 0.9375  f1_dev: 0.7375\n",
      "[16/30] [69/594], loss: 0.0906, accyrasy: 0.9673, f1: 0.9364  f1_dev: 0.8448\n",
      "[16/30] [74/594], loss: 0.0882, accyrasy: 0.965, f1: 0.9342  f1_dev: 0.7654\n",
      "[16/30] [79/594], loss: 0.0858, accyrasy: 0.9696, f1: 0.9397  f1_dev: 0.7665\n",
      "[16/30] [84/594], loss: 0.0937, accyrasy: 0.9651, f1: 0.9324  f1_dev: 0.7617\n",
      "[16/30] [89/594], loss: 0.0861, accyrasy: 0.966, f1: 0.9364  f1_dev: 0.768\n",
      "[16/30] [94/594], loss: 0.0886, accyrasy: 0.9656, f1: 0.9333  f1_dev: 0.7311\n",
      "[16/30] [99/594], loss: 0.091, accyrasy: 0.9635, f1: 0.9227  f1_dev: 0.7951\n",
      "[16/30] [104/594], loss: 0.096, accyrasy: 0.9708, f1: 0.9472  f1_dev: 0.7446\n",
      "[16/30] [109/594], loss: 0.0958, accyrasy: 0.9609, f1: 0.9238  f1_dev: 0.7326\n",
      "[16/30] [114/594], loss: 0.1019, accyrasy: 0.9484, f1: 0.8999  f1_dev: 0.7815\n",
      "[16/30] [119/594], loss: 0.0908, accyrasy: 0.9619, f1: 0.9326  f1_dev: 0.876\n",
      "[16/30] [124/594], loss: 0.0903, accyrasy: 0.9698, f1: 0.9442  f1_dev: 0.8593\n",
      "[16/30] [129/594], loss: 0.0948, accyrasy: 0.9616, f1: 0.9147  f1_dev: 0.8373\n",
      "[16/30] [134/594], loss: 0.0874, accyrasy: 0.9696, f1: 0.9416  f1_dev: 0.8515\n",
      "[16/30] [139/594], loss: 0.0865, accyrasy: 0.9671, f1: 0.9374  f1_dev: 0.8491\n",
      "[16/30] [144/594], loss: 0.0848, accyrasy: 0.9676, f1: 0.9366  f1_dev: 0.7538\n",
      "[16/30] [149/594], loss: 0.0845, accyrasy: 0.9684, f1: 0.9358  f1_dev: 0.788\n",
      "[16/30] [154/594], loss: 0.0844, accyrasy: 0.9685, f1: 0.938  f1_dev: 0.7979\n",
      "[16/30] [159/594], loss: 0.0859, accyrasy: 0.9698, f1: 0.9387  f1_dev: 0.8007\n",
      "[16/30] [164/594], loss: 0.0841, accyrasy: 0.969, f1: 0.9384  f1_dev: 0.8579\n",
      "[16/30] [169/594], loss: 0.0888, accyrasy: 0.9646, f1: 0.9291  f1_dev: 0.7758\n",
      "[16/30] [174/594], loss: 0.0842, accyrasy: 0.9699, f1: 0.9423  f1_dev: 0.8734\n",
      "[16/30] [179/594], loss: 0.0895, accyrasy: 0.969, f1: 0.937  f1_dev: 0.7748\n",
      "[16/30] [184/594], loss: 0.0845, accyrasy: 0.9681, f1: 0.9424  f1_dev: 0.8215\n",
      "[16/30] [189/594], loss: 0.0878, accyrasy: 0.9712, f1: 0.943  f1_dev: 0.8459\n",
      "[16/30] [194/594], loss: 0.0834, accyrasy: 0.9651, f1: 0.9337  f1_dev: 0.8642\n",
      "[16/30] [199/594], loss: 0.0864, accyrasy: 0.9641, f1: 0.9257  f1_dev: 0.7457\n",
      "[16/30] [204/594], loss: 0.0877, accyrasy: 0.9665, f1: 0.9342  f1_dev: 0.7262\n",
      "[16/30] [209/594], loss: 0.0893, accyrasy: 0.9645, f1: 0.9303  f1_dev: 0.7967\n",
      "[16/30] [214/594], loss: 0.0866, accyrasy: 0.969, f1: 0.9389  f1_dev: 0.8986\n",
      "[16/30] [219/594], loss: 0.0872, accyrasy: 0.9659, f1: 0.9335  f1_dev: 0.8838\n",
      "[16/30] [224/594], loss: 0.087, accyrasy: 0.9686, f1: 0.9413  f1_dev: 0.7776\n",
      "[16/30] [229/594], loss: 0.0941, accyrasy: 0.9636, f1: 0.9304  f1_dev: 0.8933\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[16/30] [234/594], loss: 0.0983, accyrasy: 0.9635, f1: 0.9308  f1_dev: 0.8953\n",
      "[16/30] [239/594], loss: 0.0929, accyrasy: 0.9635, f1: 0.9305  f1_dev: 0.7772\n",
      "[16/30] [244/594], loss: 0.0908, accyrasy: 0.9686, f1: 0.9429  f1_dev: 0.7153\n",
      "[16/30] [249/594], loss: 0.0949, accyrasy: 0.9634, f1: 0.9288  f1_dev: 0.7456\n",
      "[16/30] [254/594], loss: 0.0949, accyrasy: 0.9616, f1: 0.9263  f1_dev: 0.853\n",
      "[16/30] [259/594], loss: 0.0928, accyrasy: 0.964, f1: 0.9295  f1_dev: 0.849\n",
      "[16/30] [264/594], loss: 0.0945, accyrasy: 0.9663, f1: 0.9373  f1_dev: 0.6988\n",
      "[16/30] [269/594], loss: 0.0916, accyrasy: 0.968, f1: 0.9353  f1_dev: 0.7368\n",
      "[16/30] [274/594], loss: 0.0903, accyrasy: 0.9639, f1: 0.9277  f1_dev: 0.8522\n",
      "[16/30] [279/594], loss: 0.0923, accyrasy: 0.9666, f1: 0.9424  f1_dev: 0.8757\n",
      "[16/30] [284/594], loss: 0.0893, accyrasy: 0.9649, f1: 0.9275  f1_dev: 0.8592\n",
      "[16/30] [289/594], loss: 0.0815, accyrasy: 0.969, f1: 0.943  f1_dev: 0.7717\n",
      "[16/30] [294/594], loss: 0.0896, accyrasy: 0.9688, f1: 0.9351  f1_dev: 0.7239\n",
      "[16/30] [299/594], loss: 0.0857, accyrasy: 0.9695, f1: 0.9431  f1_dev: 0.7823\n",
      "[16/30] [304/594], loss: 0.085, accyrasy: 0.9665, f1: 0.9359  f1_dev: 0.764\n",
      "[16/30] [309/594], loss: 0.087, accyrasy: 0.9675, f1: 0.9304  f1_dev: 0.7779\n",
      "[16/30] [314/594], loss: 0.086, accyrasy: 0.9702, f1: 0.9435  f1_dev: 0.7794\n",
      "[16/30] [319/594], loss: 0.0862, accyrasy: 0.9676, f1: 0.9411  f1_dev: 0.8579\n",
      "[16/30] [324/594], loss: 0.0971, accyrasy: 0.968, f1: 0.9353  f1_dev: 0.836\n",
      "[16/30] [329/594], loss: 0.0885, accyrasy: 0.9677, f1: 0.9366  f1_dev: 0.8106\n",
      "[16/30] [334/594], loss: 0.0895, accyrasy: 0.967, f1: 0.9382  f1_dev: 0.7675\n",
      "[16/30] [339/594], loss: 0.0902, accyrasy: 0.9684, f1: 0.9396  f1_dev: 0.8716\n",
      "[16/30] [344/594], loss: 0.0905, accyrasy: 0.9644, f1: 0.9277  f1_dev: 0.8603\n",
      "[16/30] [349/594], loss: 0.089, accyrasy: 0.9647, f1: 0.9324  f1_dev: 0.7742\n",
      "[16/30] [354/594], loss: 0.088, accyrasy: 0.97, f1: 0.9395  f1_dev: 0.7821\n",
      "[16/30] [359/594], loss: 0.0842, accyrasy: 0.969, f1: 0.9421  f1_dev: 0.8787\n",
      "[16/30] [364/594], loss: 0.0885, accyrasy: 0.9677, f1: 0.932  f1_dev: 0.8021\n",
      "[16/30] [369/594], loss: 0.0845, accyrasy: 0.9645, f1: 0.9352  f1_dev: 0.7719\n",
      "[16/30] [374/594], loss: 0.0868, accyrasy: 0.9671, f1: 0.9345  f1_dev: 0.7805\n",
      "[16/30] [379/594], loss: 0.0893, accyrasy: 0.9656, f1: 0.935  f1_dev: 0.8775\n",
      "[16/30] [384/594], loss: 0.0887, accyrasy: 0.968, f1: 0.938  f1_dev: 0.7749\n",
      "[16/30] [389/594], loss: 0.0889, accyrasy: 0.9676, f1: 0.9366  f1_dev: 0.7924\n",
      "[16/30] [394/594], loss: 0.0875, accyrasy: 0.9663, f1: 0.9349  f1_dev: 0.8788\n",
      "[16/30] [399/594], loss: 0.092, accyrasy: 0.9649, f1: 0.935  f1_dev: 0.8812\n",
      "[16/30] [404/594], loss: 0.0875, accyrasy: 0.9647, f1: 0.9308  f1_dev: 0.8756\n",
      "[16/30] [409/594], loss: 0.0884, accyrasy: 0.9669, f1: 0.9359  f1_dev: 0.7415\n",
      "[16/30] [414/594], loss: 0.0879, accyrasy: 0.9636, f1: 0.9245  f1_dev: 0.7714\n",
      "[16/30] [419/594], loss: 0.0932, accyrasy: 0.9685, f1: 0.9336  f1_dev: 0.8981\n",
      "[16/30] [424/594], loss: 0.0858, accyrasy: 0.9721, f1: 0.9492  f1_dev: 0.8763\n",
      "[16/30] [429/594], loss: 0.089, accyrasy: 0.9611, f1: 0.9248  f1_dev: 0.8416\n",
      "[16/30] [434/594], loss: 0.0905, accyrasy: 0.9649, f1: 0.9288  f1_dev: 0.8405\n",
      "[16/30] [439/594], loss: 0.0877, accyrasy: 0.9613, f1: 0.9207  f1_dev: 0.8792\n",
      "[16/30] [444/594], loss: 0.0873, accyrasy: 0.9659, f1: 0.9334  f1_dev: 0.7941\n",
      "[16/30] [449/594], loss: 0.0861, accyrasy: 0.9654, f1: 0.9279  f1_dev: 0.7816\n",
      "[16/30] [454/594], loss: 0.0911, accyrasy: 0.9643, f1: 0.9241  f1_dev: 0.8428\n",
      "[16/30] [459/594], loss: 0.0898, accyrasy: 0.968, f1: 0.9368  f1_dev: 0.8759\n",
      "[16/30] [464/594], loss: 0.0851, accyrasy: 0.9677, f1: 0.941  f1_dev: 0.8652\n",
      "[16/30] [469/594], loss: 0.0886, accyrasy: 0.9674, f1: 0.934  f1_dev: 0.8205\n",
      "[16/30] [474/594], loss: 0.0868, accyrasy: 0.9665, f1: 0.9372  f1_dev: 0.7612\n",
      "[16/30] [479/594], loss: 0.0872, accyrasy: 0.9677, f1: 0.9363  f1_dev: 0.8674\n",
      "[16/30] [484/594], loss: 0.0889, accyrasy: 0.9679, f1: 0.9406  f1_dev: 0.8672\n",
      "[16/30] [489/594], loss: 0.0888, accyrasy: 0.9701, f1: 0.9432  f1_dev: 0.7013\n",
      "[16/30] [494/594], loss: 0.0872, accyrasy: 0.9649, f1: 0.9398  f1_dev: 0.81\n",
      "[16/30] [499/594], loss: 0.0895, accyrasy: 0.9625, f1: 0.9263  f1_dev: 0.8547\n",
      "[16/30] [504/594], loss: 0.0913, accyrasy: 0.9653, f1: 0.9306  f1_dev: 0.7901\n",
      "[16/30] [509/594], loss: 0.0867, accyrasy: 0.968, f1: 0.9393  f1_dev: 0.7384\n",
      "[16/30] [514/594], loss: 0.0911, accyrasy: 0.9631, f1: 0.9333  f1_dev: 0.799\n",
      "[16/30] [519/594], loss: 0.0874, accyrasy: 0.9688, f1: 0.9388  f1_dev: 0.839\n",
      "[16/30] [524/594], loss: 0.0856, accyrasy: 0.9665, f1: 0.9345  f1_dev: 0.8268\n",
      "[16/30] [529/594], loss: 0.0913, accyrasy: 0.9665, f1: 0.9369  f1_dev: 0.7137\n",
      "[16/30] [534/594], loss: 0.0857, accyrasy: 0.9676, f1: 0.9373  f1_dev: 0.7647\n",
      "[16/30] [539/594], loss: 0.0869, accyrasy: 0.9688, f1: 0.9421  f1_dev: 0.8786\n",
      "[16/30] [544/594], loss: 0.0846, accyrasy: 0.9649, f1: 0.9321  f1_dev: 0.8148\n",
      "[16/30] [549/594], loss: 0.0835, accyrasy: 0.973, f1: 0.9472  f1_dev: 0.789\n",
      "[16/30] [554/594], loss: 0.0885, accyrasy: 0.9684, f1: 0.9421  f1_dev: 0.7905\n",
      "[16/30] [559/594], loss: 0.0918, accyrasy: 0.9614, f1: 0.921  f1_dev: 0.8606\n",
      "[16/30] [564/594], loss: 0.0896, accyrasy: 0.9634, f1: 0.9312  f1_dev: 0.8777\n",
      "[16/30] [569/594], loss: 0.0917, accyrasy: 0.964, f1: 0.9335  f1_dev: 0.8797\n",
      "[16/30] [574/594], loss: 0.0917, accyrasy: 0.9651, f1: 0.9338  f1_dev: 0.8701\n",
      "[16/30] [579/594], loss: 0.0868, accyrasy: 0.9669, f1: 0.9339  f1_dev: 0.8427\n",
      "[16/30] [584/594], loss: 0.0908, accyrasy: 0.9699, f1: 0.9437  f1_dev: 0.8743\n",
      "[16/30] [589/594], loss: 0.0847, accyrasy: 0.9688, f1: 0.941  f1_dev: 0.8746\n",
      "[17/30] [4/594], loss: 0.0917, accyrasy: 0.963, f1: 0.9256  f1_dev: 0.8663\n",
      "[17/30] [9/594], loss: 0.0894, accyrasy: 0.967, f1: 0.9354  f1_dev: 0.8665\n",
      "[17/30] [14/594], loss: 0.0832, accyrasy: 0.9681, f1: 0.937  f1_dev: 0.773\n",
      "[17/30] [19/594], loss: 0.0878, accyrasy: 0.9651, f1: 0.9304  f1_dev: 0.7739\n",
      "[17/30] [24/594], loss: 0.0893, accyrasy: 0.9643, f1: 0.9361  f1_dev: 0.8533\n",
      "[17/30] [29/594], loss: 0.0868, accyrasy: 0.9691, f1: 0.9382  f1_dev: 0.8464\n",
      "[17/30] [34/594], loss: 0.0901, accyrasy: 0.9639, f1: 0.9264  f1_dev: 0.768\n",
      "[17/30] [39/594], loss: 0.0938, accyrasy: 0.9643, f1: 0.9267  f1_dev: 0.7345\n",
      "[17/30] [44/594], loss: 0.0894, accyrasy: 0.9659, f1: 0.9372  f1_dev: 0.7915\n",
      "[17/30] [49/594], loss: 0.0877, accyrasy: 0.9659, f1: 0.931  f1_dev: 0.8718\n",
      "[17/30] [54/594], loss: 0.0889, accyrasy: 0.9643, f1: 0.9238  f1_dev: 0.8452\n",
      "[17/30] [59/594], loss: 0.0843, accyrasy: 0.9691, f1: 0.9402  f1_dev: 0.762\n",
      "[17/30] [64/594], loss: 0.0856, accyrasy: 0.9663, f1: 0.9356  f1_dev: 0.865\n",
      "[17/30] [69/594], loss: 0.0865, accyrasy: 0.9674, f1: 0.9376  f1_dev: 0.7908\n",
      "[17/30] [74/594], loss: 0.0844, accyrasy: 0.9649, f1: 0.9314  f1_dev: 0.8454\n",
      "[17/30] [79/594], loss: 0.0926, accyrasy: 0.9646, f1: 0.9325  f1_dev: 0.8594\n",
      "[17/30] [84/594], loss: 0.0923, accyrasy: 0.9631, f1: 0.9322  f1_dev: 0.8465\n",
      "[17/30] [89/594], loss: 0.0902, accyrasy: 0.9651, f1: 0.9377  f1_dev: 0.8544\n",
      "[17/30] [94/594], loss: 0.0916, accyrasy: 0.963, f1: 0.9253  f1_dev: 0.8715\n",
      "[17/30] [99/594], loss: 0.0875, accyrasy: 0.9651, f1: 0.9352  f1_dev: 0.8752\n",
      "[17/30] [104/594], loss: 0.0871, accyrasy: 0.9669, f1: 0.9324  f1_dev: 0.8823\n",
      "[17/30] [109/594], loss: 0.0908, accyrasy: 0.9681, f1: 0.938  f1_dev: 0.847\n",
      "[17/30] [114/594], loss: 0.0845, accyrasy: 0.9665, f1: 0.9325  f1_dev: 0.8463\n",
      "[17/30] [119/594], loss: 0.0879, accyrasy: 0.9665, f1: 0.9357  f1_dev: 0.8474\n",
      "[17/30] [124/594], loss: 0.0837, accyrasy: 0.9699, f1: 0.9402  f1_dev: 0.874\n",
      "[17/30] [129/594], loss: 0.0887, accyrasy: 0.9643, f1: 0.9314  f1_dev: 0.8858\n",
      "[17/30] [134/594], loss: 0.0922, accyrasy: 0.9657, f1: 0.9301  f1_dev: 0.786\n",
      "[17/30] [139/594], loss: 0.0814, accyrasy: 0.9695, f1: 0.9354  f1_dev: 0.8462\n",
      "[17/30] [144/594], loss: 0.0867, accyrasy: 0.9674, f1: 0.9331  f1_dev: 0.8612\n",
      "[17/30] [149/594], loss: 0.0896, accyrasy: 0.9664, f1: 0.9331  f1_dev: 0.8549\n",
      "[17/30] [154/594], loss: 0.0908, accyrasy: 0.9683, f1: 0.9408  f1_dev: 0.7548\n",
      "[17/30] [159/594], loss: 0.0869, accyrasy: 0.9686, f1: 0.9375  f1_dev: 0.8129\n",
      "[17/30] [164/594], loss: 0.0841, accyrasy: 0.9685, f1: 0.9371  f1_dev: 0.879\n",
      "[17/30] [169/594], loss: 0.0846, accyrasy: 0.9681, f1: 0.939  f1_dev: 0.7834\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[17/30] [174/594], loss: 0.0848, accyrasy: 0.9646, f1: 0.9308  f1_dev: 0.76\n",
      "[17/30] [179/594], loss: 0.0952, accyrasy: 0.9664, f1: 0.9379  f1_dev: 0.8863\n",
      "[17/30] [184/594], loss: 0.0916, accyrasy: 0.9651, f1: 0.9341  f1_dev: 0.8128\n",
      "[17/30] [189/594], loss: 0.0994, accyrasy: 0.9666, f1: 0.9381  f1_dev: 0.6795\n",
      "[17/30] [194/594], loss: 0.095, accyrasy: 0.9667, f1: 0.9359  f1_dev: 0.7043\n",
      "[17/30] [199/594], loss: 0.0911, accyrasy: 0.9689, f1: 0.9379  f1_dev: 0.8426\n",
      "[17/30] [204/594], loss: 0.0896, accyrasy: 0.9675, f1: 0.9348  f1_dev: 0.7389\n",
