09-train_SIMPLE_CNN_RNN_rel_pos_only_Glove300.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "606fa719",
   "metadata": {},
   "source": [
    "# Train 09-train_SIMPLE_CNN_RNN_rel_pos_only"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "a8a5bd39",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "from torch.nn.utils import clip_grad_norm_\n",
    "\n",
    "from tqdm import tqdm\n",
    "\n",
    "from rouge_score import rouge_scorer\n",
    "\n",
    "from utils.GloveMgr import GloveMgr\n",
    "from utils.Dataset import Dataset\n",
    "from utils.DataLoader import DataLoader\n",
    "from utils.preprocess_df import preprocess_df\n",
    "from utils.accuracy_nb_sent_per_doc import accuracy_nb_sent_per_doc_fn\n",
    "from utils.accuracy_prop_sent_per_doc import accuracy_prop_sent_per_doc_fn\n",
    "from utils.split_all_docs import split_all_docs\n",
    "\n",
    "\n",
    "from time import time\n",
    "\n",
    "import json\n",
    "import argparse"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "818410ea",
   "metadata": {},
   "outputs": [],
   "source": [
    "def is_notebook() -> bool:\n",
    "    try:\n",
    "        shell = get_ipython().__class__.__name__\n",
    "        if shell == 'ZMQInteractiveShell':\n",
    "            return True   # Jupyter notebook or qtconsole\n",
    "        elif shell == 'TerminalInteractiveShell':\n",
    "            return False  # Terminal running IPython\n",
    "        else:\n",
    "            return False  # Other type (?)\n",
    "    except NameError:\n",
    "        return False      # Probably standard Python interpreter"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "690fa032",
   "metadata": {},
   "outputs": [],
   "source": [
    "pd.set_option('display.max_columns', None)\n",
    "pd.set_option('display.max_rows', None)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "b2636e0b",
   "metadata": {},
   "source": [
    "## (Hyper-)parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "6e7aad39",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Parse args if script mode\n",
    "parser = argparse.ArgumentParser(description='extractive summary')\n",
    "\n",
    "parser.add_argument('-is_graphic',type=int,default=1,choices=[0,1])\n",
    "parser.add_argument('-gpu_num',type=int,default=1)\n",
    "parser.add_argument('-batch_size',type=int,default=8)\n",
    "parser.add_argument('-epochs',type=int,default=100)\n",
    "parser.add_argument('-dataset',type=str,default=\"./data/nyt_corpus_LDC2008T19_50\")\n",
    "parser.add_argument('-doc_col_name',type=str,default=\"docs\")\n",
    "parser.add_argument('-label_col_name',type=str,default=\"labels\")\n",
    "parser.add_argument('-sum_col_name',type=str,default=\"summaries\")\n",
    "parser.add_argument('-max_doc_len',type=int,default=50)\n",
    "\n",
    "args = None\n",
    "\n",
    "if is_notebook():\n",
    "    args = parser.parse_args(\"\")\n",
    "else:\n",
    "    args = parser.parse_args()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "824eacf5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "parse:\n",
      "is_graphic: True\n",
      "cuda_num: 1\n",
      "epochs 100\n",
      "batch_size 8\n"
     ]
    }
   ],
   "source": [
    "# parameters\n",
    "is_graphic = args.is_graphic != 0\n",
    "cuda_num = args.gpu_num\n",
    "embed_name = \"glove.6B.300\"\n",
    "\n",
    "# hyper-parameters\n",
    "vocab_size = 150000\n",
    "batch_size = args.batch_size\n",
    "epochs = args.epochs\n",
    "learning_rate = 1e-3\n",
    "early_stopping = 3\n",
    "model_name = \"09-train_SIMPLE_CNN_RNN_rel_pos_only_Glove300\"\n",
    "sub_folder_name = \"model_name__{}__time__{}__embed_name__{}__lr__{}__batch_size__{}__vocab_size__{}__cuda_num__{}\".format(model_name, time(), embed_name, learning_rate, batch_size, vocab_size, cuda_num)\n",
    "checkpoints_folder = \"./checkpoints/\" + sub_folder_name\n",
    "average_number_of_sentences_per_document = 3\n",
    "\n",
    "# print\n",
    "print(\"parse:\")\n",
    "print(\"is_graphic:\", is_graphic)\n",
    "print(\"cuda_num:\", cuda_num)\n",
    "print(\"epochs\", epochs)\n",
    "print(\"batch_size\", batch_size)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "c8ddd4d5",
   "metadata": {},
   "source": [
    "## PyTorch initialisation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "d09bc3ed",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of available GPUs: 6\n",
      "GPU 0: NVIDIA GeForce GTX 1080 Ti\n",
      "GPU 1: NVIDIA GeForce GTX 1080 Ti\n",
      "GPU 2: NVIDIA GeForce GTX 1080 Ti\n",
      "GPU 3: NVIDIA GeForce GTX 1080\n",
      "GPU 4: NVIDIA GeForce GTX 1080\n",
      "GPU 5: NVIDIA GeForce GTX 1080\n"
     ]
    }
   ],
   "source": [
    "# Check if a GPU is available\n",
    "if torch.cuda.is_available():\n",
    "    # Display the number of available GPUs\n",
    "    print(f\"Number of available GPUs: {torch.cuda.device_count()}\")\n",
    "    # Display the name of each GPU\n",
    "    for i in range(torch.cuda.device_count()):\n",
    "        print(f\"GPU {i}: {torch.cuda.get_device_name(i)}\")\n",
    "else:\n",
    "    print(\"No GPU available.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "b0c5786e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "device(type='cuda', index=1)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "if torch.cuda.is_available():  \n",
    "  dev = \"cuda:\" + str(cuda_num) \n",
    "else:  \n",
    "  dev = \"cpu\" \n",
    "\n",
    "device = torch.device(dev)\n",
    "device"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "3393fe4e",
   "metadata": {},
   "source": [
    "## Chargement des données"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "97644745",
   "metadata": {},
   "outputs": [],
   "source": [
    "glovemgr = GloveMgr(\"./data/glove.6B/glove.6B.300d.txt\", vocab_size=vocab_size)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "fe745c8d",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train = pd.read_json(args.dataset + \"/train.json\")\n",
    "train_dataset = Dataset(preprocess_df(df_train, glovemgr=glovemgr, is_sep_n=False, remove_stop_word=True, stemming=False, trunc_sent=50, padding_sent=50, trunc_doc=args.max_doc_len, doc_column_name=args.doc_col_name, labels_column_name=args.label_col_name))\n",
    "train_iter = DataLoader(dataset=train_dataset, batch_size=batch_size, shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "f8b39f3e",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_val = pd.read_json(args.dataset + \"/val.json\")\n",
    "val_dataset = Dataset(preprocess_df(df_val, glovemgr=glovemgr, is_sep_n=False, remove_stop_word=True, stemming=False, trunc_sent=50, padding_sent=50, trunc_doc=args.max_doc_len, doc_column_name=args.doc_col_name, labels_column_name=args.label_col_name))\n",
    "val_iter = DataLoader(dataset=val_dataset, batch_size=1, shuffle=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "0f086978",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_test = pd.read_json(args.dataset + \"/test.json\")\n",
    "test_dataset = Dataset(preprocess_df(df_test, glovemgr=glovemgr, is_sep_n=False, remove_stop_word=True, stemming=False, trunc_sent=50, padding_sent=50, trunc_doc=args.max_doc_len, doc_column_name=args.doc_col_name, labels_column_name=args.label_col_name))\n",
    "test_iter = DataLoader(dataset=test_dataset, batch_size=1, shuffle=False)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "fc41eb07",
   "metadata": {},
   "source": [
    "## Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "9683721f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from models.BasicModel import BasicModel\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from torch.autograd import Variable\n",
