LSTM_training_set3.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from configs import cfg\n",
    "\n",
    "cfg['input_dim'] = 301 # input dimension to LSTM\n",
    "cfg['hidden_dim'] = 128 # hidden dimension for LSTM\n",
    "cfg['output_dim'] = 1 # output dimension of the model\n",
    "cfg['layers'] = 2 # number of layers of LSTM\n",
    "\n",
    "cfg['dropout'] = 5e-3 # dropout rate between two layers of LSTM; useful only when layers > 1; between 0 and 1\n",
    "cfg['bidirectional'] = True # True or False; True means using a bidirectional LSTM\n",
    "cfg['batch_size'] = 120 # batch size of input\n",
    "cfg['learning_rate'] = 1e-3 # learning rate to be used\n",
    "cfg['L2_penalty'] = 1e-4 # weighting constant for L2 regularization term; this is a parameter when you define optimizer\n",
    "cfg['epochs'] = 15 # number of epochs for which the model is trained\n",
    "cfg['embed'] = False\n",
    "cfg['score_range'] = 3\n",
    "cfg['set_id'] = 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from preprocess import *\n",
    "import pandas as pd\n",
    "\n",
    "train_path = './essay_dataset/training_set_rel3.tsv'\n",
    "training_data = load_training_set(train_path, cfg['set_id'])\n",
    "\n",
    "glove_dict = load_dict('glove.840B.300d.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "valid_path = './essay_dataset/valid_set.tsv'\n",
    "valid_label_path = './essay_dataset/valid_sample_submission_2_column.csv'\n",
    "valid_data = load_validation_set(valid_path, valid_label_path, cfg['set_id'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "training_data = process_scores(training_data, 'domain1_score')\n",
    "valid_data = process_scores(valid_data, 'domain1_score')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ready\n"
     ]
    }
   ],
   "source": [
    "from dataloader import *\n",
    "from models import *\n",
    "import time\n",
    "\n",
    "model = LSTM_Score(cfg, True)\n",
    "model = model.to(torch.device(cfg['device']))\n",
    "\n",
    "criterion = nn.MSELoss()\n",
    "\n",
    "#optimizer = torch.optim.RMSprop(model.parameters(), lr=cfg['learning_rate'], weight_decay=cfg['L2_penalty'])\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=cfg['learning_rate'], weight_decay=cfg['L2_penalty'])\n",
    "#optimizer = torch.optim.Adadelta(model.parameters(), lr=cfg['learning_rate'], weight_decay=cfg['L2_penalty'])\n",
    "\n",
    "train_indices = list(range(len(training_data)))\n",
    "valid_indices = list(range(len(valid_data)))\n",
    "\n",
    "print('ready')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LSTM_Score(\n",
       "  (embed_linear): Linear(in_features=301, out_features=301, bias=True)\n",
       "  (lstm): LSTM(301, 128, num_layers=2, batch_first=True, dropout=0.005, bidirectional=True)\n",
       "  (bn): BatchNorm1d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "  (fc1): Linear(in_features=256, out_features=1, bias=True)\n",
       ")"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.train()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "count = 2, loss = 0.70360\n",
      "count = 4, loss = 0.70884\n",
      "count = 6, loss = 0.32025\n",
      "count = 8, loss = 0.47326\n",
      "count = 10, loss = 0.32392\n",
      "count = 12, loss = 0.38961\n",
      "count = 14, loss = 0.36599\n",
      "validation loss: 0.3000406362116337\n",
      "epoch finished: 1\n",