      "[17/30] [209/594], loss: 0.0877, accyrasy: 0.9674, f1: 0.9286  f1_dev: 0.721\n",
      "[17/30] [214/594], loss: 0.0875, accyrasy: 0.9674, f1: 0.9338  f1_dev: 0.849\n",
      "[17/30] [219/594], loss: 0.0865, accyrasy: 0.966, f1: 0.9324  f1_dev: 0.8218\n",
      "[17/30] [224/594], loss: 0.0854, accyrasy: 0.9631, f1: 0.9347  f1_dev: 0.8096\n",
      "[17/30] [229/594], loss: 0.0873, accyrasy: 0.9684, f1: 0.934  f1_dev: 0.8773\n",
      "[17/30] [234/594], loss: 0.0875, accyrasy: 0.9683, f1: 0.9348  f1_dev: 0.8711\n",
      "[17/30] [239/594], loss: 0.0967, accyrasy: 0.9584, f1: 0.9238  f1_dev: 0.8707\n",
      "[17/30] [244/594], loss: 0.0959, accyrasy: 0.9643, f1: 0.9319  f1_dev: 0.8578\n",
      "[17/30] [249/594], loss: 0.0902, accyrasy: 0.9674, f1: 0.9329  f1_dev: 0.8773\n",
      "[17/30] [254/594], loss: 0.0927, accyrasy: 0.9671, f1: 0.9253  f1_dev: 0.8635\n",
      "[17/30] [259/594], loss: 0.0919, accyrasy: 0.9603, f1: 0.9194  f1_dev: 0.7978\n",
      "[17/30] [264/594], loss: 0.0915, accyrasy: 0.9586, f1: 0.9274  f1_dev: 0.7941\n",
      "[17/30] [269/594], loss: 0.0972, accyrasy: 0.9623, f1: 0.9226  f1_dev: 0.8976\n",
      "[17/30] [274/594], loss: 0.092, accyrasy: 0.9659, f1: 0.9282  f1_dev: 0.7946\n",
      "[17/30] [279/594], loss: 0.0903, accyrasy: 0.9659, f1: 0.9337  f1_dev: 0.777\n",
      "[17/30] [284/594], loss: 0.0901, accyrasy: 0.964, f1: 0.9331  f1_dev: 0.7626\n",
      "[17/30] [289/594], loss: 0.0885, accyrasy: 0.966, f1: 0.9309  f1_dev: 0.8643\n",
      "[17/30] [294/594], loss: 0.087, accyrasy: 0.9681, f1: 0.9395  f1_dev: 0.8634\n",
      "[17/30] [299/594], loss: 0.0916, accyrasy: 0.9649, f1: 0.9322  f1_dev: 0.8045\n",
      "[17/30] [304/594], loss: 0.0858, accyrasy: 0.9655, f1: 0.9325  f1_dev: 0.8002\n",
      "[17/30] [309/594], loss: 0.087, accyrasy: 0.9695, f1: 0.939  f1_dev: 0.8786\n",
      "[17/30] [314/594], loss: 0.0829, accyrasy: 0.9688, f1: 0.9314  f1_dev: 0.8748\n",
      "[17/30] [319/594], loss: 0.0832, accyrasy: 0.9663, f1: 0.9317  f1_dev: 0.8823\n",
      "[17/30] [324/594], loss: 0.0935, accyrasy: 0.9643, f1: 0.9302  f1_dev: 0.8626\n",
      "[17/30] [329/594], loss: 0.0872, accyrasy: 0.9674, f1: 0.9375  f1_dev: 0.7303\n",
      "[17/30] [334/594], loss: 0.0861, accyrasy: 0.9685, f1: 0.9395  f1_dev: 0.7208\n",
      "[17/30] [339/594], loss: 0.0912, accyrasy: 0.9649, f1: 0.9341  f1_dev: 0.7936\n",
      "[17/30] [344/594], loss: 0.0866, accyrasy: 0.9692, f1: 0.9421  f1_dev: 0.877\n",
      "[17/30] [349/594], loss: 0.0873, accyrasy: 0.966, f1: 0.9327  f1_dev: 0.7822\n",
      "[17/30] [354/594], loss: 0.0866, accyrasy: 0.9653, f1: 0.9328  f1_dev: 0.8691\n",
      "[17/30] [359/594], loss: 0.0875, accyrasy: 0.9694, f1: 0.9432  f1_dev: 0.8822\n",
      "[17/30] [364/594], loss: 0.0863, accyrasy: 0.9669, f1: 0.9325  f1_dev: 0.8816\n",
      "[17/30] [369/594], loss: 0.0866, accyrasy: 0.9644, f1: 0.9318  f1_dev: 0.8601\n",
      "[17/30] [374/594], loss: 0.0874, accyrasy: 0.97, f1: 0.9439  f1_dev: 0.851\n",
      "[17/30] [379/594], loss: 0.0917, accyrasy: 0.9629, f1: 0.9267  f1_dev: 0.7804\n",
      "[17/30] [384/594], loss: 0.0948, accyrasy: 0.9626, f1: 0.9157  f1_dev: 0.7979\n",
      "[17/30] [389/594], loss: 0.089, accyrasy: 0.9657, f1: 0.9402  f1_dev: 0.8573\n",
      "[17/30] [394/594], loss: 0.0883, accyrasy: 0.9684, f1: 0.9361  f1_dev: 0.8097\n",
      "[17/30] [399/594], loss: 0.091, accyrasy: 0.9643, f1: 0.9343  f1_dev: 0.7418\n",
      "[17/30] [404/594], loss: 0.0883, accyrasy: 0.9641, f1: 0.9303  f1_dev: 0.8807\n",
      "[17/30] [409/594], loss: 0.0868, accyrasy: 0.9704, f1: 0.9353  f1_dev: 0.8756\n",
      "[17/30] [414/594], loss: 0.0868, accyrasy: 0.9677, f1: 0.9372  f1_dev: 0.7854\n",
      "[17/30] [419/594], loss: 0.0856, accyrasy: 0.9654, f1: 0.9329  f1_dev: 0.852\n",
      "[17/30] [424/594], loss: 0.09, accyrasy: 0.9679, f1: 0.9381  f1_dev: 0.8701\n",
      "[17/30] [429/594], loss: 0.0921, accyrasy: 0.9661, f1: 0.9293  f1_dev: 0.8685\n",
      "[17/30] [434/594], loss: 0.0866, accyrasy: 0.9705, f1: 0.9475  f1_dev: 0.8784\n",
      "[17/30] [439/594], loss: 0.0854, accyrasy: 0.9634, f1: 0.9282  f1_dev: 0.9003\n",
      "[17/30] [444/594], loss: 0.0824, accyrasy: 0.9711, f1: 0.9466  f1_dev: 0.8834\n",
      "[17/30] [449/594], loss: 0.0845, accyrasy: 0.9674, f1: 0.9385  f1_dev: 0.8917\n",
      "[17/30] [454/594], loss: 0.0856, accyrasy: 0.9679, f1: 0.9385  f1_dev: 0.746\n",
      "[17/30] [459/594], loss: 0.0834, accyrasy: 0.9663, f1: 0.9367  f1_dev: 0.736\n",
      "[17/30] [464/594], loss: 0.0888, accyrasy: 0.9669, f1: 0.939  f1_dev: 0.8484\n",
      "[17/30] [469/594], loss: 0.0884, accyrasy: 0.967, f1: 0.9278  f1_dev: 0.6918\n",
      "[17/30] [474/594], loss: 0.0933, accyrasy: 0.9643, f1: 0.9343  f1_dev: 0.8641\n",
      "[17/30] [479/594], loss: 0.0874, accyrasy: 0.9679, f1: 0.938  f1_dev: 0.7393\n",
      "[17/30] [484/594], loss: 0.0859, accyrasy: 0.9686, f1: 0.9387  f1_dev: 0.8011\n",
      "[17/30] [489/594], loss: 0.0843, accyrasy: 0.971, f1: 0.9434  f1_dev: 0.7976\n",
      "[17/30] [494/594], loss: 0.0852, accyrasy: 0.966, f1: 0.932  f1_dev: 0.8485\n",
      "[17/30] [499/594], loss: 0.0932, accyrasy: 0.9661, f1: 0.9348  f1_dev: 0.8795\n",
      "[17/30] [504/594], loss: 0.0912, accyrasy: 0.9671, f1: 0.9295  f1_dev: 0.8137\n",
      "[17/30] [509/594], loss: 0.0934, accyrasy: 0.9637, f1: 0.9332  f1_dev: 0.7901\n",
      "[17/30] [514/594], loss: 0.0883, accyrasy: 0.9641, f1: 0.9256  f1_dev: 0.8809\n",
      "[17/30] [519/594], loss: 0.094, accyrasy: 0.964, f1: 0.9366  f1_dev: 0.8789\n",
      "[17/30] [524/594], loss: 0.0883, accyrasy: 0.9683, f1: 0.9386  f1_dev: 0.8749\n",
      "[17/30] [529/594], loss: 0.0913, accyrasy: 0.963, f1: 0.9303  f1_dev: 0.761\n",
      "[17/30] [534/594], loss: 0.0867, accyrasy: 0.9679, f1: 0.9341  f1_dev: 0.8586\n",
      "[17/30] [539/594], loss: 0.0881, accyrasy: 0.968, f1: 0.932  f1_dev: 0.7844\n",
      "[17/30] [544/594], loss: 0.0855, accyrasy: 0.9673, f1: 0.9277  f1_dev: 0.7827\n",
      "[17/30] [549/594], loss: 0.0858, accyrasy: 0.9644, f1: 0.9245  f1_dev: 0.8693\n",
      "[17/30] [554/594], loss: 0.0916, accyrasy: 0.9674, f1: 0.9329  f1_dev: 0.8133\n",
      "[17/30] [559/594], loss: 0.0884, accyrasy: 0.9667, f1: 0.9343  f1_dev: 0.8173\n",
      "[17/30] [564/594], loss: 0.0852, accyrasy: 0.9702, f1: 0.9454  f1_dev: 0.8857\n",
      "[17/30] [569/594], loss: 0.0843, accyrasy: 0.9669, f1: 0.9312  f1_dev: 0.7859\n",
      "[17/30] [574/594], loss: 0.0925, accyrasy: 0.9631, f1: 0.9249  f1_dev: 0.8779\n",
      "[17/30] [579/594], loss: 0.0946, accyrasy: 0.962, f1: 0.9259  f1_dev: 0.8648\n",
      "[17/30] [584/594], loss: 0.0902, accyrasy: 0.9639, f1: 0.9281  f1_dev: 0.864\n",
      "[17/30] [589/594], loss: 0.0876, accyrasy: 0.9669, f1: 0.9275  f1_dev: 0.8706\n",
      "[18/30] [4/594], loss: 0.0925, accyrasy: 0.9645, f1: 0.9296  f1_dev: 0.7596\n",
      "[18/30] [9/594], loss: 0.0884, accyrasy: 0.9631, f1: 0.9311  f1_dev: 0.8103\n",
      "[18/30] [14/594], loss: 0.0856, accyrasy: 0.966, f1: 0.9361  f1_dev: 0.8719\n",
      "[18/30] [19/594], loss: 0.0859, accyrasy: 0.966, f1: 0.9361  f1_dev: 0.8778\n",
      "[18/30] [24/594], loss: 0.088, accyrasy: 0.9674, f1: 0.9384  f1_dev: 0.8797\n",
      "[18/30] [29/594], loss: 0.089, accyrasy: 0.967, f1: 0.9331  f1_dev: 0.8777\n",
      "[18/30] [34/594], loss: 0.0888, accyrasy: 0.9673, f1: 0.9348  f1_dev: 0.8514\n",
      "[18/30] [39/594], loss: 0.085, accyrasy: 0.967, f1: 0.9291  f1_dev: 0.8691\n",
      "[18/30] [44/594], loss: 0.0841, accyrasy: 0.9683, f1: 0.938  f1_dev: 0.8785\n",
      "[18/30] [49/594], loss: 0.0871, accyrasy: 0.9677, f1: 0.9344  f1_dev: 0.8016\n",
      "[18/30] [54/594], loss: 0.0868, accyrasy: 0.9644, f1: 0.9286  f1_dev: 0.8474\n",
      "[18/30] [59/594], loss: 0.088, accyrasy: 0.9667, f1: 0.9284  f1_dev: 0.863\n",
      "[18/30] [64/594], loss: 0.085, accyrasy: 0.9681, f1: 0.9413  f1_dev: 0.7504\n",
      "[18/30] [69/594], loss: 0.0882, accyrasy: 0.9695, f1: 0.9429  f1_dev: 0.892\n",
      "[18/30] [74/594], loss: 0.0862, accyrasy: 0.9646, f1: 0.9209  f1_dev: 0.8623\n",
      "[18/30] [79/594], loss: 0.0865, accyrasy: 0.9664, f1: 0.9333  f1_dev: 0.8696\n",
      "[18/30] [84/594], loss: 0.0913, accyrasy: 0.9635, f1: 0.9269  f1_dev: 0.8985\n",
      "[18/30] [89/594], loss: 0.0918, accyrasy: 0.9685, f1: 0.9429  f1_dev: 0.8985\n",
      "[18/30] [94/594], loss: 0.09, accyrasy: 0.9651, f1: 0.934  f1_dev: 0.7444\n",
      "[18/30] [99/594], loss: 0.0852, accyrasy: 0.9649, f1: 0.9216  f1_dev: 0.8009\n",
      "[18/30] [104/594], loss: 0.0924, accyrasy: 0.967, f1: 0.9348  f1_dev: 0.8626\n",
      "[18/30] [109/594], loss: 0.0928, accyrasy: 0.9665, f1: 0.9341  f1_dev: 0.809\n",
      "[18/30] [114/594], loss: 0.0896, accyrasy: 0.9645, f1: 0.9319  f1_dev: 0.7975\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[18/30] [119/594], loss: 0.0865, accyrasy: 0.968, f1: 0.9376  f1_dev: 0.8688\n",
      "[18/30] [124/594], loss: 0.0851, accyrasy: 0.9667, f1: 0.9384  f1_dev: 0.7993\n",
      "[18/30] [129/594], loss: 0.086, accyrasy: 0.9673, f1: 0.931  f1_dev: 0.8897\n",
      "[18/30] [134/594], loss: 0.0877, accyrasy: 0.9671, f1: 0.9364  f1_dev: 0.7903\n",
      "[18/30] [139/594], loss: 0.086, accyrasy: 0.9639, f1: 0.9314  f1_dev: 0.8115\n",
      "[18/30] [144/594], loss: 0.091, accyrasy: 0.9653, f1: 0.9314  f1_dev: 0.8106\n",
      "[18/30] [149/594], loss: 0.0895, accyrasy: 0.9655, f1: 0.937  f1_dev: 0.822\n",
      "[18/30] [154/594], loss: 0.088, accyrasy: 0.9631, f1: 0.9294  f1_dev: 0.8111\n",
      "[18/30] [159/594], loss: 0.0865, accyrasy: 0.9691, f1: 0.9458  f1_dev: 0.8029\n",
      "[18/30] [164/594], loss: 0.0881, accyrasy: 0.9683, f1: 0.932  f1_dev: 0.8014\n",
      "[18/30] [169/594], loss: 0.0874, accyrasy: 0.9661, f1: 0.9347  f1_dev: 0.7556\n",
      "[18/30] [174/594], loss: 0.0829, accyrasy: 0.9716, f1: 0.9465  f1_dev: 0.7787\n",
      "[18/30] [179/594], loss: 0.0858, accyrasy: 0.9695, f1: 0.9411  f1_dev: 0.8602\n",
      "[18/30] [184/594], loss: 0.0897, accyrasy: 0.9647, f1: 0.9319  f1_dev: 0.8649\n",
      "[18/30] [189/594], loss: 0.0916, accyrasy: 0.9615, f1: 0.9176  f1_dev: 0.884\n",
      "[18/30] [194/594], loss: 0.0931, accyrasy: 0.9644, f1: 0.9322  f1_dev: 0.802\n",
      "[18/30] [199/594], loss: 0.0853, accyrasy: 0.9667, f1: 0.9324  f1_dev: 0.7886\n",
      "[18/30] [204/594], loss: 0.0923, accyrasy: 0.9679, f1: 0.9346  f1_dev: 0.8057\n",
      "[18/30] [209/594], loss: 0.0858, accyrasy: 0.969, f1: 0.9373  f1_dev: 0.7876\n",
      "[18/30] [214/594], loss: 0.0873, accyrasy: 0.966, f1: 0.9305  f1_dev: 0.8605\n",
      "[18/30] [219/594], loss: 0.0858, accyrasy: 0.9679, f1: 0.9304  f1_dev: 0.808\n",
      "[18/30] [224/594], loss: 0.0869, accyrasy: 0.967, f1: 0.9392  f1_dev: 0.7916\n",
      "[18/30] [229/594], loss: 0.0858, accyrasy: 0.9675, f1: 0.9336  f1_dev: 0.7723\n",
      "[18/30] [234/594], loss: 0.0882, accyrasy: 0.9685, f1: 0.9329  f1_dev: 0.7612\n",
      "[18/30] [239/594], loss: 0.0869, accyrasy: 0.9695, f1: 0.9476  f1_dev: 0.8112\n",
      "[18/30] [244/594], loss: 0.0915, accyrasy: 0.9675, f1: 0.94  f1_dev: 0.8626\n",
      "[18/30] [249/594], loss: 0.0908, accyrasy: 0.9683, f1: 0.9363  f1_dev: 0.8068\n",
      "[18/30] [254/594], loss: 0.0937, accyrasy: 0.967, f1: 0.9307  f1_dev: 0.7414\n",
      "[18/30] [259/594], loss: 0.0935, accyrasy: 0.9646, f1: 0.9287  f1_dev: 0.7245\n",
      "[18/30] [264/594], loss: 0.0914, accyrasy: 0.964, f1: 0.927  f1_dev: 0.7244\n",
      "[18/30] [269/594], loss: 0.0889, accyrasy: 0.9661, f1: 0.9337  f1_dev: 0.8487\n",
      "[18/30] [274/594], loss: 0.0895, accyrasy: 0.9639, f1: 0.9311  f1_dev: 0.8556\n",
      "[18/30] [279/594], loss: 0.085, accyrasy: 0.9643, f1: 0.9186  f1_dev: 0.777\n",
      "[18/30] [284/594], loss: 0.09, accyrasy: 0.9679, f1: 0.9326  f1_dev: 0.776\n",
      "[18/30] [289/594], loss: 0.0876, accyrasy: 0.9663, f1: 0.9296  f1_dev: 0.8612\n",
      "[18/30] [294/594], loss: 0.0927, accyrasy: 0.9646, f1: 0.9255  f1_dev: 0.8546\n",
      "[18/30] [299/594], loss: 0.084, accyrasy: 0.9699, f1: 0.9416  f1_dev: 0.863\n",
      "[18/30] [304/594], loss: 0.0884, accyrasy: 0.9679, f1: 0.9343  f1_dev: 0.7809\n",
      "[18/30] [309/594], loss: 0.0826, accyrasy: 0.9641, f1: 0.928  f1_dev: 0.7505\n",
      "[18/30] [314/594], loss: 0.0858, accyrasy: 0.9661, f1: 0.9351  f1_dev: 0.7907\n",
      "[18/30] [319/594], loss: 0.0849, accyrasy: 0.9694, f1: 0.9378  f1_dev: 0.7922\n",
      "[18/30] [324/594], loss: 0.0841, accyrasy: 0.9681, f1: 0.9417  f1_dev: 0.884\n",
      "[18/30] [329/594], loss: 0.0852, accyrasy: 0.9667, f1: 0.9318  f1_dev: 0.8859\n",
      "[18/30] [334/594], loss: 0.0819, accyrasy: 0.968, f1: 0.9383  f1_dev: 0.8606\n",
      "[18/30] [339/594], loss: 0.0917, accyrasy: 0.9665, f1: 0.9292  f1_dev: 0.8633\n",
      "[18/30] [344/594], loss: 0.0837, accyrasy: 0.969, f1: 0.9396  f1_dev: 0.8622\n",