    "\n",
    "class SIMPLE_CNN_RNN(BasicModel):\n",
    "    def __init__(self, device, vocab_size, word_embed = None, dim_word_embed = 100):\n",
    "        super(SIMPLE_CNN_RNN, self).__init__(device)\n",
    "\n",
    "        self.device = device\n",
    "\n",
    "        # 10: relative position range size, with segment size = 10\n",
    "        self.rel_pos_emb = nn.Embedding(11, 100)\n",
    "\n",
    "        self.Wrel_pos = nn.Linear(100,1,bias=False)\n",
    "        self.bias = nn.Parameter(torch.empty(1).uniform_(-0.1, 0.1))\n",
    "\n",
    "    def avg_pool1d(self,x,seq_lens):\n",
    "        out = []\n",
    "        for index,t in enumerate(x):\n",
    "            if seq_lens[index] == 0:\n",
    "                t = t[:1]\n",
    "            else:\n",
    "                t = t[:seq_lens[index],:]\n",
    "            t = torch.t(t).unsqueeze(0)\n",
    "            out.append(torch.avg_pool1d(t,t.size(2)))\n",
    "        \n",
    "        out = torch.cat(out).squeeze(2)\n",
    "        return out\n",
    "\n",
    "    def forward(self, arr_x, doc_lens):\n",
    "        probs = []\n",
    "\n",
    "        max_doc_lens = max(doc_lens)\n",
    "\n",
    "        # Compute relative position embedding\n",
    "        rel_poses = torch.tensor([[int(round(position / 10))] for position in range(max_doc_lens)], dtype=torch.long)\n",
    "        rel_poses = rel_poses.to(self.device)\n",
    "        rel_poses = self.rel_pos_emb(rel_poses).squeeze(1)\n",
    "        rps = self.Wrel_pos(rel_poses)\n",
    "\n",
    "        zeros = torch.zeros((len(doc_lens), 1)).to(self.device)\n",
    "\n",
    "        # for each document, compute probabilities\n",
    "        for position in range(max_doc_lens):\n",
    "            prob = zeros + rps[position] + self.bias\n",
    "            prob = torch.sigmoid(prob)\n",
    "            probs.append(prob)\n",
    "\n",
    "        probs = torch.cat(probs).squeeze()\n",
    "        if len(probs.shape) == 0:\n",
    "            probs = probs.unsqueeze(0)\n",
    "        probs_filtered = []\n",
    "        for idx, doc_len in enumerate(doc_lens):\n",
    "            prob = probs[idx::len(doc_lens)]\n",
    "            prob = prob[:doc_len]\n",
    "            probs_filtered.append(prob)\n",
    "        probs_filtered = torch.cat(probs_filtered).squeeze()\n",
    "        if len(probs_filtered.shape) == 0:\n",
    "            probs_filtered = probs_filtered.unsqueeze(0)\n",
    "        #probs = probs.to(self.device)\n",
    "        return probs_filtered\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "380f8ba1",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = SIMPLE_CNN_RNN(device=device, vocab_size=vocab_size, word_embed=glovemgr.getEmbeddings(), dim_word_embed=300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "2ae1a236",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SIMPLE_CNN_RNN(\n",
       "  (rel_pos_emb): Embedding(11, 100)\n",
       "  (Wrel_pos): Linear(in_features=100, out_features=1, bias=False)\n",
       ")"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "4bc411ce",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_fn = nn.BCELoss()\n",
    "optimizer = torch.optim.Adamax(model.parameters(),lr=learning_rate)\n",
    "scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=1, gamma=0.5)\n",
    "mae_fn = nn.L1Loss()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "1852182a",
   "metadata": {},
   "source": [
    "## Train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "5c803c4f",
   "metadata": {},
   "outputs": [],
   "source": [
    "if not os.path.exists(checkpoints_folder):\n",
    "    os.makedirs(checkpoints_folder)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "c6618af5",
   "metadata": {},
   "outputs": [],
   "source": [
    "df_val[\"doc_splitted\"] = split_all_docs(df_val[args.doc_col_name], is_sep_n = False)\n",
    "val_set = df_val\n",
    "scorer = rouge_scorer.RougeScorer(['rouge1', 'rouge2', 'rougeL'], use_stemmer=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "0991d3c5",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 1: 100%|██████████| 13/13 [00:00<00:00, 19.64batch/s, accuracy=0.567, loss=0.872, mae=0.55] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1 : val loss = 0.755, val mae = 0.516, val accuracy = 0.746, r1 = 0.303, r2 = 0.088, rL = 0.177\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 2: 100%|██████████| 13/13 [00:00<00:00, 38.49batch/s, accuracy=0.568, loss=0.708, mae=0.495]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 2 : val loss = 0.673, val mae = 0.483, val accuracy = 0.746, r1 = 0.303, r2 = 0.088, rL = 0.177\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 3: 100%|██████████| 13/13 [00:00<00:00, 38.25batch/s, accuracy=0.566, loss=0.648, mae=0.468]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3 : val loss = 0.638, val mae = 0.466, val accuracy = 0.746, r1 = 0.303, r2 = 0.088, rL = 0.177\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 4: 100%|██████████| 13/13 [00:00<00:00, 38.88batch/s, accuracy=0.575, loss=0.621, mae=0.455]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 4 : val loss = 0.622, val mae = 0.457, val accuracy = 0.746, r1 = 0.303, r2 = 0.088, rL = 0.177\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 5: 100%|██████████| 13/13 [00:00<00:00, 38.79batch/s, accuracy=0.747, loss=0.608, mae=0.447]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 5 : val loss = 0.615, val mae = 0.453, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 6: 100%|██████████| 13/13 [00:00<00:00, 38.77batch/s, accuracy=0.752, loss=0.601, mae=0.444]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 6 : val loss = 0.611, val mae = 0.451, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 7: 100%|██████████| 13/13 [00:00<00:00, 40.63batch/s, accuracy=0.754, loss=0.599, mae=0.443]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 7 : val loss = 0.609, val mae = 0.450, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 8: 100%|██████████| 13/13 [00:00<00:00, 38.80batch/s, accuracy=0.746, loss=0.598, mae=0.442]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 8 : val loss = 0.608, val mae = 0.450, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 9: 100%|██████████| 13/13 [00:00<00:00, 38.95batch/s, accuracy=0.741, loss=0.598, mae=0.442]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 9 : val loss = 0.608, val mae = 0.450, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 10: 100%|██████████| 13/13 [00:00<00:00, 39.05batch/s, accuracy=0.748, loss=0.596, mae=0.441]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 10 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 11: 100%|██████████| 13/13 [00:00<00:00, 37.63batch/s, accuracy=0.744, loss=0.595, mae=0.44] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 11 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 12: 100%|██████████| 13/13 [00:00<00:00, 38.09batch/s, accuracy=0.745, loss=0.595, mae=0.44] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 12 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 13: 100%|██████████| 13/13 [00:00<00:00, 37.57batch/s, accuracy=0.749, loss=0.594, mae=0.44] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 13 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 14: 100%|██████████| 13/13 [00:00<00:00, 38.24batch/s, accuracy=0.745, loss=0.596, mae=0.441]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 14 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 15: 100%|██████████| 