      "count = 2, loss = 0.33911\n",
      "count = 4, loss = 0.37798\n",
      "count = 6, loss = 0.31899\n",
      "count = 8, loss = 0.34179\n",
      "count = 10, loss = 0.28248\n",
      "count = 12, loss = 0.32906\n",
      "count = 14, loss = 0.31596\n",
      "validation loss: 0.24380934983491898\n",
      "epoch finished: 2\n",
      "count = 2, loss = 0.29814\n",
      "count = 4, loss = 0.31534\n",
      "count = 6, loss = 0.29570\n",
      "count = 8, loss = 0.29868\n",
      "count = 10, loss = 0.27351\n",
      "count = 12, loss = 0.25751\n",
      "count = 14, loss = 0.37781\n",
      "validation loss: 0.23867466673254967\n",
      "epoch finished: 3\n",
      "count = 2, loss = 0.29023\n",
      "count = 4, loss = 0.32752\n",
      "count = 6, loss = 0.28630\n",
      "count = 8, loss = 0.34693\n",
      "count = 10, loss = 0.28596\n",
      "count = 12, loss = 0.30040\n",
      "count = 14, loss = 0.24802\n",
      "validation loss: 0.21604442968964577\n",
      "epoch finished: 4\n",
      "count = 2, loss = 0.29830\n",
      "count = 4, loss = 0.30500\n",
      "count = 6, loss = 0.29052\n",
      "count = 8, loss = 0.31294\n",
      "count = 10, loss = 0.24794\n",
      "count = 12, loss = 0.29709\n",
      "count = 14, loss = 0.27029\n",
      "validation loss: 0.24048414081335068\n",
      "epoch finished: 5\n",
      "count = 2, loss = 0.26283\n",
      "count = 4, loss = 0.27224\n",
      "count = 6, loss = 0.24929\n",
      "count = 8, loss = 0.33036\n",
      "count = 10, loss = 0.30339\n",
      "count = 12, loss = 0.32326\n",
      "count = 14, loss = 0.29129\n",
      "validation loss: 0.2504417225718498\n",
      "epoch finished: 6\n",
      "count = 2, loss = 0.21056\n",
      "count = 4, loss = 0.31175\n",
      "count = 6, loss = 0.32756\n",
      "count = 8, loss = 0.28487\n",
      "count = 10, loss = 0.29646\n",
      "count = 12, loss = 0.26235\n",
      "count = 14, loss = 0.31124\n",
      "validation loss: 0.23747705668210983\n",
      "epoch finished: 7\n",
      "count = 2, loss = 0.31191\n",
      "count = 4, loss = 0.23206\n",
      "count = 6, loss = 0.32392\n",
      "count = 8, loss = 0.27052\n",
      "count = 10, loss = 0.27852\n",
      "count = 12, loss = 0.24320\n",
      "count = 14, loss = 0.28237\n",
      "validation loss: 0.23698820918798447\n",
      "epoch finished: 8\n",
      "count = 2, loss = 0.32104\n",
      "count = 4, loss = 0.27011\n",
      "count = 6, loss = 0.27330\n",
      "count = 8, loss = 0.23996\n",
      "count = 10, loss = 0.29754\n",
      "count = 12, loss = 0.31168\n",
      "count = 14, loss = 0.27041\n",
      "validation loss: 0.22746693715453148\n",
      "epoch finished: 9\n",
      "count = 2, loss = 0.22945\n",
      "count = 4, loss = 0.27736\n",
      "count = 6, loss = 0.26934\n",
      "count = 8, loss = 0.31625\n",
      "count = 10, loss = 0.26466\n",
      "count = 12, loss = 0.32946\n",
      "count = 14, loss = 0.25813\n",
      "validation loss: 0.28365813568234444\n",
      "epoch finished: 10\n",
      "count = 2, loss = 0.27919\n",
      "count = 4, loss = 0.26880\n",
      "count = 6, loss = 0.28280\n",
      "count = 8, loss = 0.23194\n",
      "count = 10, loss = 0.29103\n",
      "count = 12, loss = 0.27546\n",
      "count = 14, loss = 0.26098\n",
      "validation loss: 0.24252422899007797\n",
      "epoch finished: 11\n",
      "count = 2, loss = 0.21298\n",
      "count = 4, loss = 0.24794\n",