      "[18/30] [349/594], loss: 0.089, accyrasy: 0.9683, f1: 0.9426  f1_dev: 0.8031\n",
      "[18/30] [354/594], loss: 0.0853, accyrasy: 0.9686, f1: 0.9382  f1_dev: 0.8056\n",
      "[18/30] [359/594], loss: 0.0859, accyrasy: 0.9673, f1: 0.9367  f1_dev: 0.7927\n",
      "[18/30] [364/594], loss: 0.0829, accyrasy: 0.9708, f1: 0.9399  f1_dev: 0.8769\n",
      "[18/30] [369/594], loss: 0.0874, accyrasy: 0.9661, f1: 0.9268  f1_dev: 0.7804\n",
      "[18/30] [374/594], loss: 0.087, accyrasy: 0.9698, f1: 0.9397  f1_dev: 0.7854\n",
      "[18/30] [379/594], loss: 0.0875, accyrasy: 0.9619, f1: 0.9237  f1_dev: 0.7518\n",
      "[18/30] [384/594], loss: 0.0931, accyrasy: 0.9666, f1: 0.9351  f1_dev: 0.8618\n",
      "[18/30] [389/594], loss: 0.096, accyrasy: 0.9636, f1: 0.9315  f1_dev: 0.76\n",
      "[18/30] [394/594], loss: 0.0953, accyrasy: 0.9688, f1: 0.9356  f1_dev: 0.7575\n",
      "[18/30] [399/594], loss: 0.0892, accyrasy: 0.9684, f1: 0.9409  f1_dev: 0.8147\n",
      "[18/30] [404/594], loss: 0.0927, accyrasy: 0.9644, f1: 0.9297  f1_dev: 0.7549\n",
      "[18/30] [409/594], loss: 0.0844, accyrasy: 0.9689, f1: 0.9336  f1_dev: 0.727\n",
      "[18/30] [414/594], loss: 0.0855, accyrasy: 0.9709, f1: 0.9427  f1_dev: 0.78\n",
      "[18/30] [419/594], loss: 0.0844, accyrasy: 0.9659, f1: 0.9313  f1_dev: 0.8782\n",
      "[18/30] [424/594], loss: 0.0863, accyrasy: 0.9659, f1: 0.9324  f1_dev: 0.8807\n",
      "[18/30] [429/594], loss: 0.0868, accyrasy: 0.9683, f1: 0.938  f1_dev: 0.8005\n",
      "[18/30] [434/594], loss: 0.0932, accyrasy: 0.9651, f1: 0.9329  f1_dev: 0.7939\n",
      "[18/30] [439/594], loss: 0.0877, accyrasy: 0.9653, f1: 0.9308  f1_dev: 0.8032\n",
      "[18/30] [444/594], loss: 0.0915, accyrasy: 0.9645, f1: 0.9339  f1_dev: 0.7992\n",
      "[18/30] [449/594], loss: 0.0885, accyrasy: 0.9671, f1: 0.9354  f1_dev: 0.8345\n",
      "[18/30] [454/594], loss: 0.0936, accyrasy: 0.9664, f1: 0.9279  f1_dev: 0.8335\n",
      "[18/30] [459/594], loss: 0.0974, accyrasy: 0.9606, f1: 0.9112  f1_dev: 0.8588\n",
      "[18/30] [464/594], loss: 0.0918, accyrasy: 0.9694, f1: 0.9404  f1_dev: 0.7372\n",
      "[18/30] [469/594], loss: 0.0864, accyrasy: 0.9664, f1: 0.9373  f1_dev: 0.8049\n",
      "[18/30] [474/594], loss: 0.0881, accyrasy: 0.9689, f1: 0.942  f1_dev: 0.8979\n",
      "[18/30] [479/594], loss: 0.0865, accyrasy: 0.9656, f1: 0.9356  f1_dev: 0.8844\n",
      "[18/30] [484/594], loss: 0.0884, accyrasy: 0.9656, f1: 0.9322  f1_dev: 0.7849\n",
      "[18/30] [489/594], loss: 0.091, accyrasy: 0.9663, f1: 0.9353  f1_dev: 0.8765\n",
      "[18/30] [494/594], loss: 0.0871, accyrasy: 0.9702, f1: 0.9395  f1_dev: 0.7571\n",
      "[18/30] [499/594], loss: 0.0886, accyrasy: 0.9671, f1: 0.9304  f1_dev: 0.8342\n",
      "[18/30] [504/594], loss: 0.0947, accyrasy: 0.9654, f1: 0.9328  f1_dev: 0.777\n",
      "[18/30] [509/594], loss: 0.0889, accyrasy: 0.9688, f1: 0.9388  f1_dev: 0.8534\n",
      "[18/30] [514/594], loss: 0.0862, accyrasy: 0.9649, f1: 0.9295  f1_dev: 0.7696\n",
      "[18/30] [519/594], loss: 0.0876, accyrasy: 0.9679, f1: 0.9373  f1_dev: 0.8431\n",
      "[18/30] [524/594], loss: 0.0877, accyrasy: 0.9692, f1: 0.9399  f1_dev: 0.8642\n",
      "[18/30] [529/594], loss: 0.0918, accyrasy: 0.9651, f1: 0.9318  f1_dev: 0.8593\n",
      "[18/30] [534/594], loss: 0.097, accyrasy: 0.9637, f1: 0.9279  f1_dev: 0.7792\n",
      "[18/30] [539/594], loss: 0.0883, accyrasy: 0.9621, f1: 0.925  f1_dev: 0.8729\n",
      "[18/30] [544/594], loss: 0.0891, accyrasy: 0.9649, f1: 0.9153  f1_dev: 0.7406\n",
      "[18/30] [549/594], loss: 0.0853, accyrasy: 0.9655, f1: 0.9256  f1_dev: 0.8683\n",
      "[18/30] [554/594], loss: 0.0869, accyrasy: 0.9623, f1: 0.9222  f1_dev: 0.7814\n",
      "[18/30] [559/594], loss: 0.0895, accyrasy: 0.9616, f1: 0.9253  f1_dev: 0.7548\n",
      "[18/30] [564/594], loss: 0.0919, accyrasy: 0.9656, f1: 0.9367  f1_dev: 0.7432\n",
      "[18/30] [569/594], loss: 0.0868, accyrasy: 0.9651, f1: 0.9302  f1_dev: 0.8325\n",
      "[18/30] [574/594], loss: 0.0938, accyrasy: 0.9626, f1: 0.9312  f1_dev: 0.8292\n",
      "[18/30] [579/594], loss: 0.0941, accyrasy: 0.9646, f1: 0.9323  f1_dev: 0.7828\n",
      "[18/30] [584/594], loss: 0.0893, accyrasy: 0.9684, f1: 0.9366  f1_dev: 0.6943\n",
      "[18/30] [589/594], loss: 0.09, accyrasy: 0.9667, f1: 0.9315  f1_dev: 0.7385\n",
      "[19/30] [4/594], loss: 0.0859, accyrasy: 0.9663, f1: 0.9281  f1_dev: 0.8589\n",
      "[19/30] [9/594], loss: 0.0888, accyrasy: 0.965, f1: 0.9323  f1_dev: 0.8606\n",
      "[19/30] [14/594], loss: 0.0865, accyrasy: 0.9667, f1: 0.9392  f1_dev: 0.7738\n",
      "[19/30] [19/594], loss: 0.0834, accyrasy: 0.9695, f1: 0.9435  f1_dev: 0.7683\n",
      "[19/30] [24/594], loss: 0.0833, accyrasy: 0.9667, f1: 0.9336  f1_dev: 0.783\n",
      "[19/30] [29/594], loss: 0.0887, accyrasy: 0.9699, f1: 0.9383  f1_dev: 0.74\n",
      "[19/30] [34/594], loss: 0.0871, accyrasy: 0.9691, f1: 0.9373  f1_dev: 0.7355\n",
      "[19/30] [39/594], loss: 0.0871, accyrasy: 0.9712, f1: 0.9419  f1_dev: 0.7743\n",
      "[19/30] [44/594], loss: 0.0886, accyrasy: 0.9633, f1: 0.9273  f1_dev: 0.7512\n",
      "[19/30] [49/594], loss: 0.0884, accyrasy: 0.9654, f1: 0.9241  f1_dev: 0.7402\n",
      "[19/30] [54/594], loss: 0.0893, accyrasy: 0.9641, f1: 0.9283  f1_dev: 0.7719\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[19/30] [59/594], loss: 0.0906, accyrasy: 0.9666, f1: 0.9374  f1_dev: 0.7775\n",
      "[19/30] [64/594], loss: 0.0891, accyrasy: 0.9624, f1: 0.9252  f1_dev: 0.8553\n",
      "[19/30] [69/594], loss: 0.093, accyrasy: 0.9651, f1: 0.9287  f1_dev: 0.8752\n",
      "[19/30] [74/594], loss: 0.0895, accyrasy: 0.9671, f1: 0.9271  f1_dev: 0.7782\n",
      "[19/30] [79/594], loss: 0.0839, accyrasy: 0.9692, f1: 0.9424  f1_dev: 0.8755\n",
      "[19/30] [84/594], loss: 0.0882, accyrasy: 0.9671, f1: 0.9399  f1_dev: 0.8779\n",
      "[19/30] [89/594], loss: 0.0888, accyrasy: 0.9634, f1: 0.9308  f1_dev: 0.8736\n",
      "[19/30] [94/594], loss: 0.0849, accyrasy: 0.9661, f1: 0.9378  f1_dev: 0.7995\n",
      "[19/30] [99/594], loss: 0.0858, accyrasy: 0.9631, f1: 0.9278  f1_dev: 0.7946\n",
      "[19/30] [104/594], loss: 0.0845, accyrasy: 0.9694, f1: 0.9408  f1_dev: 0.7963\n",
      "[19/30] [109/594], loss: 0.0849, accyrasy: 0.9673, f1: 0.9352  f1_dev: 0.8875\n",
      "[19/30] [114/594], loss: 0.0911, accyrasy: 0.9646, f1: 0.9276  f1_dev: 0.8763\n",
      "[19/30] [119/594], loss: 0.0913, accyrasy: 0.9607, f1: 0.9243  f1_dev: 0.8777\n",
      "[19/30] [124/594], loss: 0.0895, accyrasy: 0.9651, f1: 0.9252  f1_dev: 0.8802\n",
      "[19/30] [129/594], loss: 0.0858, accyrasy: 0.9694, f1: 0.939  f1_dev: 0.8685\n",
      "[19/30] [134/594], loss: 0.0888, accyrasy: 0.9685, f1: 0.9416  f1_dev: 0.8751\n",
      "[19/30] [139/594], loss: 0.0942, accyrasy: 0.9627, f1: 0.9313  f1_dev: 0.7732\n",
      "[19/30] [144/594], loss: 0.0897, accyrasy: 0.9667, f1: 0.9391  f1_dev: 0.7949\n",
      "[19/30] [149/594], loss: 0.09, accyrasy: 0.9679, f1: 0.9408  f1_dev: 0.751\n",
      "[19/30] [154/594], loss: 0.0923, accyrasy: 0.958, f1: 0.92  f1_dev: 0.7267\n",
      "[19/30] [159/594], loss: 0.0866, accyrasy: 0.968, f1: 0.9365  f1_dev: 0.8576\n",
      "[19/30] [164/594], loss: 0.0859, accyrasy: 0.9691, f1: 0.9348  f1_dev: 0.8694\n",
      "[19/30] [169/594], loss: 0.0827, accyrasy: 0.9705, f1: 0.943  f1_dev: 0.801\n",
      "[19/30] [174/594], loss: 0.0836, accyrasy: 0.968, f1: 0.9393  f1_dev: 0.7989\n",
      "[19/30] [179/594], loss: 0.0852, accyrasy: 0.969, f1: 0.9373  f1_dev: 0.8733\n",
      "[19/30] [184/594], loss: 0.0904, accyrasy: 0.9649, f1: 0.9313  f1_dev: 0.8618\n",
      "[19/30] [189/594], loss: 0.0899, accyrasy: 0.9676, f1: 0.9332  f1_dev: 0.8787\n",
      "[19/30] [194/594], loss: 0.0903, accyrasy: 0.9576, f1: 0.9191  f1_dev: 0.7835\n",
      "[19/30] [199/594], loss: 0.1023, accyrasy: 0.955, f1: 0.9114  f1_dev: 0.872\n",
      "[19/30] [204/594], loss: 0.0984, accyrasy: 0.9665, f1: 0.9321  f1_dev: 0.8969\n",
      "[19/30] [209/594], loss: 0.092, accyrasy: 0.965, f1: 0.9307  f1_dev: 0.8828\n",
      "[19/30] [214/594], loss: 0.0856, accyrasy: 0.9685, f1: 0.9385  f1_dev: 0.7733\n",
      "[19/30] [219/594], loss: 0.0857, accyrasy: 0.9674, f1: 0.9387  f1_dev: 0.7667\n",
      "[19/30] [224/594], loss: 0.0874, accyrasy: 0.9681, f1: 0.933  f1_dev: 0.724\n",
      "[19/30] [229/594], loss: 0.0894, accyrasy: 0.9675, f1: 0.9356  f1_dev: 0.7772\n",
      "[19/30] [234/594], loss: 0.0933, accyrasy: 0.9626, f1: 0.9222  f1_dev: 0.7432\n",
      "[19/30] [239/594], loss: 0.0862, accyrasy: 0.9683, f1: 0.934  f1_dev: 0.7836\n",
      "[19/30] [244/594], loss: 0.0865, accyrasy: 0.963, f1: 0.9304  f1_dev: 0.8691\n",
      "[19/30] [249/594], loss: 0.0863, accyrasy: 0.9666, f1: 0.9331  f1_dev: 0.8814\n",
      "[19/30] [254/594], loss: 0.0868, accyrasy: 0.9659, f1: 0.929  f1_dev: 0.8671\n",
      "[19/30] [259/594], loss: 0.0831, accyrasy: 0.9676, f1: 0.9379  f1_dev: 0.8624\n",
      "[19/30] [264/594], loss: 0.086, accyrasy: 0.9639, f1: 0.9304  f1_dev: 0.875\n",
      "[19/30] [269/594], loss: 0.0843, accyrasy: 0.9683, f1: 0.9374  f1_dev: 0.882\n",
      "[19/30] [274/594], loss: 0.0887, accyrasy: 0.9645, f1: 0.9336  f1_dev: 0.7699\n",
      "[19/30] [279/594], loss: 0.0864, accyrasy: 0.9676, f1: 0.9404  f1_dev: 0.7638\n",
      "[19/30] [284/594], loss: 0.0961, accyrasy: 0.9651, f1: 0.9321  f1_dev: 0.8548\n",
      "[19/30] [289/594], loss: 0.0886, accyrasy: 0.9679, f1: 0.9362  f1_dev: 0.853\n",
      "[19/30] [294/594], loss: 0.0842, accyrasy: 0.9674, f1: 0.9368  f1_dev: 0.8442\n",
      "[19/30] [299/594], loss: 0.0858, accyrasy: 0.969, f1: 0.9396  f1_dev: 0.8735\n",
      "[19/30] [304/594], loss: 0.0866, accyrasy: 0.9677, f1: 0.9364  f1_dev: 0.8675\n",
      "[19/30] [309/594], loss: 0.0874, accyrasy: 0.9653, f1: 0.9348  f1_dev: 0.7691\n"
     ]
    }
   ],
   "source": [
    "num_epochs = 30\n",
    "monit_res = 5\n",
    "for epoch in range(num_epochs):\n",
    "\n",
    "    running_loss = 0.0\n",
    "    for i, batch in enumerate(dataloader):\n",
    "        model.train()\n",
    "        optimizer.zero_grad()\n",
    "        \n",
    "        input, labels = batch['input'], batch['open_channels']\n",
    "        input = input.to(device).float().view(len(input), 1, ws+1)\n",
    "        labels = labels.to(device).float()     \n",
    "    \n",
    "        output = model(input).reshape(len(input), 11)\n",
    "        \n",
    "        loss = criterion(output, torch.max(labels, 1)[1])\n",
    "        loss.backward()\n",
    "        optimizer.step()        \n",
    "        \n",
    "        running_loss += loss.item()\n",
    "        \n",
    "        \n",
    "        if i%monit_res == monit_res-1:\n",
    "            model.eval()\n",
    "            predicted_idx = torch.max(output, 1)[1]\n",
    "            true_idx = torch.max(labels, 1)[1]\n",
    "            predicted_idx = predicted_idx.cpu().detach()\n",
    "            true_idx = true_idx.cpu().detach()\n",
    "            accurasy = sum(true_idx == predicted_idx).item()/len(true_idx)     \n",
    "            f1 = f1_score(true_idx, predicted_idx, average='macro')\n",
    "            train_accurasy = np.append(train_accurasy, accurasy)\n",
    "            train_f1 = np.append(train_f1, f1)\n",
    "            \n",
    "            t_accurasy = 0\n",
    "            t_f1 = 0\n",
    "            \n",
    "            input, labels = batch_test['input'], batch_test['open_channels']\n",
    "            input = input.to(device).float().view(len(batch_test['input']), 1, ws+1)\n",
    "            labels = labels.to(device).float()\n",
    "            \n",
    "            with torch.no_grad():\n",
    "                output = model(input).reshape(len(input), 11)\n",
    "            predicted_idx = torch.max(output, 1)[1]\n",
    "            true_idx = torch.max(labels, 1)[1]\n",
    "            predicted_idx = predicted_idx.cpu().detach()\n",
    "            true_idx = true_idx.cpu().detach()\n",
    "            t_accurasy = sum(true_idx == predicted_idx).item()/len(true_idx)     \n",
    "            t_f1 = f1_score(true_idx, predicted_idx, average='macro')\n",
    "            test_accurasy = np.append(test_accurasy, t_accurasy)\n",
    "            \n",
    "            test_f1 = np.append(test_f1, t_f1)\n",
    "\n",
    "            \n",
    "            train_losses = np.append(train_losses, running_loss/monit_res)\n",
    "            print('[{}/{}] [{}/{}], loss: {}, accyrasy: {}, f1: {}  f1_dev: {}'.format(epoch,\n",
    "                                                     num_epochs,\n",
    "                                                     i,\n",
    "                                                     len(dataloader),\n",
    "                                                     round(running_loss/monit_res, 4),\n",
    "                                                     round(accurasy, 4),\n",
    "                                                     round(f1, 4),\n",