13/13 [00:00<00:00, 38.60batch/s, accuracy=0.75, loss=0.595, mae=0.44]  \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 15 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 16: 100%|██████████| 13/13 [00:00<00:00, 37.69batch/s, accuracy=0.745, loss=0.596, mae=0.441]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 16 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 17: 100%|██████████| 13/13 [00:00<00:00, 38.68batch/s, accuracy=0.748, loss=0.596, mae=0.441]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 17 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 18: 100%|██████████| 13/13 [00:00<00:00, 39.13batch/s, accuracy=0.741, loss=0.597, mae=0.441]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 18 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 19: 100%|██████████| 13/13 [00:00<00:00, 39.16batch/s, accuracy=0.745, loss=0.596, mae=0.441]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 19 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 20: 100%|██████████| 13/13 [00:00<00:00, 36.60batch/s, accuracy=0.745, loss=0.596, mae=0.441]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 20 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 21: 100%|██████████| 13/13 [00:00<00:00, 38.77batch/s, accuracy=0.749, loss=0.596, mae=0.441]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 21 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 22: 100%|██████████| 13/13 [00:00<00:00, 38.73batch/s, accuracy=0.747, loss=0.597, mae=0.441]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 22 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Epoch 23: 100%|██████████| 13/13 [00:00<00:00, 38.65batch/s, accuracy=0.75, loss=0.596, mae=0.441] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 23 : val loss = 0.608, val mae = 0.449, val accuracy = 0.866, r1 = 0.490, r2 = 0.293, rL = 0.375\n",
      "Training duration = 30.974300861358643\n"
     ]
    }
   ],
   "source": [
    "arr_train_loss = []\n",
    "arr_train_mae = []\n",
    "arr_train_acc = []\n",
    "arr_val_loss = []\n",
    "arr_val_mae = []\n",
    "arr_val_acc = []\n",
    "\n",
    "t1 = time()\n",
    "for epoch in range(1, epochs+1):\n",
    "    # train 1 epoch\n",
    "    model.train()\n",
    "    nb_batch_train = 0\n",
    "    total_train_loss = 0\n",
    "    total_train_mae = 0\n",
    "    total_train_acc = 0\n",
    "    with tqdm(train_iter, unit=\"batch\", total=len(train_iter)) as tepoch:\n",
    "        for batch in tepoch:\n",
    "            tepoch.set_description(f\"Epoch {epoch}\")\n",
    "            features = []\n",
    "            doc_lens = []\n",
    "\n",
    "            for j in range(batch_size):\n",
    "                doc_lens.append(len(batch[j][\"docs\"]))\n",
    "                features = features + batch[j][\"docs\"]\n",
    "            \n",
    "            features = torch.tensor(features, dtype=torch.long).to(device)\n",
    "\n",
    "            targets = [torch.tensor(batch[j][\"labels\"], dtype=torch.float) for j in range(batch_size)]\n",
    "            targets = torch.cat(targets)\n",
    "            targets = targets.to(device)\n",
    "            \n",
    "            probs = model(features, doc_lens)\n",
    "            \n",
    "            loss = loss_fn(probs, targets)\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            clip_grad_norm_(model.parameters(), 1.0)\n",
    "            optimizer.step()\n",
    "            nb_batch_train += 1\n",
    "            total_train_loss += loss.item()\n",
    "            total_train_mae += mae_fn(probs, targets).item()\n",
    "            total_train_acc += accuracy_prop_sent_per_doc_fn(probs=probs.cpu().detach().numpy(), targets=targets.cpu().detach().numpy(), doc_lens=doc_lens)\n",
    "            tepoch.set_postfix(loss=total_train_loss/nb_batch_train, mae=total_train_mae/nb_batch_train, accuracy=total_train_acc/nb_batch_train)\n",
    "\n",
    "    # Save model\n",
    "    model.save(checkpoints_folder + \"/\" + model_name + \"-\" + str(epoch) + \".pt\")\n",
    "\n",
    "    # Show train and val score\n",
    "    model.eval()\n",
    "    nb_batch_val = 0\n",
    "    total_val_loss = 0\n",
    "    total_val_mae = 0\n",
    "    total_val_acc = 0\n",
    "\n",
    "    total_r1 = 0\n",
    "    total_r2 = 0\n",
    "    total_rl = 0\n",
    "\n",
    "    for i, batch in enumerate(val_iter):\n",
    "        features = []\n",
    "        doc_lens = []\n",
    "\n",
    "        for j in range(1):\n",
    "            doc_lens.append(len(batch[j][\"docs\"]))\n",
    "            features = features + batch[j][\"docs\"]\n",
    "            \n",
    "        features = torch.tensor(features, dtype=torch.long).to(device)\n",
    "\n",
    "        targets = [torch.tensor(batch[j][\"labels\"], dtype=torch.float) for j in range(1)]\n",
    "        targets = torch.cat(targets)\n",
    "        targets = targets.to(device)\n",
    "        \n",
    "        y_pred = model(features, doc_lens)\n",
    "\n",
    "        loss = loss_fn(y_pred, targets)\n",
    "\n",
    "        nb_batch_val += 1\n",
    "        total_val_loss += loss.item()\n",
    "\n",
    "        doc = val_set[\"doc_splitted\"].iloc[i]\n",
    "        summaries = val_set[args.sum_col_name].iloc[i]\n",
    "\n",
    "        indices = torch.argsort(y_pred, descending=True)\n",
    "\n",
    "        y_pred_thresh = []\n",
    "\n",
    "        txt = \"\"\n",
    "\n",
    "        for j in range(min(len(doc), y_pred.shape[0])):\n",
    "            txt = txt + \". \" + doc[indices[j]]\n",
    "            y_pred_thresh.append(indices[j])\n",
    "            if len(txt) >= len(summaries):\n",
    "                break\n",
    "\n",
    "        y_pred_thresh.sort()\n",
    "\n",
    "        txt = \"\"\n",
    "\n",
    "        for j in y_pred_thresh:\n",
    "            txt = txt + \". \" + doc[j]\n",
    "\n",
    "        n = min(len(txt), len(summaries))\n",
    "\n",
    "        while n < len(txt) and txt[n].isalnum():\n",
    "            n += 1\n",
    "\n",
    "        txt = txt[:n]\n",
    "\n",
    "        #assert len(txt) - len(summaries) <= 20\n",
    "\n",
    "        scores = scorer.score(summaries, txt)\n",
    "        total_r1 += scores[\"rouge1\"].fmeasure\n",
    "        total_r2 += scores[\"rouge2\"].fmeasure\n",
    "        total_rl += scores[\"rougeL\"].fmeasure\n",
    "\n",
    "        probs = y_pred.tolist() # compute_probs(y_pred)\n",
    "        probs_tensor = y_pred #torch.tensor(probs, dtype=torch.float).to(device)\n",
    "\n",
    "        total_val_mae += mae_fn(probs_tensor, targets).item()\n",
    "        total_val_acc += accuracy_nb_sent_per_doc_fn(probs=probs, targets=targets.cpu().detach().numpy(), doc_lens=doc_lens, average_number_of_sentences_per_document=average_number_of_sentences_per_document)\n",
    "\n",
    "    arr_train_loss.append(total_train_loss / nb_batch_train)\n",
    "    arr_train_mae.append(total_train_mae / nb_batch_train)\n",
    "    arr_train_acc.append(total_train_acc / nb_batch_train)\n",
    "\n",
    "    arr_val_loss.append(total_val_loss / nb_batch_val)\n",
    "    arr_val_mae.append(total_val_mae / nb_batch_val)\n",
    "    arr_val_acc.append(total_val_acc / nb_batch_val)\n",
    "    \n",
    "    print(\"Epoch {} : val loss = {:.3f}, val mae = {:.3f}, val accuracy = {:.3f}, r1 = {:.3f}, r2 = {:.3f}, rL = {:.3f}\".format(epoch, total_val_loss / nb_batch_val, total_val_mae / nb_batch_val, total_val_acc / nb_batch_val, total_r1 / nb_batch_val, total_r2 / nb_batch_val, total_rl / nb_batch_val))\n",
    "\n",
    "    if len(arr_val_loss) >= early_stopping+1:\n",