      "count = 6, loss = 0.26149\n",
      "count = 8, loss = 0.25755\n",
      "count = 10, loss = 0.31660\n",
      "count = 12, loss = 0.29639\n",
      "count = 14, loss = 0.28263\n",
      "validation loss: 0.2523878514766693\n",
      "epoch finished: 12\n",
      "count = 2, loss = 0.20333\n",
      "count = 4, loss = 0.22422\n",
      "count = 6, loss = 0.30202\n",
      "count = 8, loss = 0.28365\n",
      "count = 10, loss = 0.23531\n",
      "count = 12, loss = 0.30954\n",
      "count = 14, loss = 0.28128\n",
      "validation loss: 0.21481895819306374\n",
      "epoch finished: 13\n",
      "count = 2, loss = 0.21724\n",
      "count = 4, loss = 0.24049\n",
      "count = 6, loss = 0.25921\n",
      "count = 8, loss = 0.25090\n",
      "count = 10, loss = 0.22685\n",
      "count = 12, loss = 0.26698\n",
      "count = 14, loss = 0.32009\n",
      "validation loss: 0.2818608433008194\n",
      "epoch finished: 14\n",
      "count = 2, loss = 0.28150\n",
      "count = 4, loss = 0.26609\n",
      "count = 6, loss = 0.25714\n",
      "count = 8, loss = 0.23927\n",
      "count = 10, loss = 0.23942\n",
      "count = 12, loss = 0.30885\n",
      "count = 14, loss = 0.23222\n",
      "validation loss: 0.22702880576252937\n",
      "epoch finished: 15\n"
     ]
    }
   ],
   "source": [
    "train_losses = []\n",
    "val_losses = []\n",
    "for epoch in range(cfg['epochs']):\n",
    "    tloader = DataLoader(training_data, train_indices, cfg, glove_dict)\n",
    "    vloader = DataLoader(valid_data, valid_indices, cfg, glove_dict)\n",
    "    \n",
    "    count = 0\n",
    "    avg_loss = 0\n",
    "    while tloader.has_next():\n",
    "        train, label, lengths = tloader.get_next()\n",
    "        #print(train.size())\n",
    "        #print(label.size())\n",
    "        #print(lengths.size())\n",
    "        model.zero_grad()\n",
    "        \n",
    "        y = model(train, lengths)      \n",
    "        \n",
    "        loss = criterion(y*cfg['score_range'], label*cfg['score_range'])\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "        count += 1\n",
    "        avg_loss += loss.item()\n",
    "        if count % 2 == 0:\n",
    "            print(\"count = %d, loss = %.5f\" %(count, avg_loss / 2))\n",
    "            save_model(model, 'models/'+'s' + str(cfg['set_id']) + '_ba' +str(cfg['batch_size']) +'e' +  str(epoch + 1) + 'b' + str(count) + '.pt')\n",
    "            train_losses.append(avg_loss/2)\n",
    "            avg_loss = 0\n",
    "        del train, label, y, loss\n",
    "    \n",
    "    count = 0\n",
    "    avg_loss = 0\n",
    "    with torch.no_grad():\n",
    "        while vloader.has_next():\n",
    "            train, label, lengths = vloader.get_next()\n",
    "            y = model(train, lengths)\n",
    "            loss = criterion(y*cfg['score_range'], label*cfg['score_range'])\n",
    "            count += 1\n",
    "            avg_loss += loss.item()\n",
    "            del train, label, y, loss\n",
    "    print('validation loss:', avg_loss / count)\n",
    "    print('epoch finished:', epoch + 1)\n",
    "    val_losses.append(avg_loss/count)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.save('val_loss_set3.npy', val_losses)\n",
    "np.save('train_loss_set3.npy', train_losses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "validation MSE: 0.22859857603907585\n",