    "                                                     round(t_f1, 4)\n",
    "                                                    )\n",
    "                 )\n",
    "            running_loss = 0.0\n",
    "#             if t_f1 > 0.892:\n",
    "#                 print(asd)\n",
    "#             if t_f1 > 0.965:\n",
    "#                 print(asd)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_epochs = 10\n",
    "monit_res = 1\n",
    "for epoch in range(num_epochs):\n",
    "\n",
    "    running_loss = 0.0\n",
    "    for i, batch in enumerate(dataloader):\n",
    "        model.train()\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        ################################################################################################\n",
    "        ################################################################################################\n",
    "        labels = batch['label']\n",
    "        \n",
    "        inputs = batch['seq'][labels>-0.5].float().to(device)\n",
    "        labels = labels[labels>-0.5].long().to(device)\n",
    "\n",
    "        outputs = model(inputs)\n",
    "        loss = criterion(outputs, labels)\n",
    "  \n",
    "        optimizer.zero_grad()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        ################################################################################################\n",
    "        ################################################################################################  \n",
    "        \n",
    "        running_loss += loss.item()\n",
    "    \n",
    "        if i%monit_res == monit_res-1:\n",
    "            model.eval()\n",
    "            predicted_idx = torch.max(outputs, 1)[1]\n",
    "            true_idx = labels\n",
    "            predicted_idx = predicted_idx.cpu().detach()\n",
    "            true_idx = true_idx.cpu().detach()\n",
    "            accurasy = sum(true_idx == predicted_idx).item()/len(true_idx)     \n",
    "            f1 = f1_score(true_idx, predicted_idx, average='macro')\n",
    "            train_accurasy = np.append(train_accurasy, accurasy)\n",
    "            train_f1 = np.append(train_f1, f1)\n",
    "#             if f1 > 0.7:\n",
    "#                 print(f1)\n",
    "#                 print(sfsdf)\n",
    "            \n",
    "            train_losses = np.append(train_losses, running_loss/monit_res)\n",
    "            if i==0:\n",
    "                print('[{}/{}] [{}/{}], loss: {}, accyrasy: {}, f1: {}'.format(epoch,\n",
    "                                                         num_epochs,\n",
    "                                                         i,\n",
    "                                                         len(dataloader),\n",
    "                                                         round(running_loss/monit_res, 4),\n",
    "                                                         round(accurasy, 4),\n",
    "                                                         round(f1, 4)\n",
    "                                                        )\n",
    "                     )\n",
    "            running_loss = 0.0\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x504 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "plt.figure(figsize=(10,7))\n",
    "plt.plot(train_losses, label='cost: {}'.format(round(train_losses[-1], 4)))\n",
    "plt.plot(train_accurasy, label='accurasy: {}'.format(accurasy))\n",
    "plt.plot(train_f1, label='F1 score: {}'.format(round(train_f1[-1], 4)))\n",
    "plt.plot(test_accurasy, label='test accurasy: {}'.format(t_accurasy))\n",
    "plt.plot(test_f1, label='test F1 score: {}'.format(round(test_f1[-1], 4)))\n",
    "plt.plot([], [], ' ', label=\"epoch: {},  lr: {},  ws: {},  bs: {}\\n\\n{}\".format(num_epochs, lr, ws, btch_sz, model))\n",
    "plt.title('test #18', fontsize=15, loc='left')\n",
    "plt.legend(fontsize=15, bbox_to_anchor=(1, 1))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(\"./data/train_thrCK.csv\")\n",
    "\n",
    "for c in range(11):\n",
    "    df['s'+str(c)] = np.ones(len(df))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_csv(\"./data/test_thrCK_seqCNN.csv\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "import dataUtils.ionUtils as ionUtils\n",
    "reload(ionUtils);\n",
    "\n",
    "num_classes = 11\n",
    "bNr_train = 6"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(\"./data/train_thrCK.csv\")\n",
    "for c in range(11):\n",
    "    df['s'+str(c)] = np.ones(len(df))\n",
    "\n",
    "# Test the model\n",
    "softmax=nn.Softmax(dim=1)\n",
    "with torch.no_grad():\n",
    "    for i,batch in enumerate(dataloader):\n",
    "#         inputs = batch['seq'].float().to(device)\n",
    "    \n",
    "        inputs = batch['signal'].float().to(device)\n",
    "        inputs = inputs.view(len(inputs), 1, 49)\n",
    "        \n",
    "        indices = batch['index']\n",
    "        outputs = model(inputs)\n",
    "        outputs=softmax(outputs)\n",
    "        outputs = outputs.cpu().detach().numpy()\n",
    "        df.loc[indices,16:16+num_classes] = outputs\n",
    "        \n",
    "#         indices = indices[(outputs.max(axis=1) > 0.9) & (df.iloc[indices,4:15].values.max(axis=1) < 0.4)]\n",
    "#         print(sum(outputs.max(axis=1) > 0.9))\n",
    "#         df.loc[indices,4:15] = 1\n",
    "        \n",
    "#         indices = batch['index']\n",
    "#         indices = indices[(outputs.max(axis=1) < 0.8) & (df.iloc[indices,4:15].values.max(axis=1) > 0.8)]\n",
    "#         print(sum(outputs.max(axis=1) < 0.4) )\n",
    "#         df.loc[indices,16:16+num_classes] = 1\n",
    "        \n",
    "        print('{}/{}'.format(i, len(dataloader)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.loc[:,4:15] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.5 {'thr': '0.500', 'batch': 6, 'acc': '0.973288', 'f1': '0.966793'}\n",
      "0.55 {'thr': '0.550', 'batch': 6, 'acc': '0.974182', 'f1': '0.967798'}\n",
      "0.6000000000000001 {'thr': '0.600', 'batch': 6, 'acc': '0.975030', 'f1': '0.968491'}\n",
      "0.6500000000000001 {'thr': '0.650', 'batch': 6, 'acc': '0.975538', 'f1': '0.968884'}\n",
      "0.7000000000000002 {'thr': '0.700', 'batch': 6, 'acc': '0.975844', 'f1': '0.968825'}\n",
      "0.7500000000000002 {'thr': '0.750', 'batch': 6, 'acc': '0.976054', 'f1': '0.969019'}\n",
      "0.8000000000000003 {'thr': '0.800', 'batch': 6, 'acc': '0.976184', 'f1': '0.968870'}\n",
      "0.8500000000000003 {'thr': '0.850', 'batch': 6, 'acc': '0.976234', 'f1': '0.968877'}\n",
      "0.9000000000000004 {'thr': '0.900', 'batch': 6, 'acc': '0.976274', 'f1': '0.968999'}\n",
      "0.9500000000000004 {'thr': '0.950', 'batch': 6, 'acc': '0.976274', 'f1': '0.968999'}\n",
      "1.0000000000000004 {'thr': '1.000', 'batch': 6, 'acc': '0.976274', 'f1': '0.968999'}\n"
     ]
    }
   ],
   "source": [
    "for j in np.arange(0.5, 1.01, 0.05):\n",
    "    print(j,ionUtils.getFullScores(df, thr=j, BR=[bNr_train])[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "indx = df[df.batch == 6].index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.loc[indx, 16:16+11] = probs_seq"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df[df.batch == 6].iloc[:, 16:16+11]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Format submmit file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "NB = 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_test = pd.read_csv('data/test_thrCK.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "# indx = df_test[df_test.batch == NB].index\n",
    "indx = df_test.index\n",
    "gaus_probs = df_test.iloc[indx, 3:14].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2000000, 11)"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gaus_probs.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2000000, 11)"
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs_seq.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_test.loc[indx, 3:14] = probs_seq*gaus_probs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[9.99999732e-01, 3.96214796e-16, 0.00000000e+00, ...,\n",
       "        0.00000000e+00, 0.00000000e+00, 0.00000000e+00],\n",
       "       [9.99999996e-01, 4.89253477e-17, 0.00000000e+00, ...,\n",
       "        0.00000000e+00, 0.00000000e+00, 0.00000000e+00],\n",
       "       [9.99999639e-01, 1.08038434e-15, 0.00000000e+00, ...,\n",
       "        0.00000000e+00, 0.00000000e+00, 0.00000000e+00],\n",
       "       ...,\n",
       "       [9.99999983e-01, 1.17989895e-19, 1.12297593e-37, ...,\n",
       "        0.00000000e+00, 0.00000000e+00, 0.00000000e+00],\n",
       "       [9.99999997e-01, 9.34892089e-20, 3.00691933e-39, ...,\n",
       "        0.00000000e+00, 0.00000000e+00, 0.00000000e+00],\n",
       "       [9.99999897e-01, 5.33756051e-19, 5.45052451e-36, ...,\n",
       "        0.00000000e+00, 0.00000000e+00, 0.00000000e+00]])"
      ]
     },
     "execution_count": 78,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs_seq*gaus_probs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2000000"
      ]
     },
     "execution_count": 79,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(df_test.iloc[indx, 3:14].values.argmax(axis=1) == (probs_seq*gaus_probs).argmax(axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "prediction = df_test.iloc[indx, 3:14].values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [],
   "source": [
    "prediction = prediction.argmax(axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 0, 0, ..., 0, 0, 0])"
      ]
     },
     "execution_count": 82,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.993609"
      ]
     },
     "execution_count": 83,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submmit = pd.read_csv('submmit.csv')\n",
    "# sum(submmit[submmit.batch == NB].open_channels.values == prediction)/100000\n",
    "sum(submmit.open_channels.values == prediction)/2000000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "submmit.open_channels[submmit.batch == NB] = prediction\n",
    "submmit.to_csv('submmit.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submmit = pd.read_csv('submmit.csv')\n",
    "sum(submmit[submmit.batch == NB].open_channels.values == prediction)/100000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 266,
   "metadata": {},
   "outputs": [],
   "source": [
    "sub_thrCK = pd.read_csv('sub_thrCK_CNN.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 267,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([10, 10, 10, ...,  7,  9, 10])"
      ]
     },
     "execution_count": 267,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 325,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.97355"
      ]
     },
     "execution_count": 325,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(sub_thrCK.iloc[submmit_5.index, 1] == prediction)/len(prediction)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 323,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.97355"
      ]
     },
     "execution_count": 323,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(sub_thrCK.iloc[submmit_5.index, 1] == submmit[submmit.batch == 5].open_channels)/len(prediction)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 273,
   "metadata": {},
   "outputs": [],
   "source": [
    "sub_thrCK = pd.read_csv('sub_thrCK.csv')\n",
    "submmit = pd.read_csv('submmit.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 269,
   "metadata": {},
   "outputs": [],
   "source": [
    "submmit_7 = submmit[submmit.batch == 7].open_channels\n",
    "submmit_5 = submmit[submmit.batch == 5].open_channels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 338,
   "metadata": {},
   "outputs": [],
   "source": [
    "sub_thrCK.iloc[submmit_5.index, 1] = submmit_5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 274,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9676"
      ]
     },
     "execution_count": 274,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(sub_thrCK.loc[submmit_5.index].open_channels == submmit_5)/100000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 340,
   "metadata": {},
   "outputs": [],
   "source": [
    "sub_thrCK.to_csv('sub_thrCK_CNN.csv', float_format='%0.4f', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 217,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>open_channels</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>500.0001</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>500.0002</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>500.0003</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>500.0004</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>500.0005</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1999995</th>\n",
       "      <td>699.9996</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1999996</th>\n",
       "      <td>699.9997</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1999997</th>\n",