    "        if min(arr_val_loss[-early_stopping:]) >= arr_val_loss[-(early_stopping+1)]:\n",
    "            break\n",
    "\n",
    "    # update lr\n",
    "    scheduler.step()\n",
    "\n",
    "t2 = time()\n",
    "print(\"Training duration =\", t2-t1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "2f39ceed",
   "metadata": {},
   "outputs": [],
   "source": [
    "training_metrics = {}\n",
    "training_metrics[\"duration\"]   = t2 - t1\n",
    "training_metrics[\"train_loss\"] = arr_train_loss\n",
    "training_metrics[\"train_mae\"]  = arr_train_mae\n",
    "training_metrics[\"train_acc\"]  = arr_train_acc\n",
    "training_metrics[\"val_loss\"]   = arr_val_loss\n",
    "training_metrics[\"val_mae\"]    = arr_val_mae\n",
    "training_metrics[\"val_acc\"]    = arr_val_acc\n",
    "\n",
    "# Save to file in JSON format\n",
    "\n",
    "with open(checkpoints_folder + \"/training_metrics.json\", 'w') as fp:\n",
    "    json.dump(training_metrics, fp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "8cd8fbee",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAkAAAAHHCAYAAABXx+fLAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8/fFQqAAAACXBIWXMAAA9hAAAPYQGoP6dpAABPhUlEQVR4nO3deXhU9d3+8feZmcxkIQshKxAIiyAoi6JEFpc+oqAWl1pFpFWpy6+KrUrRigq4VGm1UuojlS5StU+17ksLpWJasCqLghuK7MiaBALZk5lk5vz+mMwkgRCyzplk7td1zTUnZ75n8jmEMDff5RzDNE0TERERkQhis7oAERERkVBTABIREZGIowAkIiIiEUcBSERERCKOApCIiIhEHAUgERERiTgKQCIiIhJxFIBEREQk4igAiYiISMRRABIREZGIowAkIp3Kc889h2EYfPLJJ1aXIiKdmAKQiIiIRBwFIBEREYk4CkAi0uV8+umnXHTRRSQkJNCtWzfOP/981qxZ06BNdXU1Dz30ECeddBLR0dH06NGD8ePHs2LFimCbvLw8pk+fTu/evXG5XGRmZnLZZZexa9euEJ+RiLQ3h9UFiIi0p6+++oqzzz6bhIQE7rnnHqKiovj973/Peeedx6pVq8jJyQHgwQcfZP78+dx0002MHj2akpISPvnkEzZs2MAFF1wAwJVXXslXX33FT37yE7KzsykoKGDFihXs3r2b7OxsC89SRNrKME3TtLoIEZHmeu6555g+fToff/wxZ5xxxjGvX3HFFSxbtoxNmzbRv39/AA4cOMDgwYM57bTTWLVqFQAjR46kd+/e/OMf/2j0+xQVFdG9e3eeeOIJZs2a1XEnJCKW0BCYiHQZXq+Xd999l8svvzwYfgAyMzO59tpr+eCDDygpKQEgKSmJr776iq1btzb6XjExMTidTlauXMmRI0dCUr+IhI4CkIh0GQcPHqSiooLBgwcf89qQIUPw+Xzs2bMHgIcffpiioiIGDRrEsGHDuPvuu/niiy+C7V0uF7/61a/45z//SXp6Oueccw6PP/44eXl5ITsfEek4CkAiEpHOOecctm/fzpIlSzj11FP505/+xOmnn86f/vSnYJs777yTLVu2MH/+fKKjo5kzZw5Dhgzh008/tbByEWkPCkAi0mWkpqYSGxvL5s2bj3ntm2++wWazkZWVFdyXnJzM9OnTeemll9izZw/Dhw/nwQcfbHDcgAED+NnPfsa7777Lxo0b8Xg8PPnkkx19KiLSwRSARKTLsNvtXHjhhbz99tsNlqrn5+fz4osvMn78eBISEgAoLCxscGy3bt0YOHAgbrcbgIqKCqqqqhq0GTBgAPHx8cE2ItJ5aRm8iHRKS5YsYfny5cfsf/DBB1mxYgXjx4/ntttuw+Fw8Pvf/x63283jjz8ebDd06FDOO+88Ro0aRXJyMp988gmvvfYat99+OwBbtmzh/PPP5+qrr2bo0KE4HA7efPNN8vPzueaaa0J2niLSMbQMXkQ6lcAy+OPZs2cPBw8eZPbs2Xz44Yf4fD5ycnJ49NFHGTNmTLDdo48+yjvvvMOWLVtwu9307duXH/7wh9x9991ERUVRWFjIvHnzyM3NZc+ePTgcDk4++WR+9rOfcdVVV4XiVEWkAykAiYiISMTRHCARERGJOApAIiIiEnEUgERERCTiKACJiIhIxFEAEhERkYijACQiIiIRRxdCbITP52P//v3Ex8djGIbV5YiIiEgzmKZJaWkpPXv2xGZruo9HAagR+/fvb3C/IBEREek89uzZQ+/evZtsowDUiPj4eMD/Bxi4b5CIiIiEt5KSErKysoKf401RAGpEYNgrISFBAUhERKSTac70FU2CFhERkYijACQiIiIRRwFIREREIo7mALWB1+ulurra6jI6JafTecIliiIiIh1FAagVTNMkLy+PoqIiq0vptGw2G/369cPpdFpdioiIRCAFoFYIhJ+0tDRiY2N1scQWClxo8sCBA/Tp00d/fiIiEnIKQC3k9XqD4adHjx5Wl9Nppaamsn//fmpqaoiKirK6HBERiTCahNFCgTk/sbGxFlfSuQWGvrxer8WViIhIJFIAaiUN27SN/vxERMRKCkAiIiIScRSApFWys7NZuHCh1WWIiIi0iiZBR5DzzjuPkSNHtktw+fjjj4mLi2t7USIiIhZQAAohn2lS4/VhYBDlCL/ON9M08Xq9OBwn/muRmpoagopEREQ6Rvh9CndhBSVVfJNXSkGZO+Tf+4YbbmDVqlX89re/xTAMDMPgueeewzAM/vnPfzJq1ChcLhcffPAB27dv57LLLiM9PZ1u3bpx5pln8t577zV4v6OHwAzD4E9/+hNXXHEFsbGxnHTSSbzzzjshPksREZHmUQBqB6ZpUuGpOeGj2mtSVe2ltLK6We1P9DBNs9k1/va3v2XMmDHcfPPNHDhwgAMHDpCVlQXAvffeyy9/+Us2bdrE8OHDKSsr4+KLLyY3N5dPP/2USZMmMXnyZHbv3t3k93jooYe4+uqr+eKLL7j44ouZNm0ahw8fbtOfrYiISEfQEFg7qKz2MnTuv0L+fb9+eCKxzub9CBMTE3E6ncTGxpKRkQHAN998A8DDDz/MBRdcEGybnJzMiBEjgl8/8sgjvPnmm7zzzjvcfvvtx/0eN9xwA1OnTgXgscce46mnnmLdunVMmjSpxecmIiLSkdQDJJxxxhkNvi4rK2PWrFkMGTKEpKQkunXrxqZNm07YAzR8+PDgdlxcHAkJCRQUFHRIzSIiIm2hHqB2EBNl5+uHJ56wnafGx5b8UgwMhvaMb/PFAGOi7G06PuDo1VyzZs1ixYoV/PrXv2bgwIHExMTw/e9/H4/H0+T7HH1LC8Mw8Pl87VKjiIhIe1IAageGYTRrKComyiS6NrQ4HXai7KHtgHM6nc269cSHH37IDTfcwBVXXAH4e4R27drVwdWJiIiEjobAQsgwDBy1oafGG/qekezsbNauXcuuXbs4dOjQcXtnTjrpJN544w0+++wzPv/8c6699lr15IiISJeiABRiUTb/sFe1t/kruNrLrFmzsNvtDB06lNTU1OPO6VmwYAHdu3dn7NixTJ48mYkTJ3L66aeHuFoREZGOY5gtWUsdIUpKSkhMTKS4uJiEhIQGr1VVVbFz50769etHdHR0i99716FySqqq6ZUUQ49urvYqudNp65+jiIjI0Zr6/D6aeoBCLMpuXQ+QiIiI+CkAhZiVc4BERETETwEoxII9QD71AImIiFhFASjEAkvfq9UDJCIiYhkFoBBz2AJDYOoBEhERsYoCUIgFhsBqfD58WoAnIiJiCQWgELPbjOAtMDQRWkRExBoKQCFmGIaWwouIiFhMAcgCUTZNhBYREbGSApAFHJ20Byg7O5uFCxdaXYaIiEibKQBZILAUvkY3GBUREbGEApAFNAdIRETEWgpAFrDiYoh/+MMf6NmzJ76jep0uu+wyfvSjH7F9+3Yuu+wy0tPT6datG2eeeSbvvfdeyOoTEREJJQWg9mCa4Clv9sNRU4lRXYG3qvnHNPpowXWErrrqKgoLC/nPf/4T3Hf48GGWL1/OtGnTKCsr4+KLLyY3N5dPP/2USZMmMXnyZHbv3t0Rf2IiIiKWclhdQJdQXQGP9Wx2827AsPb4vvftB2dcs5p2796diy66iBdffJHzzz8fgNdee42UlBS+853vYLPZGDFiRLD9I488wptvvsk777zD7bff3h7VioiIhA3Le4AWLVpEdnY20dHR5OTksG7duibbL1y4kMGDBxMTE0NWVhZ33XUXVVVVwdcffPBBDMNo8Dj55JM7+jQ6hWnTpvH666/jdrsB+Otf/8o111yDzWajrKyMWbNmMWTIEJKSkujWrRubNm1SD5CIiHRJlvYAvfzyy8ycOZPFixeTk5PDwoULmThxIps3byYtLe2Y9i+++CL33nsvS5YsYezYsWzZsoUbbrgBwzBYsGBBsN0pp5zSYP6Kw9HBpxkV6++NaYGv95fgNU1OSutGdJS99d+3BSZPnoxpmixdupQzzzyT//73v/zmN78BYNasWaxYsYJf//rXDBw4kJiYGL7//e/j8XhaV5uIiEgYszQALViwgJtvvpnp06cDsHjxYpYuXcqSJUu49957j2n/0UcfMW7cOK699lrAf12aqVOnsnbt2gbtHA4HGRkZHX8CAYbR7KGoAHu0j5oaLzX2GHBGdVBhDUVHR/O9732Pv/71r2zbto3Bgwdz+umnA/Dhhx9yww03cMUVVwBQVlbGrl27QlKXiIhIqFk2BObxeFi/fj0TJkyoK8ZmY8KECaxevbrRY8aOHcv69euDw2Q7duxg2bJlXHzxxQ3abd26lZ49e9K/f3+mTZt2wmEct9tNSUlJg0dHs2op/LRp04Ihc9q0acH9J510Em+88QafffYZn3/+Oddee+0xK8ZERES6CssC0KFDh/B6vaSnpzfYn56eTl5eXqPHXHvttTz88MOMHz+eqKgoBgwYwHnnncd9990XbJOTk8Nzzz3H8uXLeeaZZ9i5cydnn302paWlx61l/vz5JCYmBh9ZWVntc5JNCC6FD3HI+J//+R+Sk5PZvHlzsCcN/L1x3bt3Z+zYsUyePJmJEycGe4dERES6mk61CmzlypU89thj/O53vyMnJ4dt27Zxxx138MgjjzBnzhwALrroomD74cOHk5OTQ9++fXnllVe48cYbG33f2bNnM3PmzODXJSUlHR6CArfDqAlxD5DNZmP//mPnK2VnZ/Pvf/+7wb4ZM2Y0+FpDYiIi0lVYFoBSUlKw2+3k5+c32J+fn3/c+Ttz5szhhz/8ITfddBMAw4YNo7y8nFtuuYX7778fm+3YDq2kpCQGDRrEtm3bjluLy+XC5XK14WxazoqLIYqIiIifZUNgTqeTUaNGkZubG9zn8/nIzc1lzJgxjR5TUVFxTMix2/0rqMzjXBSwrKyM7du3k5mZ2U6Vtw/dDkNERMQ6lg6BzZw5k+uvv54zzjiD0aNHs3DhQsrLy4Orwq677jp69erF/PnzAf8y7gULFnDaaacFh8DmzJnD5MmTg0Fo1qxZTJ48mb59+7J//37mzZuH3W5n6tSplp1nYxw29QCJiIhYxdIANGXKFA4ePMjcuXPJy8tj5MiRLF++PDgxevfu3Q16fB544AEMw+CBBx5g3759pKamMnnyZB599NFgm7179zJ16lQKCwtJTU1l/PjxrFmzhtTU1JCfX1OCd4T3mpimiWEYFlckIiISOQzzeGNHEaykpITExESKi4tJSEho8FpVVRU7d+4kOzubmJiYVn8Pn2mycV8xAEMyE4KBKFJUVlaya9cu+vXrR3R0tNXliIhIF9DU5/fRIutTtx1ERfkvWlhRUdGm97EZRnAYrCYCh8ECV5gODF2KiIiEUqdaBh8O7HY7SUlJFBQUABAbG9vq4Subrxqzxkt5RSWGLzRXgw4HPp+PgwcPEhsb2/G3KREREWmEPn1aIbBMPxCCWquwzE1ltY/qoijiXJH1o7DZbPTp00dzn0RExBKR9anbTgzDIDMzk7S0NKqrq1v9Pm++u4WlX+bxw7P6csO4fu1YYfhzOp2NXrdJREQkFBSA2sBut7dpDktsbDT7Sr3sKqrWRGAREZEQ0n/BLZSR4A89+SVVFlciIiISWRSALJQeDEBuiysRERGJLApAFkpXD5CIiIglFIAslJ7gvwFrYbkHT03kXQtIRETEKgpAFkqOcwZvilpQql4gERGRUFEAspBhGKTFax6QiIhIqCkAWSwwDKZ5QCIiIqGjAGSxjERNhBYREQk1BSCLaQhMREQk9BSALKYeIBERkdBTALKY5gCJiIiEngKQxQIXQ8xTABIREQkZBSCLBQJQgeYAiYiIhIwCkMUCAajMXUOZu8biakRERCKDApDFurkcdHM5AM0DEhERCRUFoDCQponQIiIiIaUAFAYydFd4ERGRkFIACgPpCboYooiISCgpAIWB4FL4YvUAiYiIhIICUBgIXAyxoFQBSEREJBQUgMKAeoBERERCSwEoDGgOkIiISGgpAIWB+kNgpmlaXI2IiEjXpwAUBtLi/T1A1V6Tw+Uei6sRERHp+hSAwoDTYaNHnBPQMJiIiEgoKACFiXRdDFFERCRkFIDCRLpuhyEiIhIyCkBhIrgUXgFIRESkwykAhQkthRcREQkdBaAwEQhABeoBEhER6XAKQGEiI9E/B0hDYCIiIh1PAShMBK4FpCEwERGRjqcAFCYyEv0BqLDcTbXXZ3E1IiIiXZsCUJhIjnXisBmYJhwsVS+QiIhIR1IAChM2m0FavOYBiYiIhIICUBhJT9RKMBERkVBQAAoj6ZoILSIiEhIKQGEkMBFaQ2AiIiIdSwEojKTpfmAiIiIhoQAURjJ0R3gREZGQUAAKI7ofmIiISGgoAIWR9MAQWLF6gERERDqSAlAYCfQAlbprKHfXWFyNiIhI16UAFEa6uRzEOu0AFOhq0CIiIh3G8gC0aNEisrOziY6OJicnh3Xr1jXZfuHChQwePJiYmBiysrK46667qKpqOGTU0vcMF4ZhBCdC52kYTEREpMNYGoBefvllZs6cybx589iwYQMjRoxg4sSJFBQUNNr+xRdf5N5772XevHls2rSJZ599lpdffpn77ruv1e8ZbgJL4QtKFYBEREQ6iqUBaMGCBdx8881Mnz6doUOHsnjxYmJjY1myZEmj7T/66CPGjRvHtddeS3Z2NhdeeCFTp05t0MPT0vcMN+nqARIREelwlgUgj8fD+vXrmTBhQl0xNhsTJkxg9erVjR4zduxY1q9fHww8O3bsYNmyZVx88cWtfk8At9tNSUlJg4dVMrQUXkREpMM5rPrGhw4dwuv1kp6e3mB/eno633zzTaPHXHvttRw6dIjx48djmiY1NTX8+Mc/Dg6BteY9AebPn89DDz3UxjNqH2m6GKKIiEiHs3wSdEusXLmSxx57jN/97nds2LCBN954g6VLl/LII4+06X1nz55NcXFx8LFnz552qrjldDVoERGRjmdZD1BKSgp2u538/PwG+/Pz88nIyGj0mDlz5vDDH/6Qm266CYBhw4ZRXl7OLbfcwv3339+q9wRwuVy4XK42nlH7CF4MUZOgRUREOoxlPUBOp5NRo0aRm5sb3Ofz+cjNzWXMmDGNHlNRUYHN1rBku91/3RzTNFv1nuGm/u0wTNO0uBoREZGuybIeIICZM2dy/fXXc8YZZzB69GgWLlxIeXk506dPB+C6666jV69ezJ8/H4DJkyezYMECTjvtNHJycti2bRtz5sxh8uTJwSB0ovcMd4Fl8J4aH0UV1XSPc1pckYiISNdjaQCaMmUKBw8eZO7cueTl5TFy5EiWL18enMS8e/fuBj0+DzzwAIZh8MADD7Bv3z5SU1OZPHkyjz76aLPfM9y5HHa6x0ZxpKKavJIqBSAREZEOYJgaZzlGSUkJiYmJFBcXk5CQEPLvP2nh+3yTV8pz08/kvMFpIf/+IiIinVFLPr871SqwSJGulWAiIiIdSgEoDOliiCIiIh1LASgMBZfCqwdIRESkQygAhaH0RA2BiYiIdCQFoDCUHq8hMBERkY6kABSGgneEVw+QiIhIh1AACkPpif45QIfK3NR4fRZXIyIi0vUoAIWhHnEu7DYD04SDZRoGExERaW8KQGHIbjNIiw+sBFMAEhERaW8KQGEqTRdDFBER6TAKQGEqQ9cCEhER6TAKQGFKt8MQERHpOApAYSq4FL5Yc4BERETamwJQmAoEoIJS9QCJiIi0NwWgMBW4H1hesQKQiIhIe1MAClMZmgMkIiLSYRSAwlRgGXxJVQ2VHq/F1YiIiHQtCkBhKiHaQUyUHVAvkIiISHtTAApThmEE5wEpAImIiLQvBaAwlqa7wouIiHQIBaAwFpgIXaD7gYmIiLQrBaAwFlwKrx4gERGRdqUAFMZ0OwwREZGOoQAUxtI1BCYiItIhFIDCWEaiJkGLiIh0BAWgMJYeXzcEZpqmxdWIiIh0HQpAYSytdhK0u8ZHcWW1xdWIiIh0HQpAYSw6yk5SbBQA+ZoHJCIi0m4UgMJcYBhM84BERETajwJQmEtP1FJ4ERGR9qYAFObS4/3zgAoUgERERNqNAlCY01J4ERGR9qcAFObSgleD1iRoERGR9qIAFOYCQ2CaAyQiItJ+FIDCXIYmQYuIiLQ7BaAwF7gf2MFSNzVen8XViIiIdA0KQGEupZsLmwE+EwrLPVaXIyIi0iUoAIU5u80gVfOARERE2pUCUCeQUTsMllesACQiItIeFIA6geBS+FIthRcREWkPCkCdQHrtXeHz1QMkIiLSLhSAOoGMBC2FFxERaU8KQJ1AYAhMt8MQERFpHwpAnUCgB6hAt8MQERFpFwpAnUB6cBK0eoBERETagwJQqNV4oKq4RYcEeoCKKqqpqvZ2RFUiIiIRRQEolFY9AfN7wYe/bdFhCTEOXA7/j0rDYCIiIm2nABRKcT3A64F9G1p0mGEYwWEwTYQWERFpOwWgUOp5uv95/6dgmi06VEvhRURE2k9YBKBFixaRnZ1NdHQ0OTk5rFu37rhtzzvvPAzDOOZxySWXBNvccMMNx7w+adKkUJxK09KGgt0JVUVwZGfLDk3Q/cBERETai+UB6OWXX2bmzJnMmzePDRs2MGLECCZOnEhBQUGj7d944w0OHDgQfGzcuBG73c5VV13VoN2kSZMatHvppZdCcTpNczghY5h/u4XDYOoBEhERaT+WB6AFCxZw8803M336dIYOHcrixYuJjY1lyZIljbZPTk4mIyMj+FixYgWxsbHHBCCXy9WgXffu3UNxOidWfxisBYJL4TUJWkREpM0sDUAej4f169czYcKE4D6bzcaECRNYvXp1s97j2Wef5ZprriEuLq7B/pUrV5KWlsbgwYO59dZbKSwsbNfaW63naf7nlgagRE2CFhERaS8OK7/5oUOH8Hq9pKenN9ifnp7ON998c8Lj161bx8aNG3n22Wcb7J80aRLf+9736NevH9u3b+e+++7joosuYvXq1djt9mPex+1243bX9ayUlJS08oyaoVegB+gz8HnBdmw9jUmP988BKlAAEhERaTNLA1BbPfvsswwbNozRo0c32H/NNdcEt4cNG8bw4cMZMGAAK1eu5Pzzzz/mfebPn89DDz3U4fUCkDIIouKguhwObYW0k5t1WP1l8KZpYhhGR1YpIiLSpVk6BJaSkoLdbic/P7/B/vz8fDIyMpo8try8nL/97W/ceOONJ/w+/fv3JyUlhW3btjX6+uzZsykuLg4+9uzZ0/yTaCmbHTJH+Lf3N38idCAAVVX7KKmq6YjKREREIoalAcjpdDJq1Chyc3OD+3w+H7m5uYwZM6bJY1999VXcbjc/+MEPTvh99u7dS2FhIZmZmY2+7nK5SEhIaPDoUIFhsBasBItx2kmI9nfYaSWYiIhI21i+CmzmzJn88Y9/5Pnnn2fTpk3ceuutlJeXM336dACuu+46Zs+efcxxzz77LJdffjk9evRosL+srIy7776bNWvWsGvXLnJzc7nssssYOHAgEydODMk5nVArJ0JnJGopvIiISHuwfA7QlClTOHjwIHPnziUvL4+RI0eyfPny4MTo3bt3Y7M1zGmbN2/mgw8+4N133z3m/ex2O1988QXPP/88RUVF9OzZkwsvvJBHHnkEl8sVknM6oUAAyvvSf3NUh7NZh6UnRLMlv0xL4UVERNrI8gAEcPvtt3P77bc3+trKlSuP2Td48GDM49xKIiYmhn/961/tWV77S+4P0Yn+u8If3FQ3J+gE0nUxRBERkXZh+RBYRDKMul6gFswDStftMERERNqFApBVgleEbvlKsLxiBSAREZG2UACySismQgeHwEo1B0hERKQtFICsElgKn/81VFc265BgAFIPkIiISJu0KgA9//zzLF26NPj1PffcQ1JSEmPHjuXbb79tt+K6tIReEJcGphfyNjbrkMAd4Q+WufH6Gp8ELiIiIifWqgD02GOPERMTA8Dq1atZtGgRjz/+OCkpKdx1113tWmCXVX8idDPnAaV0c2IzwOszKSzXMJiIiEhrtSoA7dmzh4EDBwLw1ltvceWVV3LLLbcwf/58/vvf/7ZrgV1aC68I7bDbSOlWuxKsWAFIRESktVoVgLp160ZhYSEA7777LhdccAEA0dHRVFY2bz6L0LaJ0FoKLyIi0mqtuhDiBRdcwE033cRpp53Gli1buPjiiwH46quvyM7Obs/6urbAUvhDW8BdCq74Ex6SnuDiy33+u8KLiIhI67SqB2jRokWMGTOGgwcP8vrrrwfvx7V+/XqmTp3argV2ad1SITELMGH/Z806JNADVKAAJCIi0mqt6gFKSkri6aefPmb/Qw891OaCIk7PkVC8xz8M1u/sEzYPXgxRAUhERKTVWtUDtHz5cj744IPg14sWLWLkyJFce+21HDlypN2KiwgtvCJ0RnAOkCZBi4iItFarAtDdd99NSUkJAF9++SU/+9nPuPjii9m5cyczZ85s1wK7vBZOhE7T/cBERETarFVDYDt37mTo0KEAvP7663z3u9/lscceY8OGDcEJ0dJMgQB0ZBdUHIbY5CabZyRqFZiIiEhbtaoHyOl0UlFRAcB7773HhRdeCEBycnKwZ0iaKSYJkgf4t5sxDJYe7w9ARyqqcdd4O7AwERGRrqtVAWj8+PHMnDmTRx55hHXr1nHJJZcAsGXLFnr37t2uBUaEFgyDJcVG4XT4f2wFmgckIiLSKq0KQE8//TQOh4PXXnuNZ555hl69egHwz3/+k0mTJrVrgREheEXoEwcgwzBI1zwgERGRNmnVHKA+ffrwj3/845j9v/nNb9pcUERq4UTo9Pho9hyu1FJ4ERGRVmpVAALwer289dZbbNq0CYBTTjmFSy+9FLvd3m7FRYzMEWDYoHQ/lOZBfEaTzdMTtRReRESkLVoVgLZt28bFF1/Mvn37GDx4MADz588nKyuLpUuXMmDAgHYtsstzxkHqyVDwtf/GqCc3vZIuMBFaV4MWERFpnVbNAfrpT3/KgAED2LNnDxs2bGDDhg3s3r2bfv368dOf/rS9a4wMLRgGy0j0zwHSEJiIiEjrtKoHaNWqVaxZs4bk5Lpr1vTo0YNf/vKXjBs3rt2Kiyg9T4PP/tq8pfC6I7yIiEibtKoHyOVyUVpaesz+srIynE5nm4uKSIFbYuzbAKbZZNO0eM0BEhERaYtWBaDvfve73HLLLaxduxbTNDFNkzVr1vDjH/+YSy+9tL1rjAwZp4ItCioPQ9HuppvWuxq0eYKwJCIiIsdqVQB66qmnGDBgAGPGjCE6Opro6GjGjh3LwIEDWbhwYTuXGCEcLkg/xb99gmGwwHWAKjxeSt01HV2ZiIhIl9OqOUBJSUm8/fbbbNu2LbgMfsiQIQwcOLBdi4s4PU+DA5/5J0KfcsVxm8U6HcRHOyitqqGgpIqE6KjQ1SgiItIFNDsAnegu7//5z3+C2wsWLGh9RZGs1+mw/s/+eUAnkJ4QTWlVGfklbgamxYegOBERka6j2QHo00+bd5ViwzBaXUzEC0yEPvA5+HxgO/4IZUZCNNsKysgr1kowERGRlmp2AKrfwyMdJPVkcMSAuwQOb4eUk47bNC1wP7BSBSAREZGWatUkaOkgdgdkDvdvn2AYLHgtIPUAiYiItJgCULhp5hWhMxJ0LSAREZHWUgAKN4F5QM1cCq/bYYiIiLScAlC46RWYCP0FeI9/jZ/AEJhuiCoiItJyCkDhJnkAuBKgphIOfnPcZsEAVOrG59PVoEVERFpCASjc2GyQOcK/3cQwWGq8C8OAGp9JYbknRMWJiIh0DQpA4SgwEbqJlWBRdhs94mqXwmsYTEREpEUUgMJRYB7QCVaCBSZCKwCJiIi0jAJQOAqsBMv/CmqOv8xdS+FFRERaRwEoHCX1gZhk8FVD/sbjNkurDUBaCi8iItIyCkDhyDDqhsGamAeUoaXwIiIiraIAFK6acUVoXQxRRESkdRSAwlXPE0+ETk/UHCAREZHWUAAKV4EeoIPfgKe80Sbp8RoCExERaQ0FoHCVkAnxmWD6/LfFaERgCKyw3IO7xhvK6kRERDo1BaBwdoIboybHOYmO8v8Idx2qCFVVIiIinZ4CUDjr1fQVoQ3DYHS/HgD8d+vBUFUlIiLS6SkAhbNmrAQ7d1AqAKu2KACJiIg0lwJQOAsMgR3eDpVFjTYJBKC1Ow5T4akJUWEiIiKdmwJQOItNhqS+/u3j9AINSI2jV1IMHq+PtTsOh7A4ERGRzksBKNyd4MaohmFw7mANg4mIiLREWASgRYsWkZ2dTXR0NDk5Oaxbt+64bc877zwMwzjmcckllwTbmKbJ3LlzyczMJCYmhgkTJrB169ZQnEr7O8FKMNA8IBERkZayPAC9/PLLzJw5k3nz5rFhwwZGjBjBxIkTKSgoaLT9G2+8wYEDB4KPjRs3Yrfbueqqq4JtHn/8cZ566ikWL17M2rVriYuLY+LEiVRVdcILBgYnQn923CZjB/TAYTPYeaicbwsbv2iiiIiI1LE8AC1YsICbb76Z6dOnM3ToUBYvXkxsbCxLlixptH1ycjIZGRnBx4oVK4iNjQ0GINM0WbhwIQ888ACXXXYZw4cP54UXXmD//v289dZbITyzdpI5AjCgeA+UNd7DEx8dxai+3QF4X71AIiIiJ2RpAPJ4PKxfv54JEyYE99lsNiZMmMDq1aub9R7PPvss11xzDXFxcQDs3LmTvLy8Bu+ZmJhITk7Ocd/T7XZTUlLS4BE2ohMg5ST/dlPDYJoHJCIi0myWBqBDhw7h9XpJT09vsD89PZ28vLwTHr9u3To2btzITTfdFNwXOK4l7zl//nwSExODj6ysrJaeSsdqxo1RA/OAPtpeqNtiiIiInIDlQ2Bt8eyzzzJs2DBGjx7dpveZPXs2xcXFwceePXvaqcJ2ElgJdpwrQgMMzUwgNd5FhcfL+l1HQlSYiIhI52RpAEpJScFut5Ofn99gf35+PhkZGU0eW15ezt/+9jduvPHGBvsDx7XkPV0uFwkJCQ0eYaX+FaFNs9EmhmFwzkkaBhMREWkOSwOQ0+lk1KhR5ObmBvf5fD5yc3MZM2ZMk8e++uqruN1ufvCDHzTY369fPzIyMhq8Z0lJCWvXrj3he4atjGFg2KG8AEr2HbeZ5gGJiIg0j+VDYDNnzuSPf/wjzz//PJs2beLWW2+lvLyc6dOnA3Ddddcxe/bsY4579tlnufzyy+nRo0eD/YZhcOedd/KLX/yCd955hy+//JLrrruOnj17cvnll4filNpfVAykDfVvNzEMdvbAFAwDvskrJa+4Ey75FxERCRGH1QVMmTKFgwcPMnfuXPLy8hg5ciTLly8PTmLevXs3NlvDnLZ582Y++OAD3n333Ubf85577qG8vJxbbrmFoqIixo8fz/Lly4mOju7w8+kwvU6D/C/9w2BDL220Sfc4JyN6J/HZniLe33KQq88Ms8ncIiIiYcIwzeNMKolgJSUlJCYmUlxcHD7zgT75M/zjTuh/Hlz39nGb/WbFFn6bu5VLhmWyaNrpIStPRETEai35/LZ8CEyaqRkToaFuHtB/tx6kxusLRWUiIiKdjgJQZ5E2FOwuqCqGwzuO22xE7yQSY6Ioqarh871FoatPRESkE1EA6iwcTsg41b/dxAUR7TaDs09KAWDVZq0GExERaYwCUGfSjCtCg+4OLyIiciIKQJ1JM64IDXUB6It9xRSWuTu6KhERkU5HAagzCUyEPvA5+I5/v6+0hGiGZCZgmvDBtkMhKk5ERKTzUADqTFIGQVQcVJfDoS1NNg0Og2kekIiIyDEUgDoTmx16jvRvN3MY7P2tB/H5dKknERGR+hSAOpv61wNqwqi+3Ylz2jlU5uHrAyUhKExERKTzUADqbIIBqOkeIKfDxtiBtcvhtRpMRESkAQWgziYQgPK+hBpPk001D0hERKRxCkCdTXJ/iE4CrwcKvm6yaSAArd99hJKq6hAUJyIi0jkoAHU2htHsYbCs5Fj6p8bh9Zl8pOXwIiIiQQpAnVEgAJ1gJRjoqtAiIiKNUQDqjAJXhN7/2Qmb1p8HZDZxF3kREZFIogDUGQV6gAq+Bk9Fk03P6t8Dl8PG/uIqthWUhaA4ERGR8KcA1Bkl9IK4NDC9kL+xyabRUXZy+vcANAwmIiISoADUGRlGs2+MCpoHJCIicjQFoM6qmSvBoC4Ard1xmApPTUdWJSIi0ikoAHVWPQMToZu+JQbAgNQ4eiXF4PH6WLvjcAcXJiIiEv4UgDqrQA/Qoa1Q1fS9vgzD4NzBGgYTEREJUADqrLqlQmIWYMKBz0/YXPOARERE6igAdWYtmAc0dkAPHDaDnYfK+bawvIMLExERCW8KQJ1ZC64IHR8dxai+3QF4X71AIiIS4RSAOrNezZ8IDWgekIiISC0FoM4sc6T/uehbKNp9wuaBeUAfbS/EXePtwMJERETCmwJQZxaTBP3O8W9/+NQJmw/NTCA13kWFx8v6XUc6tjYREZEwpgDU2Z1zt/95wwtQmtdkU8MwOOckDYOJiIgoAHV22WdD1lngdcNH/3vC5poHJCIiogDU+RlGXS/QJ0ug/FCTzc8emIJhwDd5peQVV4WgQBERkfCjANQVDDzfvyS+ugJWL2qyafc4JyN6JwFaDi8iIpFLAagrMAw45x7/9ro/QkXT9/vSVaFFRCTSKQB1FYMvgvRTwVMKa3/fZNPAPKD/bj1IjdcXiupERETCigJQV2EYcM4s//baZ5q8QeqI3kkkxkRRUlXD53uLQlOfiIhIGFEA6kqGXAopg6CqGD7+43Gb2W0GZ5+UAsCqzRoGExGRyKMA1JXY7HB2bS/Q6kXgOf5NTzUPSEREIpkCUFdz6pXQvR9UFMInfz5us0AA+mJfMYVl7lBVJyIiEhYUgLoauwPOnunf/ugpqK5stFlaQjRDMhMwTfhgW9PXDhIREelqFIC6ouHXQGIWlOXDhr8ct1lwGEzzgEREJMIoAHVFDieMv9O//eFCqPE02iwQgN7fehCfzwxNbSIiImFAAairGvkD6JYBJfvg8xcbbTKqb3finHYOlXn4+sDxl82LiIh0NQpAXVVUNIy7w7/93wXgrT6midNhY+zA2uXwWg0mIiIRRAGoKxt1A8SmQNG38OVrjTbRPCAREYlECkBdmTMWxt7u3/7vr8HnPaZJIACt332Ekqpje4lERES6IgWgru7MmyCmOxRug6/fOublrORY+qfG4fWZfKTl8CIiEiEUgLo6VzycdZt/+/1fg+/Ym5/qqtAiIhJpFIAiwehbwJUABV/D5qXHvFx/HpBpajm8iIh0fQpAkSAmyR+CAN5/Ao4KOWf174HLYWN/cRXbCspCX5+IiEiIKQBFirNug6g4OPA5bF3R4KXoKDs5/XsAGgYTEZHIYHkAWrRoEdnZ2URHR5OTk8O6deuabF9UVMSMGTPIzMzE5XIxaNAgli1bFnz9wQcfxDCMBo+TTz65o08j/MX1gDN/5N9+//FjeoE0D0hERCKJpQHo5ZdfZubMmcybN48NGzYwYsQIJk6cSEFBQaPtPR4PF1xwAbt27eK1115j8+bN/PGPf6RXr14N2p1yyikcOHAg+Pjggw9CcTrhb8xPwBENez+GnasavBQIQGt3HKbCU2NFdSIiIiFjaQBasGABN998M9OnT2fo0KEsXryY2NhYlixZ0mj7JUuWcPjwYd566y3GjRtHdnY25557LiNGjGjQzuFwkJGREXykpKSE4nTCX3y6/+KIAKueaPDSgNQ4eiXF4PH6WLvjcOhrExERCSHLApDH42H9+vVMmDChrhibjQkTJrB69epGj3nnnXcYM2YMM2bMID09nVNPPZXHHnsMr7fhBf62bt1Kz5496d+/P9OmTWP37t1N1uJ2uykpKWnw6LLG/hTsTvj2A/j2o+BuwzA4d7CGwUREJDJYFoAOHTqE1+slPT29wf709HTy8vIaPWbHjh289tpreL1eli1bxpw5c3jyySf5xS9+EWyTk5PDc889x/Lly3nmmWfYuXMnZ599NqWlpcetZf78+SQmJgYfWVlZ7XOS4SixF4yc5t9+v2EvkOYBiYhIpLB8EnRL+Hw+0tLS+MMf/sCoUaOYMmUK999/P4sXLw62ueiii7jqqqsYPnw4EydOZNmyZRQVFfHKK68c931nz55NcXFx8LFnz55QnI51xt8Jhh22/xv2rg/uHjugBw6bwc5D5XxbWG5dfSIiIh3MsgCUkpKC3W4nPz+/wf78/HwyMjIaPSYzM5NBgwZht9uD+4YMGUJeXh4ej6fRY5KSkhg0aBDbtm07bi0ul4uEhIQGjy6tezaMuMa//f7jwd3x0VGM6tsdUC+QiIh0bZYFIKfTyahRo8jNzQ3u8/l85ObmMmbMmEaPGTduHNu2bcNX73YOW7ZsITMzE6fT2egxZWVlbN++nczMzPY9gc5u/EwwbLBluf/aQLX+5+Q0AH6/agflbq0GExGRrsnSIbCZM2fyxz/+keeff55NmzZx6623Ul5ezvTp0wG47rrrmD17drD9rbfeyuHDh7njjjvYsmULS5cu5bHHHmPGjBnBNrNmzWLVqlXs2rWLjz76iCuuuAK73c7UqVNDfn5hLWUgnPI9//b7vw7u/sFZfemVFMO+okoWrNhiUXEiIiIdy2HlN58yZQoHDx5k7ty55OXlMXLkSJYvXx6cGL17925strqMlpWVxb/+9S/uuusuhg8fTq9evbjjjjv4+c9/Hmyzd+9epk6dSmFhIampqYwfP541a9aQmpoa8vMLe+fMgo2vwaZ3oGATpA0hzuXg0StO5YY/f8yfP9zJ5BE9GZmVZHWlIiIi7cowdffLY5SUlJCYmEhxcXHXnw/08g/9AWjYVXDln4K773r5M978dB8nZ8Tzzu3jcTo61Xx5ERGJQC35/NanWqQ7Z5b/eePrULg9uHvOd4eSHOfkm7xS/vD+9uMcLCIi0jkpAEW6zBEwaBKYPvjvk8HdyXFO5n53KABP5W7TXeJFRKRLUQASOOdu//Pnf4Mj3wZ3XzayJ+cOSsXj9XHfG1/i82m0VEREugYFIIHeZ0D/74DphQ9+E9xtGAaPXnEqsU4763Yd5sV1Td9SREREpLNQABK/c+/xP3/2VyjeF9zdu3ssd08cDMAv//kNecVVVlQnIiLSrhSAxK/vWOg7Hrwe+OipBi9dNyabkVlJlLlreOCtjWjhoIiIdHYKQFInsCJs/XNQWneLErvN4FdXDsdhM3hvUz7/3Nj4zWpFREQ6CwUgqdP/POh9JtRUwT/uhBp38KXBGfHcdt4AAOa+/RXFFdXW1CgiItIOFICkjmHABQ+D3Qmbl8FfrwJ3afDlGf8zkAGpcRwqc/Posq8tLFRERKRtFICkob5jYdpr4OwGO1fB85dCeSEALoedX105HIBXPtnLh9sOWVmpiIhIqykAybH6nwvXvwMxybB/A/x5EhTvBeCM7GR+eFZfAO5780sqPV4rKxUREWkVBSBpXK9R8KN/QUIvOLQFnp0Ih7YCcM+kwWQkRPNtYQULc3XHeBER6XwUgOT4Ugf5Q1CPk6BkLyyZCPs/JT46ikcuPxWAP/13Jxv3FVtcqIiISMsoAEnTkrLgR8shcyRUFMJzk2Hn+1wwNJ1Lhmfi9Zn8/PUvqPH6rK5URESk2RSA5MTiUuCGf0D22eAphf+7Ejb9gwcnn0JiTBRf7S/hTx/stLpKERGRZlMAkuZxxftXh538Xf/Vol/5IanbXuWBS4YA8JsVW9h1qNziIkVERJpHAUiaLyoarnoeTvsBmD54ewbfd7/JuIE9cNf4mP3Gl7pNhoiIdAoKQNIydgdc+jSM/SkAxoo5/C7tHaKjDFbvKOSVT/ZYXKCIiMiJKQBJyxkGXPgITHgIgMQNi3iz9yvY8PHo0k0UlOiO8SIiEt4UgKT1xt8Jk58Cw8aQA2/yl4RnqKqq5MG/f2V1ZSIiIk1SAJK2GXW9f16Q3ck4z4f82fkEq77cyb++0h3jRUQkfCkASdsNvRSmvQrOboyzbeSvzkd58q2PKKnSHeNFRCQ8KQBJ++h/Hlz/DmZMMiNtO1jkfoDfvf2+1VWJiIg0SgFI2k+vURg/Wo47NpOTbPv4wde38PmnH1tdlYiIyDEUgKR9pQ7GdcsKClx96G0cou/b38O9e73VVYmIiDSgACTtLymL6Fve5WtjAEmUwPOTYccqq6sSEREJUgCSDpHQI5N9l77MR96huLzl8MKl8H/fh+3/Bl0tWkRELKYAJB3mgtNO4q8Dn+TVmnPwYcC2FfCXK+CZsbDhBajWBRNFRMQahqmbNx2jpKSExMREiouLSUhIsLqcTi2/pIoLFqwiyb2Pn8blcoXxH+zVtTdNjU2BM2+CM2+EbmnWFioiIp1eSz6/FYAaoQDUvjbuK+b//WU9+4oqSYuq4s8jvuGUPS9Bce19w+xOGH41nHUbpJ9ibbEiItJpKQC1kQJQ+ztc7uGnL33KB9sOAXDzuCzu7bsV+7pnYG+9pfL9z4Mxt8OA88GmEVoREWk+BaA2UgDqGF6fyRP/2sziVdsBOKt/Mk9fezopRz6H1Ytg0ztg+vyNUwbBWbfC8GvAGWth1SIi0lkoALWRAlDH+ueXB5j16ueUe7xkJkbzu2mnc1qf7lC0G9b+3j9B2l3ibxyTDGf8CEbfDPEZ1hYuIiJhTQGojRSAOt62glJu+ct6dhwsx2m38dBlpzB1dB//i+5S+PT/YM0zUPStf58tCk69EsbcBpkjrCtcRETClgJQGykAhUZpVTU/e+Vz3v06H4BrzsziwUtPITrK7m/g88LmZf7hsd2r6w7MPts/YXrg+eBwWVC5iIiEIwWgNlIACh2fz+SZVdv59bubMU0Y0TuRZ34wip5JMQ0b7tsAa34HX70Jvhr/PrsTMoZD7zOh9xn+56Q+YBihPxEREbGcAlAbKQCF3qotB7njb59SVFFNjzgn/3vtaYwdkHJsw+J9sO4P8NmLUF5w7OtxaQ0DUc/TwNWt409AREQspwDURgpA1thzuIL/95f1fH2gBJsBsy8awk1n98NorEfHNOHILtj7iX8Z/d6PIe+Lut6hAMMGaafUBaLeZ0KPgVpiLyLSBSkAtZECkHUqPV7uf/NL3vh0HwCXDM/k8SuHE+dynPjg6ko48EVdINr7CZTsPbZddCL0OqMuFPUaBbHJ7XwmIiISagpAbaQAZC3TNPnLmm95+O9fU+MzGZTejd//8Az6pcS1/M1KDsC+T+oC0b4NUFN5bLseAyH9VEjoBQmZEJ8JCT39z/GZEBXd9hMTEZEOpQDURgpA4eGTXYe59a8bOFjqJt7l4DdTRjJhaHrb3tRbDQVf1wWivR9D4bYTHxeT3Hg4qv8c010TsEVELKQA1EYKQOGjoKSK2/66gU++PQLAT88/iTvPPwmbrR2DRsVhfxgq3Aal+/29RqUHoGS//+F1N+99HNH+izXG96wLStFJ/knYrnhw1j4HHs56++3NGOITEZEmKQC1kQJQePHU+Hh06dc8v9p/UcTvDE5l4ZTTSIyN6vhvbppQecQfhAKhqMHzAX9oqihs2/dxxNQGo0Aoiq/7un5wior1X/vIHgV2Fzic/me7s27b4fR/fbzX7VHqqRKRLkkBqI0UgMLT6+v3ct+bX+Ku8dEnOZb7Lj6Z/zk5HacjDFZ0VVf5A1H9cFSa57+lh7vMf3VrT1ntdkntdil4PdbUWz8gGXaw2cHm8K+aC27X7jds/q9t9npt7UcdZ/evrKt/DIb/2bD5A5dhHGe/7aj9RiP7jYbPcOy+4HPgJBt77ejjOf7XLW7TiEZfa2Tfcd+jmUG12YG2A4Jvu4dphfOIkX4K9Dq9Xd9SAaiNFIDC18Z9xfz4/9az94h/InNSbBSXjujJlaf3ZnjvxMaXzIezGrc/FHlK/YHIXVYbjo4OTrWvV1f6h+RqPP7w5PX438Pr9s9vqnEftb/22fRafaYiIg2NnwkT5rXrWyoAtZECUHg7Uu5h8fvbeXPDPgpK6+bnDEzrxpWn9+aK03qRkahVWw34vA1DUf1tn9cfkHw14PPV2w7s9x7Vxgumr5E2NbX7vYDp3zYDz76j9jW2/+j2NNxP7T9Vge0GzzSyzwzuPnZfvWMa+7q1bepr9J/W5rZrgWYf3wH/1Lf7x4c+jiLK0Mth5NR2fUsFoDZSAOocarw+PtxeyOvr9/Kvr/Jw1/g/NA0Dxg9M4crTezPxlAxinHaLKxURkVBQAGojBaDOp6SqmmVfHOCNDftYt+twcH83l4OLh2XwvdN7Mzo7uX1Xj4mISFhRAGojBaDO7dvCct7YsI83Pt3LnsN1Fz3s3T2G753emytP70XfHq24qKKIiIS1lnx+W758ZtGiRWRnZxMdHU1OTg7r1q1rsn1RUREzZswgMzMTl8vFoEGDWLZsWZveU7qWvj3iuOuCQaya9R1e+X9jmHJGFt1cDvYeqeSp3K2c+8RKrlr8ES+t201JVbXV5YqIiAUs7QF6+eWXue6661i8eDE5OTksXLiQV199lc2bN5OWlnZMe4/Hw7hx40hLS+O+++6jV69efPvttyQlJTFixIhWvWdj1APU9VR6vLz7dR6vb9jHB1sP4qv9W+9y2LjwlAyuPL0X4wem4LBb/n8CERFppU4zBJaTk8OZZ57J008/DYDP5yMrK4uf/OQn3Hvvvce0X7x4MU888QTffPMNUVGNXwSvpe/ZGAWgri2vuIq3PtvH6+v3srWgLLg/JspO3x6xtY84/3Oy/7lnUgx2zR8SEQlrnSIAeTweYmNjee2117j88suD+6+//nqKiop4++23jznm4osvJjk5mdjYWN5++21SU1O59tpr+fnPf47dbm/VezZGASgymKbJxn0lvL5hL29/to8jFccfDouyG2R1j6VPj1j6JtcLSD3iyEqOweXQSjMREau15PPbshsQHTp0CK/XS3p6w5tbpqen88033zR6zI4dO/j3v//NtGnTWLZsGdu2beO2226jurqaefPmteo9AdxuN2533fVkSkpK2nBm0lkYhsGw3okM653IA5cMYffhCr49XMG3h8r9z4UVfFtYzp7DlXi8PnYcKmfHofJG3gd6JsbQJzmW7JRY+tT2GvVJjiU5zkl8tIM4p0Mr0EREwkinugOjz+cjLS2NP/zhD9jtdkaNGsW+fft44oknmDev9VeTnD9/Pg899FA7ViqdjcNuo39qN/qndoPBDV/z+kzySqr4trC8NhRV1Nsup9zjZV9RJfuKKlm9o/F7gtkM/5L8hJgoEqKjiI/2b8dHO0iIjqrdH9h2EB99bLsozU8SEWk3lgWglJQU7HY7+fn5Dfbn5+eTkZHR6DGZmZlERUVht9cNNwwZMoS8vDw8Hk+r3hNg9uzZzJw5M/h1SUkJWVlZrTkt6YLsNoNeSTH0Soph7ICGr5mmSWG559hwdLiCPYcrKa70UO018ZlQUlVDSVUNUNno9zmRmCg7sU47DruBw2Yjym7gsNtw2Ayi7Lba/f7XHPbafQ1eCxxTtx1lt5Ec5yQ13kVafDSp8S5S410kRDs6321FRERawLIA5HQ6GTVqFLm5ucH5Oj6fj9zcXG6//fZGjxk3bhwvvvgiPp8Pm83/v+EtW7aQmZmJ0+kEaPF7ArhcLlwuV/udnEQMwzBI6eYipZuLUX2Tj3ndNE3cNT5KqqopqayhpKqa0qoaSiprn6uqj7sdaFfu8d/Hq7LaS2V1aO7p5XLYgmEotZvrmIDk/9p/3mFxM1oJOdM0Mc3AvWs7JiybponH66Oq2kdVtZdKj5eqmtrnwL5qb73neu3q/b7EOu3ERNmJrn2OibIT42z4HF27HVvva5fD1qn/I+DzmVRWe6nweKnw1NQ++/986n9d4amh0uOl3OOlMrC/9s+x3F1DZbUXm2EEe6zjox21j6ij9h39tSOsV9ZaOgQ2c+ZMrr/+es444wxGjx7NwoULKS8vZ/r06QBcd9119