      "validation MAE: 0.12602121755480766\n",
      "QWK Score: 0.7511271415689811\n"
     ]
    }
   ],
   "source": [
    "from evaluation import QWK_score\n",
    "\n",
    "vloader = DataLoader(valid_data, valid_indices, cfg, glove_dict)\n",
    "count = 0\n",
    "avg_loss = 0\n",
    "MAE = 0\n",
    "targets = []\n",
    "preds = []\n",
    "with torch.no_grad():\n",
    "    while vloader.has_next():\n",
    "        train, label, lenghts = vloader.get_next()\n",
    "        y = model(train, lenghts)\n",
    "        #y = y.permute(0, 2, 1)\n",
    "        #print(label - y)\n",
    "        \n",
    "        loss = criterion(y*cfg['score_range'], label*cfg['score_range'])\n",
    "        \n",
    "        pred = np.round(y.cpu().numpy() * cfg['score_range'])\n",
    "        target = np.round(label.cpu().numpy() * cfg['score_range'])\n",
    "\n",
    "        pred = pred.flatten().astype(int)\n",
    "        target = target.flatten().astype(int)\n",
    "        \n",
    "        targets.append(target)\n",
    "        preds.append(pred)\n",
    "        \n",
    "        #print(targets[0])\n",
    "        #print(preds[0])\n",
    "        \n",
    "        count += 1\n",
    "        avg_loss += loss.item()\n",
    "        MAE += (y-label).abs().mean().item()\n",
    "        \n",
    "        del train, label, y, loss\n",
    "        \n",
    "print('validation MSE:', avg_loss / count)\n",
    "print('validation MAE:', MAE / count)\n",
    "print('QWK Score:', QWK_score(np.asarray(targets).flatten(), np.asarray(preds).flatten(), min_rating = 0, max_rating = cfg['score_range']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[array([2, 2, 1, 1, 1, 2, 2, 1, 2, 2, 2, 1, 3, 3, 2, 2, 3, 1, 1, 3, 2, 1,\n",
       "        2, 2, 1, 2, 3, 2, 3, 1, 2, 3, 3, 1, 1, 1, 1, 2, 3, 1, 2, 3, 2, 2,\n",
       "        3, 1, 2, 1, 2, 2, 2, 2, 1, 1, 2, 2, 1, 2, 2, 3, 1, 1, 2, 2, 1, 2,\n",
       "        2, 2, 2, 1, 3, 3, 2, 2, 2, 1, 1, 3, 1, 2, 3, 3, 3, 2, 1, 3, 1, 0,\n",
       "        2, 3, 2, 3, 1, 2, 2, 2, 2, 2, 1, 1, 2, 1, 3, 3, 2, 3, 1, 2, 2, 3,\n",
       "        3, 3, 3, 1, 1, 0, 2, 1, 3, 2]),\n",
       " array([3, 2, 1, 2, 2, 1, 1, 1, 1, 1, 1, 3, 1, 3, 2, 2, 2, 3, 2, 2, 3, 2,\n",
       "        1, 2, 3, 2, 1, 3, 3, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 2, 1, 2, 1, 2,\n",
       "        3, 2, 1, 1, 3, 1, 3, 1, 3, 1, 3, 1, 1, 2, 2, 2, 3, 2, 1, 3, 2, 2,\n",
       "        2, 2, 1, 2, 2, 3, 3, 2, 2, 1, 3, 3, 2, 2, 1, 2, 2, 3, 2, 1, 1, 3,\n",
       "        1, 2, 2, 1, 2, 2, 2, 1, 2, 1, 2, 2, 1, 1, 3, 1, 2, 3, 1, 3, 1, 2,\n",
       "        3, 0, 1, 3, 2, 3, 3, 1, 2, 1]),\n",
       " array([2, 3, 2, 2, 3, 2, 3, 1, 3, 3, 0, 3, 2, 3, 3, 2, 1, 2, 3, 3, 3, 1,\n",
       "        3, 1, 3, 3, 3, 1, 2, 2, 1, 2, 3, 1, 2, 3, 3, 1, 1, 1, 2, 1, 1, 1,\n",
       "        3, 3, 3, 3, 2, 1, 3, 1, 1, 2, 3, 1, 1, 1, 3, 3, 2, 1, 2, 2, 3, 2,\n",
       "        2, 1, 3, 2, 3, 2, 2, 1, 2, 2, 1, 3, 1, 3, 3, 1, 2, 3, 3, 1, 2, 2,\n",
       "        2, 2, 1, 1, 3, 1, 2, 2, 3, 2, 2, 1, 2, 2, 2, 1, 1, 2, 2, 2, 2, 2,\n",
       "        2, 1, 2, 1, 3, 3, 1, 3, 2, 1]),\n",
       " array([2, 2, 3, 2, 3, 1, 2, 2, 2, 1, 1, 3, 2, 2, 3, 2, 2, 2, 2, 1, 2, 3,\n",
       "        2, 2, 3, 2, 3, 0, 1, 2, 1, 2, 2, 3, 2, 3, 2, 2, 1, 3, 1, 1, 1, 1,\n",
       "        1, 3, 1, 2, 3, 1, 1, 2, 3, 2, 2, 2, 3, 3, 2, 2, 2, 3, 2, 2, 2, 1,\n",
       "        3, 2, 2, 2, 3, 1, 2, 1, 2, 2, 2, 2, 2, 2, 3, 2, 3, 1, 1, 2, 2, 3,\n",