       "      <td>699.9998</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1999998</th>\n",
       "      <td>699.9999</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1999999</th>\n",
       "      <td>700.0000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2000000 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             time  open_channels\n",
       "0        500.0001              0\n",
       "1        500.0002              0\n",
       "2        500.0003              0\n",
       "3        500.0004              0\n",
       "4        500.0005              0\n",
       "...           ...            ...\n",
       "1999995  699.9996              0\n",
       "1999996  699.9997              0\n",
       "1999997  699.9998              0\n",
       "1999998  699.9999              0\n",
       "1999999  700.0000              0\n",
       "\n",
       "[2000000 rows x 2 columns]"
      ]
     },
     "execution_count": 217,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub_thrCK"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 213,
   "metadata": {},
   "outputs": [],
   "source": [
    "def save(model, epoch, loss, path_to_save):\n",
    "    torch.save({\n",
    "                'model': model.module.state_dict(),\n",
    "                'epoch': epoch,\n",
    "                'loss': loss,\n",
    "                },\n",
    "                path_to_save)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 214,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "save(model=model, epoch=15, loss=np.array([train_losses,train_accurasy]), path_to_save='test_5_7_CNN.pth' )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 284,
   "metadata": {},
   "outputs": [],
   "source": [
    "checkpoint = torch.load('test_5_7_CNN.pth')\n",
    "model2 = FCC.OpenChannelsClassifier_CNN2(1)\n",
    "model2.load_state_dict(checkpoint['model'])\n",
    "model2.to(device)\n",
    "model2 = torch.nn.DataParallel(model2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "softmax = nn.Softmax(dim=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "probs_seq = np.empty((0,11))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "probs_singl = np.empty((0,11))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0/400]\n",
      "[5/400]\n",
      "[10/400]\n",
      "[15/400]\n",
      "[20/400]\n",
      "[25/400]\n",
      "[30/400]\n",
      "[35/400]\n",
      "[40/400]\n",
      "[45/400]\n",
      "[50/400]\n",
      "[55/400]\n",
      "[60/400]\n",
      "[65/400]\n",
      "[70/400]\n",
      "[75/400]\n",
      "[80/400]\n",
      "[85/400]\n",
      "[90/400]\n",
      "[95/400]\n",
      "[100/400]\n",
      "[105/400]\n",
      "[110/400]\n",
      "[115/400]\n",
      "[120/400]\n",
      "[125/400]\n",
      "[130/400]\n",
      "[135/400]\n",
      "[140/400]\n",
      "[145/400]\n",
      "[150/400]\n",
      "[155/400]\n",
      "[160/400]\n",
      "[165/400]\n",
      "[170/400]\n",
      "[175/400]\n",
      "[180/400]\n",
      "[185/400]\n",
      "[190/400]\n",
      "[195/400]\n",
      "[200/400]\n",
      "[205/400]\n",
      "[210/400]\n",
      "[215/400]\n",
      "[220/400]\n",
      "[225/400]\n",
      "[230/400]\n",
      "[235/400]\n",
      "[240/400]\n",
      "[245/400]\n",
      "[250/400]\n",
      "[255/400]\n",
      "[260/400]\n",
      "[265/400]\n",
      "[270/400]\n",
      "[275/400]\n",
      "[280/400]\n",
      "[285/400]\n",
      "[290/400]\n",
      "[295/400]\n",
      "[300/400]\n",
      "[305/400]\n",
      "[310/400]\n",
      "[315/400]\n",
      "[320/400]\n",
      "[325/400]\n",
      "[330/400]\n",
      "[335/400]\n",
      "[340/400]\n",
      "[345/400]\n",
      "[350/400]\n",
      "[355/400]\n",
      "[360/400]\n",
      "[365/400]\n",
      "[370/400]\n",
      "[375/400]\n",
      "[380/400]\n",
      "[385/400]\n",
      "[390/400]\n",
      "[395/400]\n"
     ]
    }
   ],
   "source": [
    "probs_seq = np.empty((0,11))\n",
    "\n",
    "model.eval()\n",
    "prediction = torch.tensor([]).to(device)\n",
    "for i, batch in enumerate(dataloader_submmit):\n",
    "    \n",
    "    signal = batch['input']\n",
    "    signal = signal.to(device).float().view(len(batch['input']), 1, ws+1)\n",
    "\n",
    "    \n",
    "    with torch.no_grad():\n",
    "        output = model(signal).view(len(signal), 11)\n",
    "  \n",
    "    \n",
    "    predicted_idx = torch.max(output, 1)[1]\n",
    "    prediction = torch.cat((prediction, predicted_idx.float()), 0)\n",
    "    probs_seq = np.concatenate((probs_seq, softmax(output).cpu().detach().numpy()), axis=0)\n",
    "#     probs_singl = np.concatenate((probs_singl, softmax(output).cpu().detach().numpy()), axis=0)\n",
    "    \n",
    "    if i%5 == 0:\n",
    "        print('[{}/{}]'.format(i, len(dataloader_submmit)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 491,
   "metadata": {},
   "outputs": [],
   "source": [
    "probs_singl_argmax = np.argmax(probs_singl, axis=1)\n",
    "probs_seq_argmax   = np.argmax(probs_seq, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 477,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(probs_singl.max(axis=1), bins=25)\n",
    "plt.hist(probs_seq.max(axis=1), bins=25)\n",
    "plt.hist(probs_singl[probs_singl_argmax == data_test.open_channels.values].max(axis=1), bins=25)\n",
    "plt.hist(probs_singl[probs_singl_argmax != data_test.open_channels.values].max(axis=1), bins=25)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/javascript": [
       "/* Put everything inside the global mpl namespace */\n",
       "window.mpl = {};\n",
       "\n",
       "\n",
       "mpl.get_websocket_type = function() {\n",
       "    if (typeof(WebSocket) !== 'undefined') {\n",
       "        return WebSocket;\n",
       "    } else if (typeof(MozWebSocket) !== 'undefined') {\n",
       "        return MozWebSocket;\n",
       "    } else {\n",
       "        alert('Your browser does not have WebSocket support. ' +\n",
       "              'Please try Chrome, Safari or Firefox ≥ 6. ' +\n",
       "              'Firefox 4 and 5 are also supported but you ' +\n",
       "              'have to enable WebSockets in about:config.');\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure = function(figure_id, websocket, ondownload, parent_element) {\n",
       "    this.id = figure_id;\n",
       "\n",
       "    this.ws = websocket;\n",
       "\n",
       "    this.supports_binary = (this.ws.binaryType != undefined);\n",
       "\n",
       "    if (!this.supports_binary) {\n",
       "        var warnings = document.getElementById(\"mpl-warnings\");\n",
       "        if (warnings) {\n",
       "            warnings.style.display = 'block';\n",
       "            warnings.textContent = (\n",
       "                \"This browser does not support binary websocket messages. \" +\n",
       "                    \"Performance may be slow.\");\n",
       "        }\n",
       "    }\n",
       "\n",
       "    this.imageObj = new Image();\n",
       "\n",
       "    this.context = undefined;\n",
       "    this.message = undefined;\n",
       "    this.canvas = undefined;\n",
       "    this.rubberband_canvas = undefined;\n",
       "    this.rubberband_context = undefined;\n",
       "    this.format_dropdown = undefined;\n",
       "\n",
       "    this.image_mode = 'full';\n",
       "\n",
       "    this.root = $('<div/>');\n",
       "    this._root_extra_style(this.root)\n",
       "    this.root.attr('style', 'display: inline-block');\n",
       "\n",
       "    $(parent_element).append(this.root);\n",
       "\n",
       "    this._init_header(this);\n",
       "    this._init_canvas(this);\n",
       "    this._init_toolbar(this);\n",
       "\n",
       "    var fig = this;\n",
       "\n",
       "    this.waiting = false;\n",
       "\n",
       "    this.ws.onopen =  function () {\n",
       "            fig.send_message(\"supports_binary\", {value: fig.supports_binary});\n",
       "            fig.send_message(\"send_image_mode\", {});\n",
       "            if (mpl.ratio != 1) {\n",
       "                fig.send_message(\"set_dpi_ratio\", {'dpi_ratio': mpl.ratio});\n",
       "            }\n",
       "            fig.send_message(\"refresh\", {});\n",
       "        }\n",
       "\n",
       "    this.imageObj.onload = function() {\n",
       "            if (fig.image_mode == 'full') {\n",
       "                // Full images could contain transparency (where diff images\n",
       "                // almost always do), so we need to clear the canvas so that\n",
       "                // there is no ghosting.\n",
       "                fig.context.clearRect(0, 0, fig.canvas.width, fig.canvas.height);\n",
       "            }\n",
       "            fig.context.drawImage(fig.imageObj, 0, 0);\n",
       "        };\n",
       "\n",
       "    this.imageObj.onunload = function() {\n",
       "        fig.ws.close();\n",
       "    }\n",
       "\n",
       "    this.ws.onmessage = this._make_on_message_function(this);\n",
       "\n",
       "    this.ondownload = ondownload;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_header = function() {\n",
       "    var titlebar = $(\n",
       "        '<div class=\"ui-dialog-titlebar ui-widget-header ui-corner-all ' +\n",
       "        'ui-helper-clearfix\"/>');\n",
       "    var titletext = $(\n",
       "        '<div class=\"ui-dialog-title\" style=\"width: 100%; ' +\n",
       "        'text-align: center; padding: 3px;\"/>');\n",
       "    titlebar.append(titletext)\n",
       "    this.root.append(titlebar);\n",
       "    this.header = titletext[0];\n",
       "}\n",
       "\n",
       "\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(canvas_div) {\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_canvas = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var canvas_div = $('<div/>');\n",
       "\n",
       "    canvas_div.attr('style', 'position: relative; clear: both; outline: 0');\n",
       "\n",
       "    function canvas_keyboard_event(event) {\n",
       "        return fig.key_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    canvas_div.keydown('key_press', canvas_keyboard_event);\n",
       "    canvas_div.keyup('key_release', canvas_keyboard_event);\n",
       "    this.canvas_div = canvas_div\n",
       "    this._canvas_extra_style(canvas_div)\n",
       "    this.root.append(canvas_div);\n",
       "\n",
       "    var canvas = $('<canvas/>');\n",
       "    canvas.addClass('mpl-canvas');\n",
       "    canvas.attr('style', \"left: 0; top: 0; z-index: 0; outline: 0\")\n",
       "\n",
       "    this.canvas = canvas[0];\n",
       "    this.context = canvas[0].getContext(\"2d\");\n",
       "\n",
       "    var backingStore = this.context.backingStorePixelRatio ||\n",
       "\tthis.context.webkitBackingStorePixelRatio ||\n",
       "\tthis.context.mozBackingStorePixelRatio ||\n",
       "\tthis.context.msBackingStorePixelRatio ||\n",
       "\tthis.context.oBackingStorePixelRatio ||\n",
       "\tthis.context.backingStorePixelRatio || 1;\n",
       "\n",
       "    mpl.ratio = (window.devicePixelRatio || 1) / backingStore;\n",
       "\n",
       "    var rubberband = $('<canvas/>');\n",
       "    rubberband.attr('style', \"position: absolute; left: 0; top: 0; z-index: 1;\")\n",
       "\n",
       "    var pass_mouse_events = true;\n",
       "\n",
       "    canvas_div.resizable({\n",
       "        start: function(event, ui) {\n",
       "            pass_mouse_events = false;\n",
       "        },\n",
       "        resize: function(event, ui) {\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "        stop: function(event, ui) {\n",
       "            pass_mouse_events = true;\n",
       "            fig.request_resize(ui.size.width, ui.size.height);\n",
       "        },\n",
       "    });\n",
       "\n",
       "    function mouse_event_fn(event) {\n",
       "        if (pass_mouse_events)\n",
       "            return fig.mouse_event(event, event['data']);\n",
       "    }\n",
       "\n",
       "    rubberband.mousedown('button_press', mouse_event_fn);\n",
       "    rubberband.mouseup('button_release', mouse_event_fn);\n",
       "    // Throttle sequential mouse events to 1 every 20ms.\n",
       "    rubberband.mousemove('motion_notify', mouse_event_fn);\n",
       "\n",
       "    rubberband.mouseenter('figure_enter', mouse_event_fn);\n",
       "    rubberband.mouseleave('figure_leave', mouse_event_fn);\n",
       "\n",
       "    canvas_div.on(\"wheel\", function (event) {\n",
       "        event = event.originalEvent;\n",
       "        event['data'] = 'scroll'\n",
       "        if (event.deltaY < 0) {\n",
       "            event.step = 1;\n",
       "        } else {\n",
       "            event.step = -1;\n",
       "        }\n",
       "        mouse_event_fn(event);\n",
       "    });\n",
       "\n",
       "    canvas_div.append(canvas);\n",
       "    canvas_div.append(rubberband);\n",
       "\n",
       "    this.rubberband = rubberband;\n",
       "    this.rubberband_canvas = rubberband[0];\n",
       "    this.rubberband_context = rubberband[0].getContext(\"2d\");\n",
       "    this.rubberband_context.strokeStyle = \"#000000\";\n",
       "\n",
       "    this._resize_canvas = function(width, height) {\n",
       "        // Keep the size of the canvas, canvas container, and rubber band\n",
       "        // canvas in synch.\n",
       "        canvas_div.css('width', width)\n",
       "        canvas_div.css('height', height)\n",
       "\n",
       "        canvas.attr('width', width * mpl.ratio);\n",
       "        canvas.attr('height', height * mpl.ratio);\n",
       "        canvas.attr('style', 'width: ' + width + 'px; height: ' + height + 'px;');\n",
       "\n",
       "        rubberband.attr('width', width);\n",
       "        rubberband.attr('height', height);\n",
       "    }\n",
       "\n",
       "    // Set the figure to an initial 600x600px, this will subsequently be updated\n",
       "    // upon first draw.\n",
       "    this._resize_canvas(600, 600);\n",
       "\n",
       "    // Disable right mouse context menu.\n",
       "    $(this.rubberband_canvas).bind(\"contextmenu\",function(e){\n",
       "        return false;\n",
       "    });\n",
       "\n",
       "    function set_focus () {\n",
       "        canvas.focus();\n",
       "        canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    window.setTimeout(set_focus, 100);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>');\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items) {\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) {\n",
       "            // put a spacer in here.\n",
       "            continue;\n",
       "        }\n",
       "        var button = $('<button/>');\n",
       "        button.addClass('ui-button ui-widget ui-state-default ui-corner-all ' +\n",
       "                        'ui-button-icon-only');\n",
       "        button.attr('role', 'button');\n",
       "        button.attr('aria-disabled', 'false');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "\n",
       "        var icon_img = $('<span/>');\n",
       "        icon_img.addClass('ui-button-icon-primary ui-icon');\n",
       "        icon_img.addClass(image);\n",
       "        icon_img.addClass('ui-corner-all');\n",
       "\n",
       "        var tooltip_span = $('<span/>');\n",
       "        tooltip_span.addClass('ui-button-text');\n",
       "        tooltip_span.html(tooltip);\n",
       "\n",
       "        button.append(icon_img);\n",
       "        button.append(tooltip_span);\n",
       "\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    var fmt_picker_span = $('<span/>');\n",
       "\n",
       "    var fmt_picker = $('<select/>');\n",
       "    fmt_picker.addClass('mpl-toolbar-option ui-widget ui-widget-content');\n",
       "    fmt_picker_span.append(fmt_picker);\n",
       "    nav_element.append(fmt_picker_span);\n",
       "    this.format_dropdown = fmt_picker[0];\n",
       "\n",
       "    for (var ind in mpl.extensions) {\n",
       "        var fmt = mpl.extensions[ind];\n",
       "        var option = $(\n",
       "            '<option/>', {selected: fmt === mpl.default_extension}).html(fmt);\n",
       "        fmt_picker.append(option);\n",
       "    }\n",
       "\n",
       "    // Add hover states to the ui-buttons\n",
       "    $( \".ui-button\" ).hover(\n",
       "        function() { $(this).addClass(\"ui-state-hover\");},\n",
       "        function() { $(this).removeClass(\"ui-state-hover\");}\n",
       "    );\n",
       "\n",
       "    var status_bar = $('<span class=\"mpl-message\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.request_resize = function(x_pixels, y_pixels) {\n",
       "    // Request matplotlib to resize the figure. Matplotlib will then trigger a resize in the client,\n",
       "    // which will in turn request a refresh of the image.\n",
       "    this.send_message('resize', {'width': x_pixels, 'height': y_pixels});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_message = function(type, properties) {\n",
       "    properties['type'] = type;\n",
       "    properties['figure_id'] = this.id;\n",
       "    this.ws.send(JSON.stringify(properties));\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.send_draw_message = function() {\n",
       "    if (!this.waiting) {\n",
       "        this.waiting = true;\n",
       "        this.ws.send(JSON.stringify({type: \"draw\", figure_id: this.id}));\n",
       "    }\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    var format_dropdown = fig.format_dropdown;\n",
       "    var format = format_dropdown.options[format_dropdown.selectedIndex].value;\n",
       "    fig.ondownload(fig, format);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.figure.prototype.handle_resize = function(fig, msg) {\n",
       "    var size = msg['size'];\n",
       "    if (size[0] != fig.canvas.width || size[1] != fig.canvas.height) {\n",
       "        fig._resize_canvas(size[0], size[1]);\n",
       "        fig.send_message(\"refresh\", {});\n",
       "    };\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_rubberband = function(fig, msg) {\n",
       "    var x0 = msg['x0'] / mpl.ratio;\n",
       "    var y0 = (fig.canvas.height - msg['y0']) / mpl.ratio;\n",
       "    var x1 = msg['x1'] / mpl.ratio;\n",
       "    var y1 = (fig.canvas.height - msg['y1']) / mpl.ratio;\n",
       "    x0 = Math.floor(x0) + 0.5;\n",
       "    y0 = Math.floor(y0) + 0.5;\n",
       "    x1 = Math.floor(x1) + 0.5;\n",
       "    y1 = Math.floor(y1) + 0.5;\n",
       "    var min_x = Math.min(x0, x1);\n",
       "    var min_y = Math.min(y0, y1);\n",
       "    var width = Math.abs(x1 - x0);\n",
       "    var height = Math.abs(y1 - y0);\n",
       "\n",
       "    fig.rubberband_context.clearRect(\n",
       "        0, 0, fig.canvas.width / mpl.ratio, fig.canvas.height / mpl.ratio);\n",
       "\n",
       "    fig.rubberband_context.strokeRect(min_x, min_y, width, height);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_figure_label = function(fig, msg) {\n",
       "    // Updates the figure title.\n",
       "    fig.header.textContent = msg['label'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_cursor = function(fig, msg) {\n",
       "    var cursor = msg['cursor'];\n",
       "    switch(cursor)\n",
       "    {\n",
       "    case 0:\n",
       "        cursor = 'pointer';\n",
       "        break;\n",
       "    case 1:\n",
       "        cursor = 'default';\n",
       "        break;\n",
       "    case 2:\n",
       "        cursor = 'crosshair';\n",
       "        break;\n",
       "    case 3:\n",
       "        cursor = 'move';\n",
       "        break;\n",
       "    }\n",
       "    fig.rubberband_canvas.style.cursor = cursor;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_message = function(fig, msg) {\n",
       "    fig.message.textContent = msg['message'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_draw = function(fig, msg) {\n",
       "    // Request the server to send over a new figure.\n",
       "    fig.send_draw_message();\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_image_mode = function(fig, msg) {\n",
       "    fig.image_mode = msg['mode'];\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Called whenever the canvas gets updated.\n",
       "    this.send_message(\"ack\", {});\n",
       "}\n",
       "\n",
       "// A function to construct a web socket function for onmessage handling.\n",
       "// Called in the figure constructor.\n",
       "mpl.figure.prototype._make_on_message_function = function(fig) {\n",
       "    return function socket_on_message(evt) {\n",
       "        if (evt.data instanceof Blob) {\n",
       "            /* FIXME: We get \"Resource interpreted as Image but\n",
       "             * transferred with MIME type text/plain:\" errors on\n",
       "             * Chrome.  But how to set the MIME type?  It doesn't seem\n",
       "             * to be part of the websocket stream */\n",
       "            evt.data.type = \"image/png\";\n",
       "\n",
       "            /* Free the memory for the previous frames */\n",
       "            if (fig.imageObj.src) {\n",
       "                (window.URL || window.webkitURL).revokeObjectURL(\n",
       "                    fig.imageObj.src);\n",
       "            }\n",
       "\n",
       "            fig.imageObj.src = (window.URL || window.webkitURL).createObjectURL(\n",
       "                evt.data);\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "        else if (typeof evt.data === 'string' && evt.data.slice(0, 21) == \"data:image/png;base64\") {\n",
       "            fig.imageObj.src = evt.data;\n",
       "            fig.updated_canvas_event();\n",
       "            fig.waiting = false;\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        var msg = JSON.parse(evt.data);\n",
       "        var msg_type = msg['type'];\n",
       "\n",
       "        // Call the  \"handle_{type}\" callback, which takes\n",
       "        // the figure and JSON message as its only arguments.\n",
       "        try {\n",
       "            var callback = fig[\"handle_\" + msg_type];\n",
       "        } catch (e) {\n",
       "            console.log(\"No handler for the '\" + msg_type + \"' message type: \", msg);\n",
       "            return;\n",
       "        }\n",
       "\n",
       "        if (callback) {\n",
       "            try {\n",
       "                // console.log(\"Handling '\" + msg_type + \"' message: \", msg);\n",
       "                callback(fig, msg);\n",
       "            } catch (e) {\n",
       "                console.log(\"Exception inside the 'handler_\" + msg_type + \"' callback:\", e, e.stack, msg);\n",
       "            }\n",
       "        }\n",
       "    };\n",
       "}\n",
       "\n",
       "// from http://stackoverflow.com/questions/1114465/getting-mouse-location-in-canvas\n",
       "mpl.findpos = function(e) {\n",
       "    //this section is from http://www.quirksmode.org/js/events_properties.html\n",
       "    var targ;\n",
       "    if (!e)\n",
       "        e = window.event;\n",
       "    if (e.target)\n",
       "        targ = e.target;\n",
       "    else if (e.srcElement)\n",
       "        targ = e.srcElement;\n",
       "    if (targ.nodeType == 3) // defeat Safari bug\n",
       "        targ = targ.parentNode;\n",
       "\n",
       "    // jQuery normalizes the pageX and pageY\n",
       "    // pageX,Y are the mouse positions relative to the document\n",
       "    // offset() returns the position of the element relative to the document\n",
       "    var x = e.pageX - $(targ).offset().left;\n",
       "    var y = e.pageY - $(targ).offset().top;\n",
       "\n",
       "    return {\"x\": x, \"y\": y};\n",
       "};\n",
       "\n",
       "/*\n",
       " * return a copy of an object with only non-object keys\n",
       " * we need this to avoid circular references\n",
       " * http://stackoverflow.com/a/24161582/3208463\n",
       " */\n",
       "function simpleKeys (original) {\n",
       "  return Object.keys(original).reduce(function (obj, key) {\n",
       "    if (typeof original[key] !== 'object')\n",
       "        obj[key] = original[key]\n",
       "    return obj;\n",
       "  }, {});\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.mouse_event = function(event, name) {\n",
       "    var canvas_pos = mpl.findpos(event)\n",
       "\n",
       "    if (name === 'button_press')\n",
       "    {\n",
       "        this.canvas.focus();\n",
       "        this.canvas_div.focus();\n",
       "    }\n",
       "\n",
       "    var x = canvas_pos.x * mpl.ratio;\n",
       "    var y = canvas_pos.y * mpl.ratio;\n",
       "\n",
       "    this.send_message(name, {x: x, y: y, button: event.button,\n",
       "                             step: event.step,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "\n",
       "    /* This prevents the web browser from automatically changing to\n",
       "     * the text insertion cursor when the button is pressed.  We want\n",
       "     * to control all of the cursor setting manually through the\n",
       "     * 'cursor' event from matplotlib */\n",
       "    event.preventDefault();\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    // Handle any extra behaviour associated with a key event\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.key_event = function(event, name) {\n",
       "\n",
       "    // Prevent repeat events\n",