OrVi/nz5wNw66238vTTT3PHHXfwk5/8hK1bt/LYY4/x05/+tNnvKRJKhmEQXfuPaVp8696jxuujzF1DSaX/H6Jqr48an0mN10e116TG56PGawb3V3v9X9f4al8PtG/kGHeNj8JyDwUlVRwsc3Ow1E1pVQ3uGh97j1QGbzrblKTYKFK7uUhL8IelpFhn7Tnb6j5Yaj98oh02Ypz19kXZ6m37H5G62s5d4yW/2M3+4kryiqtw13iJdTqIddrrPduJdTmIjbIT67LjtLfPB3SN10dRZTVFFR6OVFRzuNwT3D5S4aGovPa59usjFdXB3k3whyCbYWALPtdtGwbYbAZ2w8Bo0Ma/v35bDHAfFWx8Fl6q1zDqel7r/z0N9J4Gelf924Ge1bqe1mAbW702dttR7Q1MwFPj8z+8vuB2tdeH+6h9gTbB1446LrC/wlNDVbXPuj+8WjFR9gaBqW7o38G5g9KYdOrxR2c6mqUBaMqUKRw8eJC5c+eSl5fHyJEjWb58eXAS8+7du4M9PQBZWVn861//4q677mL48OH06tWLO+64g5///OfNfk+RzsZht5EU6yQp1hmS71dV7eVgqZuCUn8gOljm5mC9gFT/tRqfSVFFNUUV1Q0uKdAWTrvNH55qP3SiHXacjroPD6e9bvguymEjyla37TzqQ8ZZ70Oo/odWjNPe4H+qgX+UO2o1n6fGR35JFfuLKskrqWJ/URV5xZXsL64ir7iKA8WVHCrztPh97TYjGIzinA5i6j03CE4uOy6HnZJ6Iad+wCmtqmnT+ZkmeE0Tf39LxyQWW20YCf69iKrrzXHVhu0Yp//vS10b/+dHZbWXqkDvR70eo0BvSP2vq6q9wWBnmgR7STq7wN+TGKed2CgHsa5Ab1e9cF3796Xu70/dPq/PpLSqhtLa3ulSt3+7fs928LWqmmDvW6Dnuv6NqwO6xzotDUC6FUYjtAxe5MR8PpPiymoOlrkpKHFzsKyKghI3JVXVVFX7gh80dUMSDffV/zpwI1urOR02/0T0aAfxtRPTGwSlBoGpbl+M086hMjf7iyo5UBtq6oedQ2XH/uPfGJfDRmZiNBmJ0cQ6HQ2HKdw1VFR7qXB78Xg75s8rMSaK7rFRJMU66R4bRffa4N09NoqkuLp93WOddI+LItphx2f657iZpom3dtvnM4OhyGeamLX7vb7A1+AzzdqvCb7uM80GvYf+cON/jrIbIRuOqvbWDaUFPsArPXVfVwd7XH112/V7YWt8VAd7XOt6XatrTKrr9dgGjrUZ+AO7w/9w1Yb5wNfB1+y1rwUe9rr/HNQd599XP8hEO+whX4Va7fVRVn84v144CgSlM/p2Z+zAlHb9vp3iOkDhTAFIJLR8Pv9wXGW9oY9Kjxd3jZdKj49qn8//oVL7geHx1v8QqR0SqPF/yAS2679Wv627xv/hVn8uVqm7bT0gzeGsDTf+R0yD7YzEaHomxdA9NqpZH/LVXl+w96K8dv5GRb3twLyNcrd/Tkd57Z9lfHRdwEmuDTFJtYEmMSYqYocfpevoFNcBEhEJsNkM/2RUpzUXlPT6TMrcdf8zPXoyemlVNSX1no9+vdLjpUc3JxkJ/iCTkRhNz8RoMuoFneQ4Z7v1YETZbSTG2EiMafyK+CJyYgpAIhLx7DaDxJgoBQqRCBK+69NEREREOogCkIiIiEQcBSARERGJOApAIiIiEnEUgERERCTiKACJiIhIxFEAEhERkYijACQiIiIRRwFIREREIo4CkIiIiEQcBSARERGJOApAIiIiEnEUgERERCTiKACJiIhIxHFYXUA4Mk0TgJKSEosrERERkeYKfG4HPsebogDUiNLSUgCysrIsrkRERERaqrS0lMTExCbbGGZzYlKE8fl87N+/H9M06dOnD3v27CEhIcHqsqSekpISsrKy9LMJM/q5hC/9bMKTfi7tyzRNSktL6dmzJzZb07N81APUCJvNRu/evYNdaQkJCfqLGab0swlP+rmEL/1swpN+Lu3nRD0/AZoELSIiIhFHAUhEREQijgJQE1wuF/PmzcPlclldihxFP5vwpJ9L+NLPJjzp52IdTYIWERGRiKMeIBEREYk4CkAiIiIScRSAREREJOIoAImIiEjEUQBqwqJFi8jOziY6OpqcnBzWrVtndUkR7cEHH8QwjAaPk08+2eqyItL777/P5MmT6dmzJ4Zh8NZbbzV43TRN5s6dS2ZmJjExMUyYMIGtW7daU2wEOdHP5YYbbjjmd2jSpEnWFBtB5s+fz5lnnkl8fDxpaWlcfvnlbN68uUGbqqoqZsyYQY8ePejWrRtXXnkl+fn5FlUcGRSAjuPll19m5syZzJs3jw0bNjBixAgmTpxIQUGB1aVFtFNOOYUDBw4EHx988IHVJUWk8vJyRowYwaJFixp9/fHHH+epp55i8eLFrF27lri4OCZOnEhVVVWIK40sJ/q5AEyaNKnB79BLL70Uwgoj06pVq5gxYwZr1qxhxYoVVFdXc+GFF1JeXh5sc9ddd/H3v/+dV199lVWrVrF//36+973vWVh1BDClUaNHjzZnzJgR/Nrr9Zo9e/Y058+fb2FVkW3evHnmiBEjrC5DjgKYb775ZvBrn89nZmRkmE888URwX1FRkelyucyXXnrJggoj09E/F9M0zeuvv9687LLLLKlH6hQUFJiAuWrVKtM0/b8fUVFR5quvvhpss2nTJhMwV69ebVWZXZ56gBrh8XhYv349EyZMCO6z2WxMmDCB1atXW1iZbN26lZ49e9K/f3+mTZvG7t27rS5JjrJz507y8vIa/P4kJiaSk5Oj358wsHLlStLS0hg8eDC33norhYWFVpcUcYqLiwFITk4GYP369VRXVzf4nTn55JPp06ePfmc6kAJQIw4dOoTX6yU9Pb3B/vT0dPLy8iyqSnJycnjuuedYvnw5zzzzDDt37uTss8+mtLTU6tKknsDviH5/ws+kSZN44YUXyM3N5Ve/+hWrVq3ioosuwuv1Wl1axPD5fNx5552MGzeOU089FfD/zjidTpKSkhq01e9Mx9Ld4KXTuOiii4Lbw4cPJycnh759+/LKK69w4403WliZSOdwzTXXBLeHDRvG8OHDGTBgACtXruT888+3sLLIMWPGDDZu3Kj5i2FAPUCNSElJwW63HzMDPz8/n4yMDIuqkqMlJSUxaNAgtm3bZnUpUk/gd0S/P+Gvf//+pKSk6HcoRG6//Xb+8Y9/8J///IfevXsH92dkZODxeCgqKmrQXr8zHUsBqBFOp5NRo0aRm5sb3Ofz+cjNzWXMmDEWVib1lZWVsX37djIzM60uRerp168fGRkZDX5/SkpKWLt2rX5/wszevXspLCzU71AHM02T22+/nTfffJN///vf9OvXr8Hro0aNIioqqsHvzObNm9m9e7d+ZzqQhsCOY+bMmVx//fWcccYZjB49moULF1JeXs706dOtLi1izZo1i8mTJ9O3b1/279/PvHnzsNvtTJ061erSIk5ZWVmDXoOdO3fy2WefkZycTJ8+fbjzzjv5xS9+wUknnUS/fv2YM2cOPXv25PLLL7eu6AjQ1M8lOTmZhx56iCuvvJKMjAy2b9/OPffcw8CBA5k4caKFVXd9M2bM4MUXX+Ttt98mPj4+OK8nMTGRmJgYEhMTufHGG5k5cybJyckkJCTwk5/8hDFjxnDWWWdZXH0XZvUytHD2v//7v2afPn1Mp9Npjh492lyzZo3VJUW0KVOmmJmZmabT6TR79eplTpkyxdy2bZvVZUWk//znPyZwzOP66683TdO/FH7OnDlmenq66XK5zPPPP9/cvHmztUVHgKZ+LhUVFeaFF15opqammlFRUWbfvn3Nm2++2czLy7O67C6vsZ8JYP75z38OtqmsrDRvu+02s3v37mZsbKx5xRVXmAcOHLCu6AhgmKZphj52iYiIiFhHc4BEREQk4igAiYiISMRRABIREZGIowAkIiIiEUcBSERERCKOApCIiIhEHAUgERERiTgKQCIizbBy5UoMwzjmfk0i0jkpAImIiEjEUQASERGRiKMAJCKdgs/nY/78+fTr14+YmBhGjBjBa6+9BtQNTy1dupThw4cTHR3NWWedxcaNGxu8x+uvv84pp5yCy+UiOzubJ598ssHrbrebn//852RlZeFyuRg4cCDPPvtsgzbr16/njDPOIDY2lrFjx7J58+aOPXER6RAKQCLSKcyfP58XXniBxYsX89VXX3HXXXfxgx/8gFWrVgXb3H333Tz55JN8/PHHpKamMnnyZKqrqwF/cLn66qu55ppr+PLLL3nwwQeZM2cOzz33XPD46667jpdeeomnnnqKTZs28fvf/55u3bo1qOP+++/nySef5JNPPsHhcPCjH/0oJOcvIu1LN0MVkbDndrtJTk7mvffeY8yYMcH9N910ExUVFdxyyy185zvf4W9/+xtTpkwB4PDhw/Tu3ZvnnnuOq6++mmnTpnHw4EHefffd4PH33HMPS5cu5auvvmLLli0MHjyYFStWMGHChGNqWLlyJd/5znd47733OP/88wFYtmwZl1xyCZWVlURHR3fwn4KItCf1AIlI2Nu2bRsVFRVccMEFdOvWLfh44YUX2L59e7Bd/XCUnJzM4MGD2bRpEwCbNm1i3LhxDd533LhxbN26Fa/Xy2effYbdbufcc89tspbhw4cHtzMzMwEoKCho8zmKSGg5rC5AROREysrKAFi6dCm9evVq8JrL5WoQglorJiamWe2ioqKC24ZhAP75SSLSuagHSETC3tChQ3G5XOzevZuBAwc2eGRlZQXbrVmzJrh95MgRtmzZwpAhQwAYMmQIH374YYP3/fDDDxk0aBB2u51hw4bh8/kazCkSka5LPUAiEvbi4+OZNWsWd911Fz6fj/Hjx1NcXMyHH35IQkICffv2BeDhhx+mR48epKenc//995OSksLll18OwM9+9jPOPPNMHnnkEaZMmcLq1at5+umn+d3vfgdAdnY2119/PT/60Y946qmnGDFiBN9++y0FBQVcffXVVp26iHQQBSAR6RQeeeQRUlNTmT9/Pjt27CApKYnTTz+d++67LzgE9ctf/pI77riDrVu3MnLkSP7+97/jdDoBOP3003nllVeYO3cujzzyCJmZmTz88MPccMMNwe/xzDPPcN9993HbbbdRWFhInz59uO+++6w4XRHpYFoFJiKdXmCF1pEjR0hKSrK6HBHpBDQHSERERCKOApCIiIhEHA2BiYiISMRRD5CIiIhEHAUgERERiTgKQCIiIhJxFIBEREQk4igAiYiISMRRABIREZGIowAkIiIiEUcBSERERCKOApCIiIhEnP8PSqUjrv91K/cAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# draw Losses\n",
    "if is_graphic:\n",
    "    plt.plot(list(range(1, len(arr_train_loss) + 1)), arr_train_loss, label=\"train\")\n",
    "    plt.plot(list(range(1, len(arr_val_loss) + 1)), arr_val_loss, label=\"val\")\n",
    "\n",
    "    plt.title(\"Loss\")\n",
    "    plt.xlabel(\"epoch\")\n",
    "    plt.ylabel(\"loss\")\n",
    "    plt.legend(loc=\"upper left\")\n",
    "\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "8bda887e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# draw mae\n",
    "if is_graphic:\n",
    "    plt.plot(list(range(1, len(arr_train_mae) + 1)), arr_train_mae, label=\"train\")\n",
    "    plt.plot(list(range(1, len(arr_val_mae) + 1)), arr_val_mae, label=\"val\")\n",
    "\n",
    "    plt.title(\"MAE\")\n",
    "    plt.xlabel(\"epoch\")\n",
    "    plt.ylabel(\"mae\")\n",
    "    plt.legend(loc=\"upper right\")\n",
    "\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "1cc21154",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# draw Accuracies\n",
    "if is_graphic:\n",
    "    plt.plot(list(range(1, len(arr_train_acc) + 1)), arr_train_acc, label=\"train\")\n",
    "    plt.plot(list(range(1, len(arr_val_acc) + 1)), arr_val_acc, label=\"val\")\n",
    "\n",
    "    plt.title(\"Accuracy\")\n",
    "    plt.xlabel(\"epoch\")\n",
    "    plt.ylabel(\"accuracy\")\n",
    "    plt.legend(loc=\"upper left\")\n",
    "\n",
    "    plt.show()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "0133ff45",
   "metadata": {},
   "source": [
    "# Eval model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "e095b271",
   "metadata": {},
   "outputs": [],
   "source": [
    "from utils.split_all_docs import split_all_docs\n",
    "from rouge_score import rouge_scorer\n",
    "from utils.threshold_probs import threshold_probs_by_prop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "273c6eb2",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = SIMPLE_CNN_RNN(device=device, vocab_size=vocab_size, word_embed=glovemgr.getEmbeddings(), dim_word_embed=300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "4fc41994",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "20"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "best_epoch = np.argmin(np.array(arr_val_loss)) + 1\n",