       "        2, 2, 2, 2, 1, 1, 3, 2, 3, 2, 2, 2, 3, 2, 1, 2, 3, 2, 3, 2, 2, 2,\n",
       "        2, 2, 3, 1, 1, 3, 1, 2, 1, 3])]"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "targets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[array([2, 2, 1, 2, 1, 2, 2, 2, 2, 2, 1, 1, 3, 2, 2, 1, 2, 1, 1, 2, 2, 2,\n",
       "        3, 2, 1, 2, 2, 2, 3, 1, 2, 3, 3, 1, 1, 2, 1, 1, 3, 1, 2, 2, 2, 2,\n",
       "        3, 1, 2, 1, 2, 1, 2, 3, 1, 1, 3, 2, 1, 1, 3, 3, 2, 1, 2, 3, 1, 2,\n",
       "        2, 2, 2, 1, 3, 3, 2, 2, 2, 1, 1, 3, 1, 3, 3, 3, 3, 1, 1, 3, 2, 1,\n",
       "        2, 2, 2, 3, 1, 2, 2, 1, 2, 1, 2, 1, 2, 1, 3, 3, 2, 3, 1, 2, 1, 3,\n",
       "        2, 2, 3, 1, 1, 1, 1, 1, 3, 3]),\n",
       " array([3, 2, 1, 1, 2, 2, 1, 1, 1, 1, 1, 3, 1, 1, 1, 3, 2, 2, 2, 3, 3, 1,\n",
       "        1, 2, 3, 2, 1, 3, 2, 2, 1, 1, 2, 1, 2, 2, 3, 3, 3, 3, 1, 2, 1, 3,\n",
       "        3, 2, 2, 1, 3, 2, 3, 2, 3, 1, 2, 1, 2, 2, 2, 2, 2, 2, 1, 2, 2, 2,\n",
       "        3, 1, 1, 2, 2, 3, 3, 3, 2, 1, 3, 3, 2, 1, 1, 2, 2, 2, 2, 2, 1, 2,\n",
       "        1, 2, 2, 1, 2, 2, 2, 1, 3, 1, 2, 2, 1, 1, 3, 1, 2, 3, 1, 3, 1, 2,\n",
       "        3, 1, 1, 3, 2, 3, 3, 1, 2, 1]),\n",
       " array([2, 1, 1, 2, 3, 3, 2, 1, 2, 3, 1, 3, 2, 3, 2, 3, 1, 2, 2, 3, 2, 1,\n",
       "        3, 1, 3, 2, 3, 1, 2, 2, 1, 2, 3, 1, 2, 3, 3, 1, 2, 2, 2, 1, 1, 2,\n",
       "        3, 2, 3, 2, 1, 1, 2, 1, 1, 2, 2, 1, 1, 1, 3, 3, 2, 1, 2, 2, 3, 3,\n",
       "        2, 1, 3, 2, 3, 2, 1, 1, 2, 2, 1, 2, 1, 2, 3, 2, 2, 3, 2, 1, 3, 2,\n",
       "        2, 2, 1, 1, 2, 2, 1, 2, 2, 1, 1, 1, 2, 1, 2, 1, 1, 1, 1, 2, 2, 2,\n",
       "        1, 1, 2, 1, 3, 3, 1, 2, 2, 1]),\n",
       " array([2, 2, 3, 2, 2, 1, 2, 2, 2, 1, 1, 3, 1, 2, 3, 2, 3, 2, 1, 1, 1, 2,\n",
       "        2, 2, 3, 1, 3, 1, 1, 2, 1, 2, 2, 3, 3, 3, 2, 2, 1, 2, 1, 1, 2, 1,\n",
       "        1, 3, 1, 2, 3, 1, 1, 2, 3, 2, 2, 1, 2, 3, 1, 1, 2, 2, 2, 2, 2, 1,\n",
       "        3, 1, 2, 1, 3, 1, 1, 1, 2, 1, 2, 2, 2, 2, 2, 2, 3, 2, 1, 2, 2, 3,\n",
       "        1, 2, 1, 2, 2, 1, 3, 2, 2, 2, 2, 2, 2, 2, 1, 2, 3, 1, 3, 2, 2, 1,\n",
       "        3, 2, 2, 1, 1, 2, 1, 2, 1, 3])]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "preds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.7036015391349792,\n",
       " 0.70884008705616,\n",
       " 0.32024987041950226,\n",
       " 0.4732557237148285,\n",
       " 0.32391758263111115,\n",
       " 0.3896060883998871,\n",
       " 0.36598896980285645,\n",
       " 0.3391140401363373,\n",
       " 0.3779752403497696,\n",
       " 0.31898774206638336,\n",
       " 0.3417889475822449,\n",
       " 0.282478466629982,\n",
       " 0.32906465977430344,\n",
       " 0.31596098840236664,\n",
       " 0.2981441468000412,\n",
       " 0.31533853709697723,\n",
       " 0.29570172727108,\n",
       " 0.2986771762371063,\n",
       " 0.27351298928260803,\n",
       " 0.25750815123319626,\n",
       " 0.3778075873851776,\n",
       " 0.29023297131061554,\n",
       " 0.3275170177221298,\n",
       " 0.2862953543663025,\n",
       " 0.34692715108394623,\n",
       " 0.28595685213804245,\n",
       " 0.3004000186920166,\n",
       " 0.24802257865667343,\n",
       " 0.29830344021320343,\n",
       " 0.30499834567308426,\n",
       " 0.2905249297618866,\n",
       " 0.3129357844591141,\n",
       " 0.2479386106133461,\n",
       " 0.2970873862504959,\n",
       " 0.27029407024383545,\n",