       "    if (name == 'key_press')\n",
       "    {\n",
       "        if (event.which === this._key)\n",
       "            return;\n",
       "        else\n",
       "            this._key = event.which;\n",
       "    }\n",
       "    if (name == 'key_release')\n",
       "        this._key = null;\n",
       "\n",
       "    var value = '';\n",
       "    if (event.ctrlKey && event.which != 17)\n",
       "        value += \"ctrl+\";\n",
       "    if (event.altKey && event.which != 18)\n",
       "        value += \"alt+\";\n",
       "    if (event.shiftKey && event.which != 16)\n",
       "        value += \"shift+\";\n",
       "\n",
       "    value += 'k';\n",
       "    value += event.which.toString();\n",
       "\n",
       "    this._key_event_extra(event, name);\n",
       "\n",
       "    this.send_message(name, {key: value,\n",
       "                             guiEvent: simpleKeys(event)});\n",
       "    return false;\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onclick = function(name) {\n",
       "    if (name == 'download') {\n",
       "        this.handle_save(this, null);\n",
       "    } else {\n",
       "        this.send_message(\"toolbar_button\", {name: name});\n",
       "    }\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.toolbar_button_onmouseover = function(tooltip) {\n",
       "    this.message.textContent = tooltip;\n",
       "};\n",
       "mpl.toolbar_items = [[\"Home\", \"Reset original view\", \"fa fa-home icon-home\", \"home\"], [\"Back\", \"Back to previous view\", \"fa fa-arrow-left icon-arrow-left\", \"back\"], [\"Forward\", \"Forward to next view\", \"fa fa-arrow-right icon-arrow-right\", \"forward\"], [\"\", \"\", \"\", \"\"], [\"Pan\", \"Pan axes with left mouse, zoom with right\", \"fa fa-arrows icon-move\", \"pan\"], [\"Zoom\", \"Zoom to rectangle\", \"fa fa-square-o icon-check-empty\", \"zoom\"], [\"\", \"\", \"\", \"\"], [\"Download\", \"Download plot\", \"fa fa-floppy-o icon-save\", \"download\"]];\n",
       "\n",
       "mpl.extensions = [\"eps\", \"jpeg\", \"pdf\", \"png\", \"ps\", \"raw\", \"svg\", \"tif\"];\n",
       "\n",
       "mpl.default_extension = \"png\";var comm_websocket_adapter = function(comm) {\n",
       "    // Create a \"websocket\"-like object which calls the given IPython comm\n",
       "    // object with the appropriate methods. Currently this is a non binary\n",
       "    // socket, so there is still some room for performance tuning.\n",
       "    var ws = {};\n",
       "\n",
       "    ws.close = function() {\n",
       "        comm.close()\n",
       "    };\n",
       "    ws.send = function(m) {\n",
       "        //console.log('sending', m);\n",
       "        comm.send(m);\n",
       "    };\n",
       "    // Register the callback with on_msg.\n",
       "    comm.on_msg(function(msg) {\n",
       "        //console.log('receiving', msg['content']['data'], msg);\n",
       "        // Pass the mpl event to the overridden (by mpl) onmessage function.\n",
       "        ws.onmessage(msg['content']['data'])\n",
       "    });\n",
       "    return ws;\n",
       "}\n",
       "\n",
       "mpl.mpl_figure_comm = function(comm, msg) {\n",
       "    // This is the function which gets called when the mpl process\n",
       "    // starts-up an IPython Comm through the \"matplotlib\" channel.\n",
       "\n",
       "    var id = msg.content.data.id;\n",
       "    // Get hold of the div created by the display call when the Comm\n",
       "    // socket was opened in Python.\n",
       "    var element = $(\"#\" + id);\n",
       "    var ws_proxy = comm_websocket_adapter(comm)\n",
       "\n",
       "    function ondownload(figure, format) {\n",
       "        window.open(figure.imageObj.src);\n",
       "    }\n",
       "\n",
       "    var fig = new mpl.figure(id, ws_proxy,\n",
       "                           ondownload,\n",
       "                           element.get(0));\n",
       "\n",
       "    // Call onopen now - mpl needs it, as it is assuming we've passed it a real\n",
       "    // web socket which is closed, not our websocket->open comm proxy.\n",
       "    ws_proxy.onopen();\n",
       "\n",
       "    fig.parent_element = element.get(0);\n",
       "    fig.cell_info = mpl.find_output_cell(\"<div id='\" + id + \"'></div>\");\n",
       "    if (!fig.cell_info) {\n",
       "        console.error(\"Failed to find cell for figure\", id, fig);\n",
       "        return;\n",
       "    }\n",
       "\n",
       "    var output_index = fig.cell_info[2]\n",
       "    var cell = fig.cell_info[0];\n",
       "\n",
       "};\n",
       "\n",
       "mpl.figure.prototype.handle_close = function(fig, msg) {\n",
       "    var width = fig.canvas.width/mpl.ratio\n",
       "    fig.root.unbind('remove')\n",
       "\n",
       "    // Update the output cell to use the data from the current canvas.\n",
       "    fig.push_to_output();\n",
       "    var dataURL = fig.canvas.toDataURL();\n",
       "    // Re-enable the keyboard manager in IPython - without this line, in FF,\n",
       "    // the notebook keyboard shortcuts fail.\n",
       "    IPython.keyboard_manager.enable()\n",
       "    $(fig.parent_element).html('<img src=\"' + dataURL + '\" width=\"' + width + '\">');\n",
       "    fig.close_ws(fig, msg);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.close_ws = function(fig, msg){\n",
       "    fig.send_message('closing', msg);\n",
       "    // fig.ws.close()\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.push_to_output = function(remove_interactive) {\n",
       "    // Turn the data on the canvas into data in the output cell.\n",
       "    var width = this.canvas.width/mpl.ratio\n",
       "    var dataURL = this.canvas.toDataURL();\n",
       "    this.cell_info[1]['text/html'] = '<img src=\"' + dataURL + '\" width=\"' + width + '\">';\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.updated_canvas_event = function() {\n",
       "    // Tell IPython that the notebook contents must change.\n",
       "    IPython.notebook.set_dirty(true);\n",
       "    this.send_message(\"ack\", {});\n",
       "    var fig = this;\n",
       "    // Wait a second, then push the new image to the DOM so\n",
       "    // that it is saved nicely (might be nice to debounce this).\n",
       "    setTimeout(function () { fig.push_to_output() }, 1000);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._init_toolbar = function() {\n",
       "    var fig = this;\n",
       "\n",
       "    var nav_element = $('<div/>');\n",
       "    nav_element.attr('style', 'width: 100%');\n",
       "    this.root.append(nav_element);\n",
       "\n",
       "    // Define a callback function for later on.\n",
       "    function toolbar_event(event) {\n",
       "        return fig.toolbar_button_onclick(event['data']);\n",
       "    }\n",
       "    function toolbar_mouse_event(event) {\n",
       "        return fig.toolbar_button_onmouseover(event['data']);\n",
       "    }\n",
       "\n",
       "    for(var toolbar_ind in mpl.toolbar_items){\n",
       "        var name = mpl.toolbar_items[toolbar_ind][0];\n",
       "        var tooltip = mpl.toolbar_items[toolbar_ind][1];\n",
       "        var image = mpl.toolbar_items[toolbar_ind][2];\n",
       "        var method_name = mpl.toolbar_items[toolbar_ind][3];\n",
       "\n",
       "        if (!name) { continue; };\n",
       "\n",
       "        var button = $('<button class=\"btn btn-default\" href=\"#\" title=\"' + name + '\"><i class=\"fa ' + image + ' fa-lg\"></i></button>');\n",
       "        button.click(method_name, toolbar_event);\n",
       "        button.mouseover(tooltip, toolbar_mouse_event);\n",
       "        nav_element.append(button);\n",
       "    }\n",
       "\n",
       "    // Add the status bar.\n",
       "    var status_bar = $('<span class=\"mpl-message\" style=\"text-align:right; float: right;\"/>');\n",
       "    nav_element.append(status_bar);\n",
       "    this.message = status_bar[0];\n",
       "\n",
       "    // Add the close button to the window.\n",
       "    var buttongrp = $('<div class=\"btn-group inline pull-right\"></div>');\n",
       "    var button = $('<button class=\"btn btn-mini btn-primary\" href=\"#\" title=\"Stop Interaction\"><i class=\"fa fa-power-off icon-remove icon-large\"></i></button>');\n",
       "    button.click(function (evt) { fig.handle_close(fig, {}); } );\n",
       "    button.mouseover('Stop Interaction', toolbar_mouse_event);\n",
       "    buttongrp.append(button);\n",
       "    var titlebar = this.root.find($('.ui-dialog-titlebar'));\n",
       "    titlebar.prepend(buttongrp);\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._root_extra_style = function(el){\n",
       "    var fig = this\n",
       "    el.on(\"remove\", function(){\n",
       "\tfig.close_ws(fig, {});\n",
       "    });\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._canvas_extra_style = function(el){\n",
       "    // this is important to make the div 'focusable\n",
       "    el.attr('tabindex', 0)\n",
       "    // reach out to IPython and tell the keyboard manager to turn it's self\n",
       "    // off when our div gets focus\n",
       "\n",
       "    // location in version 3\n",
       "    if (IPython.notebook.keyboard_manager) {\n",
       "        IPython.notebook.keyboard_manager.register_events(el);\n",
       "    }\n",
       "    else {\n",
       "        // location in version 2\n",
       "        IPython.keyboard_manager.register_events(el);\n",
       "    }\n",
       "\n",
       "}\n",
       "\n",
       "mpl.figure.prototype._key_event_extra = function(event, name) {\n",
       "    var manager = IPython.notebook.keyboard_manager;\n",
       "    if (!manager)\n",
       "        manager = IPython.keyboard_manager;\n",
       "\n",
       "    // Check for shift+enter\n",
       "    if (event.shiftKey && event.which == 13) {\n",
       "        this.canvas_div.blur();\n",
       "        // select the cell after this one\n",
       "        var index = IPython.notebook.find_cell_index(this.cell_info[0]);\n",
       "        IPython.notebook.select(index + 1);\n",
       "    }\n",
       "}\n",
       "\n",
       "mpl.figure.prototype.handle_save = function(fig, msg) {\n",
       "    fig.ondownload(fig, null);\n",
       "}\n",
       "\n",
       "\n",
       "mpl.find_output_cell = function(html_output) {\n",
       "    // Return the cell and output element which can be found *uniquely* in the notebook.\n",
       "    // Note - this is a bit hacky, but it is done because the \"notebook_saving.Notebook\"\n",
       "    // IPython event is triggered only after the cells have been serialised, which for\n",
       "    // our purposes (turning an active figure into a static one), is too late.\n",
       "    var cells = IPython.notebook.get_cells();\n",
       "    var ncells = cells.length;\n",
       "    for (var i=0; i<ncells; i++) {\n",
       "        var cell = cells[i];\n",
       "        if (cell.cell_type === 'code'){\n",
       "            for (var j=0; j<cell.output_area.outputs.length; j++) {\n",
       "                var data = cell.output_area.outputs[j];\n",
       "                if (data.data) {\n",
       "                    // IPython >= 3 moved mimebundle to data attribute of output\n",
       "                    data = data.data;\n",
       "                }\n",
       "                if (data['text/html'] == html_output) {\n",
       "                    return [cell, data, j];\n",
       "                }\n",
       "            }\n",
       "        }\n",
       "    }\n",
       "}\n",
       "\n",
       "// Register the function which deals with the matplotlib target/channel.\n",
       "// The kernel may be null if the page has been refreshed.\n",
       "if (IPython.notebook.kernel != null) {\n",
       "    IPython.notebook.kernel.comm_manager.register_target('matplotlib', mpl.mpl_figure_comm);\n",
       "}\n"
      ],
      "text/plain": [
       "<IPython.core.display.Javascript object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<img src=\"data:image/png;base64,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\" width=\"432\">"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(probs_singl[probs_seq_argmax != probs_singl_argmax].max(axis=1), bins=25)\n",
    "plt.hist(probs_seq[probs_seq_argmax != probs_singl_argmax].max(axis=1), bins=25)\n",
    "# plt.hist(probs_singl[probs_singl_argmax != data_test.open_channels.values].max(axis=1), bins=25)\n",
    "# plt.hist(prediction.max(axis=1), bins=25)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(500000, 11)"
      ]
     },
     "execution_count": 74,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prediction.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 478,
   "metadata": {},
   "outputs": [],
   "source": [
    "prediction = prediction.cpu().detach().numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 492,
   "metadata": {},
   "outputs": [],
   "source": [
    "prediction = np.argmax(probs_seq*probs_singl, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 243,
   "metadata": {},
   "outputs": [],
   "source": [
    "prediction = np.argmax(probs_seq, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 244,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "prediction[probs_seq.max(axis=1) < 0.91] = np.argmax(probs_seq[probs_seq.max(axis=1) < 0.91]*probs_singl[probs_seq.max(axis=1) < 0.91], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 482,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "data_test = pd.read_csv('data/dev.csv')\n",
    "true_indx = data_test.index[prediction == data_test.open_channels.values]\n",
    "false_indx = data_test.index[prediction != data_test.open_channels.values]\n",
    "true_data = data_test.open_channels.values[prediction == data_test.open_channels.values]\n",
    "false_data = data_test.open_channels.values[prediction != data_test.open_channels.values]\n",
    "\n",
    "%matplotlib inline\n",
    "plt.figure(figsize=(12,6))\n",
    "plt.plot(data_test.open_channels.values)\n",
    "plt.plot(prediction)\n",
    "plt.plot(false_indx, false_data, lw=0, marker='o', color='red')\n",
    "plt.plot(true_indx, true_data, lw=0, marker='o', color='green')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 480,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accur: signl 0.823706\n",
      "f1: signl 0.7114041859388094\n"
     ]
    }
   ],
   "source": [
    "print('accur: signl', sum(prediction == data_test.open_channels.values)/len(prediction))\n",
    "print('f1: signl',f1_score(data_test.open_channels.values, prediction, average='macro'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 483,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accur seq:  0.84726\n",
      "f1 seq:  0.7351337696028376\n"
     ]
    }
   ],
   "source": [
    "print('accur seq: ', sum(prediction == data_test.open_channels.values)/len(prediction))\n",
    "print('f1 seq: ',f1_score(data_test.open_channels.values, prediction, average='macro'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 352,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accur:  0.9806505263157895\n",
      "f1:  0.982546007616218\n"
     ]
    }
   ],
   "source": [
    "print('accur: ', sum(prediction == data_test.open_channels.values)/len(prediction))\n",
    "print('f1: ',f1_score(data_test.open_channels.values, prediction, average='macro'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 384,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accur seq:  0.83509\n",
      "f1 seq:  0.699142267153876\n"
     ]
    }
   ],
   "source": [
    "print('accur seq: ', sum(prediction == data_test.open_channels.values)/len(prediction))\n",
    "print('f1 seq: ',f1_score(data_test.open_channels.values, prediction, average='macro'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 387,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accur:  0.848864\n",
      "f1:  0.7179935177758291\n"
     ]
    }
   ],
   "source": [
    "print('accur: ', sum(prediction == data_test.open_channels.values)/len(prediction))\n",
    "print('f1: ',f1_score(data_test.open_channels.values, prediction, average='macro'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 501,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(100000,)"
      ]
     },
     "execution_count": 501,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prediction.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 506,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9035"
      ]
     },
     "execution_count": 506,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum(submmit.open_channels[submmit.batch == 5] == prediction)/100000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 507,
   "metadata": {},
   "outputs": [],
   "source": [
    "submmit = pd.read_csv('submmit.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 508,
   "metadata": {},
   "outputs": [],
   "source": [
    "submmit = pd.read_csv('submmit.csv')\n",
    "submmit.open_channels[submmit.batch == 5] = prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 509,
   "metadata": {},
   "outputs": [],
   "source": [
    "submmit.to_csv('submmit.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>signal</th>\n",
       "      <th>batch</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>open_channels</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1221046</td>\n",
       "      <td>1221046</td>\n",
       "      <td>1221046</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>215572</td>\n",
       "      <td>215572</td>\n",
       "      <td>215572</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>117129</td>\n",
       "      <td>117129</td>\n",
       "      <td>117129</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>134419</td>\n",
       "      <td>134419</td>\n",
       "      <td>134419</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>81587</td>\n",
       "      <td>81587</td>\n",
       "      <td>81587</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>56372</td>\n",
       "      <td>56372</td>\n",
       "      <td>56372</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>37071</td>\n",
       "      <td>37071</td>\n",
       "      <td>37071</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>52886</td>\n",
       "      <td>52886</td>\n",
       "      <td>52886</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>49684</td>\n",
       "      <td>49684</td>\n",
       "      <td>49684</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>27231</td>\n",
       "      <td>27231</td>\n",
       "      <td>27231</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>7003</td>\n",
       "      <td>7003</td>\n",
       "      <td>7003</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  time   signal    batch\n",
       "open_channels                           \n",
       "0              1221046  1221046  1221046\n",
       "1               215572   215572   215572\n",
       "2               117129   117129   117129\n",
       "3               134419   134419   134419\n",
       "4                81587    81587    81587\n",
       "5                56372    56372    56372\n",
       "6                37071    37071    37071\n",
       "7                52886    52886    52886\n",
       "8                49684    49684    49684\n",
       "9                27231    27231    27231\n",
       "10                7003     7003     7003"
      ]
     },
     "execution_count": 122,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submmit.groupby('open_channels').count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>signal</th>\n",
       "      <th>batch</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>open_channels</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1221046</td>\n",
       "      <td>1221046</td>\n",
       "      <td>1221046</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>215572</td>\n",
       "      <td>215572</td>\n",
       "      <td>215572</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>117129</td>\n",
       "      <td>117129</td>\n",
       "      <td>117129</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>134419</td>\n",
       "      <td>134419</td>\n",
       "      <td>134419</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>81587</td>\n",
       "      <td>81587</td>\n",
       "      <td>81587</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>56372</td>\n",
       "      <td>56372</td>\n",
       "      <td>56372</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>37071</td>\n",
       "      <td>37071</td>\n",
       "      <td>37071</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>52886</td>\n",
       "      <td>52886</td>\n",
       "      <td>52886</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>49684</td>\n",
       "      <td>49684</td>\n",
       "      <td>49684</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>27231</td>\n",
       "      <td>27231</td>\n",
       "      <td>27231</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>7003</td>\n",
       "      <td>7003</td>\n",
       "      <td>7003</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  time   signal    batch\n",
       "open_channels                           \n",
       "0              1221046  1221046  1221046\n",
       "1               215572   215572   215572\n",
       "2               117129   117129   117129\n",
       "3               134419   134419   134419\n",
       "4                81587    81587    81587\n",
       "5                56372    56372    56372\n",
       "6                37071    37071    37071\n",
       "7                52886    52886    52886\n",
       "8                49684    49684    49684\n",
       "9                27231    27231    27231\n",
       "10                7003     7003     7003"
      ]
     },
     "execution_count": 120,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submmit.groupby('open_channels').count()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "data\n",
    "for c in range(10):\n",
    "    data['s'+str(c)] = np.ones(len(data))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>time</th>\n",
       "      <th>signal</th>\n",
       "      <th>open_channels</th>\n",
       "      <th>s0</th>\n",
       "      <th>s1</th>\n",
       "      <th>s2</th>\n",
       "      <th>s3</th>\n",
       "      <th>s4</th>\n",
       "      <th>s5</th>\n",
       "      <th>s6</th>\n",
       "      <th>s7</th>\n",
       "      <th>s8</th>\n",
       "      <th>s9</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0001</td>\n",
       "      <td>-2.7600</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.0002</td>\n",
       "      <td>-2.8557</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.0003</td>\n",
       "      <td>-2.4074</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.0004</td>\n",
       "      <td>-3.1404</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.0005</td>\n",
       "      <td>-3.1525</td>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4999995</th>\n",
       "      <td>499.9996</td>\n",
       "      <td>2.9194</td>\n",
       "      <td>7</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4999996</th>\n",
       "      <td>499.9997</td>\n",
       "      <td>2.6980</td>\n",
       "      <td>7</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4999997</th>\n",
       "      <td>499.9998</td>\n",
       "      <td>4.5164</td>\n",
       "      <td>8</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4999998</th>\n",
       "      <td>499.9999</td>\n",
       "      <td>5.6397</td>\n",
       "      <td>9</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4999999</th>\n",
       "      <td>500.0000</td>\n",
       "      <td>5.3792</td>\n",
       "      <td>9</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5000000 rows × 13 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             time  signal  open_channels   s0   s1   s2   s3   s4   s5   s6  \\\n",
       "0          0.0001 -2.7600              0  1.0  1.0  1.0  1.0  1.0  1.0  1.0   \n",
       "1          0.0002 -2.8557              0  1.0  1.0  1.0  1.0  1.0  1.0  1.0   \n",
       "2          0.0003 -2.4074              0  1.0  1.0  1.0  1.0  1.0  1.0  1.0   \n",
       "3          0.0004 -3.1404              0  1.0  1.0  1.0  1.0  1.0  1.0  1.0   \n",
       "4          0.0005 -3.1525              0  1.0  1.0  1.0  1.0  1.0  1.0  1.0   \n",
       "...           ...     ...            ...  ...  ...  ...  ...  ...  ...  ...   \n",
       "4999995  499.9996  2.9194              7  1.0  1.0  1.0  1.0  1.0  1.0  1.0   \n",
       "4999996  499.9997  2.6980              7  1.0  1.0  1.0  1.0  1.0  1.0  1.0   \n",
       "4999997  499.9998  4.5164              8  1.0  1.0  1.0  1.0  1.0  1.0  1.0   \n",
       "4999998  499.9999  5.6397              9  1.0  1.0  1.0  1.0  1.0  1.0  1.0   \n",
       "4999999  500.0000  5.3792              9  1.0  1.0  1.0  1.0  1.0  1.0  1.0   \n",
       "\n",
       "          s7   s8   s9  \n",
       "0        1.0  1.0  1.0  \n",
       "1        1.0  1.0  1.0  \n",
       "2        1.0  1.0  1.0  \n",
       "3        1.0  1.0  1.0  \n",
       "4        1.0  1.0  1.0  \n",
       "...      ...  ...  ...  \n",
       "4999995  1.0  1.0  1.0  \n",
       "4999996  1.0  1.0  1.0  \n",
       "4999997  1.0  1.0  1.0  \n",
       "4999998  1.0  1.0  1.0  \n",
       "4999999  1.0  1.0  1.0  \n",
       "\n",
       "[5000000 rows x 13 columns]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:py36]",
   "language": "python",
   "name": "conda-env-py36-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}