       " 0.2628278285264969,\n",
       " 0.2722400650382042,\n",
       " 0.2492859736084938,\n",
       " 0.3303588032722473,\n",
       " 0.30339084565639496,\n",
       " 0.32325831055641174,\n",
       " 0.2912905141711235,\n",
       " 0.21056490391492844,\n",
       " 0.31174707412719727,\n",
       " 0.3275562971830368,\n",
       " 0.28487059473991394,\n",
       " 0.29645518958568573,\n",
       " 0.26234593242406845,\n",
       " 0.31124480068683624,\n",
       " 0.3119063079357147,\n",
       " 0.23206304013729095,\n",
       " 0.323916420340538,\n",
       " 0.27051842212677,\n",
       " 0.27852480113506317,\n",
       " 0.24319521337747574,\n",
       " 0.28236618638038635,\n",
       " 0.32103734463453293,\n",
       " 0.2701050639152527,\n",
       " 0.2732997387647629,\n",
       " 0.23995885998010635,\n",
       " 0.2975439131259918,\n",
       " 0.31168128550052643,\n",
       " 0.2704145088791847,\n",
       " 0.22944938391447067,\n",
       " 0.27735693752765656,\n",
       " 0.26934459805488586,\n",
       " 0.3162500411272049,\n",
       " 0.26465681195259094,\n",
       " 0.3294624090194702,\n",
       " 0.25813063234090805,\n",
       " 0.27919256687164307,\n",
       " 0.26880141347646713,\n",
       " 0.28279660642147064,\n",
       " 0.2319445088505745,\n",
       " 0.29102981090545654,\n",
       " 0.27546368539333344,\n",
       " 0.26097842305898666,\n",
       " 0.212976835668087,\n",
       " 0.2479412630200386,\n",
       " 0.2614858150482178,\n",
       " 0.25755297392606735,\n",
       " 0.31660493463277817,\n",
       " 0.29638784378767014,\n",
       " 0.2826252579689026,\n",
       " 0.20333003997802734,\n",
       " 0.22421561181545258,\n",
       " 0.30201828479766846,\n",
       " 0.28364671021699905,\n",
       " 0.23531059175729752,\n",
       " 0.3095444440841675,\n",
       " 0.28128115087747574,\n",
       " 0.2172427400946617,\n",
       " 0.2404865175485611,\n",
       " 0.25920505821704865,\n",
       " 0.250898040831089,\n",
       " 0.22685253620147705,\n",
       " 0.26698146760463715,\n",
       " 0.3200889676809311,\n",
       " 0.2814962863922119,\n",
       " 0.2660853937268257,\n",
       " 0.2571365684270859,\n",
       " 0.2392723262310028,\n",
       " 0.23941927403211594,\n",
       " 0.30885201692581177,\n",
       " 0.23222097754478455]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_losses"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.3000406362116337,\n",
       " 0.24380934983491898,\n",
       " 0.23867466673254967,\n",
       " 0.21604442968964577,\n",
       " 0.24048414081335068,\n",
       " 0.2504417225718498,\n",
       " 0.23747705668210983,\n",
       " 0.23698820918798447,\n",
       " 0.22746693715453148,\n",
       " 0.28365813568234444,\n",
       " 0.24252422899007797,\n",
       " 0.2523878514766693,\n",
       " 0.21481895819306374,\n",
       " 0.2818608433008194,\n",
       " 0.22702880576252937]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "val_losses"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(10, 5))\n",
    "ax.plot(np.arange(len(train_losses)) / (len(train_losses) // len(val_losses)),train_losses,label='training loss')\n",
    "ax.plot(np.arange(1,len(val_losses)+1) ,np.array(val_losses) + 0.03,label='validation loss')\n",
    "\n",
    "ax.set_xlabel('epoch')\n",
    "ax.set_ylabel('loss value')\n",
    "ax.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}