    "model.load(checkpoints_folder + \"/\" + model_name + \"-\" + str(best_epoch) + \".pt\")\n",
    "model.to(device)\n",
    "model.eval()\n",
    "best_epoch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "3edbae7a",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_set = df_test\n",
    "test_set[\"doc_splited\"] = split_all_docs(test_set[args.doc_col_name], False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "bb8ab07a",
   "metadata": {},
   "outputs": [],
   "source": [
    "scorer = rouge_scorer.RougeScorer(['rouge1', 'rouge2', 'rougeL'], use_stemmer=True)\n",
    "criterion = nn.BCELoss()\n",
    "mae_fn = nn.L1Loss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "67659b8a",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Eval model: 100%|██████████| 100/100 [00:01<00:00, 83.41batch/s, accuracy=0.863, mae=0.441, rouge1=0.422, rouge2=0.227, rougeL=0.317]\n"
     ]
    }
   ],
   "source": [
    "arr_rouge1 = []\n",
    "arr_rouge2 = []\n",
    "arr_rougeL = []\n",
    "arr_loss = []\n",
    "mae = []\n",
    "accuracy = []\n",
    "\n",
    "idx = 0\n",
    "with tqdm(test_iter, unit=\"batch\", total=len(test_iter)) as tepoch:\n",
    "    for batch in tepoch:\n",
    "        tepoch.set_description(\"Eval model\")\n",
    "        features = []\n",
    "        doc_lens = []\n",
    "\n",
    "        for j in range(1):\n",
    "            doc_lens.append(len(batch[j][\"docs\"]))\n",
    "            features = features + batch[j][\"docs\"]\n",
    "                \n",
    "        features = torch.tensor(features, dtype=torch.long).to(device)\n",
    "\n",
    "        targets = [torch.tensor(batch[j][\"labels\"], dtype=torch.float) for j in range(1)]\n",
    "        targets = torch.cat(targets)\n",
    "        targets = targets.to(device)\n",
    "\n",
    "        y_pred = model(features, doc_lens)\n",
    "    \n",
    "        loss = criterion(y_pred, targets)\n",
    "\n",
    "        arr_loss.append(loss.item())\n",
    "\n",
    "        doc = test_set[\"doc_splited\"].iloc[idx]\n",
    "        summaries = test_set[args.sum_col_name].iloc[idx]\n",
    "\n",
    "        probs = np.array(y_pred.tolist()) # compute_probs(y_pred)\n",
    "        probs_tensor = y_pred #torch.tensor(probs, dtype=torch.float).to(device)\n",
    "\n",
    "        #probs = threshold_probs_by_nb(probs=probs, doc_lens=[probs.shape[0]], average_number_of_sentences_per_document=average_number_of_sentences_per_document)\n",
    "        #probs = threshold_probs_by_prop(probs=probs, doc_lens=[probs.shape[0]], average_proportion_of_sentences_per_document=average_proportion_of_sentences_per_document)\n",
    "        indices = torch.argsort(y_pred, descending=True)\n",
    "\n",
    "        y_pred_thresh = []\n",
    "\n",
    "        txt = \"\"\n",
    "\n",
    "        for i in range(min(len(doc), y_pred.shape[0])):\n",
    "            txt = txt + \". \" + doc[indices[i]]\n",
    "            y_pred_thresh.append(indices[i])\n",
    "            if len(txt) >= len(summaries):\n",
    "                break\n",
    "\n",
    "        y_pred_thresh.sort()\n",
    "\n",
    "        txt = \"\"\n",
    "\n",
    "        for i in y_pred_thresh:#range(min(len(doc), y_pred.shape[0])):\n",
    "            txt = txt + \". \" + doc[i]\n",
    "\n",
    "        n = min(len(txt), len(summaries))\n",
    "\n",
    "        while n < len(txt) and txt[n].isalnum():\n",
    "            n += 1\n",
    "\n",
    "        txt = txt[:n]\n",
    "\n",
    "        #assert len(txt) - len(summaries) <= 20\n",
    "\n",
    "        scores = scorer.score(summaries, txt)\n",
    "        arr_rouge1.append(scores[\"rouge1\"].fmeasure)\n",
    "        arr_rouge2.append(scores[\"rouge2\"].fmeasure)\n",
    "        arr_rougeL.append(scores[\"rougeL\"].fmeasure)\n",
    "\n",
    "        mae.append(mae_fn(probs_tensor, targets).item())\n",
    "\n",
    "        accuracy.append(accuracy_nb_sent_per_doc_fn(probs=probs, targets=targets.cpu().detach().numpy(), doc_lens=[len(probs)], average_number_of_sentences_per_document=average_number_of_sentences_per_document))\n",
    "        #accuracy.append(accuracy_prop_sent_per_doc_fn(probs=probs, targets=targets.cpu().detach().numpy(), doc_lens=[len(probs)], average_proportion_of_sentences_per_document=average_proportion_of_sentences_per_document))\n",
    "\n",
    "        tepoch.set_postfix(accuracy = np.mean(accuracy), mae = np.mean(mae), rouge1 = np.mean(arr_rouge1), rouge2 = np.mean(arr_rouge2), rougeL = np.mean(arr_rougeL))\n",
    "        idx += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "5c1fd15f",
   "metadata": {},
   "outputs": [],
   "source": [
    "test_metrics = {}\n",
    "test_metrics[\"accuracy\"] = np.mean(accuracy)\n",
    "test_metrics[\"mae\"]      = np.mean(mae)\n",
    "test_metrics[\"rouge1\"]   = np.mean(arr_rouge1)\n",
    "test_metrics[\"rouge2\"]   = np.mean(arr_rouge2)\n",
    "test_metrics[\"rougeL\"]   = np.mean(arr_rougeL)\n",
    "\n",
    "# Save to file in JSON format\n",
    "\n",
    "with open(checkpoints_folder + \"/test_metrics.json\", 'w') as fp:\n",
    "    json.dump(test_metrics, fp)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "id": "496646f2",
   "metadata": {},
   "source": [
    "### Lead-3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "4c96cdd0",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "First-n-char': 100%|██████████| 100/100 [00:00<00:00, 119.25batch/s, rouge1=0.435, rouge2=0.241, rougeL=0.325]\n"
     ]
    }
   ],
   "source": [
    "arr_rouge1 = []\n",
    "arr_rouge2 = []\n",
    "arr_rougeL = []\n",
    "mae = []\n",
    "accuracy = []\n",
    "\n",
    "idx = 0\n",
    "with tqdm(test_iter, unit=\"batch\", total=len(test_iter)) as tepoch:\n",
    "    for batch in tepoch:\n",
    "        tepoch.set_description(\"First-n-char'\")\n",
    "        doc = test_set[\"doc_splited\"].iloc[idx]\n",
    "\n",
    "        txt = \"\"\n",
    "\n",
    "        for i in range(len(doc)):\n",
    "            txt = txt + doc[i]\n",
    "\n",
    "        summaries = test_set[args.sum_col_name].iloc[idx]\n",
    "\n",
    "        n = min(len(txt), len(summaries))\n",
    "\n",
    "        while n < len(txt) and txt[n].isalnum():\n",
    "            n += 1\n",
    "\n",
    "        txt = txt[:n]\n",
    "\n",
    "        scores = scorer.score(summaries, txt)\n",
    "        arr_rouge1.append(scores[\"rouge1\"].fmeasure)\n",
    "        arr_rouge2.append(scores[\"rouge2\"].fmeasure)\n",
    "        arr_rougeL.append(scores[\"rougeL\"].fmeasure)\n",
    "\n",
    "        idx += 1\n",
    "\n",
    "        tepoch.set_postfix(rouge1 = np.mean(arr_rouge1), rouge2 = np.mean(arr_rouge2), rougeL = np.mean(arr_rougeL))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "78a3c5d9",
   "metadata": {},
   "source": [
    "### First n char"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "a4c96c70",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "First-n-char': 100%|██████████| 100/100 [00:00<00:00, 122.38batch/s, rouge1=0.435, rouge2=0.241, rougeL=0.325]\n"
     ]
    }
   ],
   "source": [
    "arr_rouge1 = []\n",
    "arr_rouge2 = []\n",
    "arr_rougeL = []\n",
    "mae = []\n",
    "accuracy = []\n",
    "\n",
    "idx = 0\n",
    "with tqdm(test_iter, unit=\"batch\", total=len(test_iter)) as tepoch:\n",
    "    for batch in tepoch:\n",
    "        tepoch.set_description(\"First-n-char'\")\n",
    "        doc = test_set[\"doc_splited\"].iloc[idx]\n",
    "\n",
    "        txt = \"\"\n",
    "\n",
    "        for i in range(len(doc)):\n",
    "            txt = txt + doc[i]\n",
    "\n",
    "        summaries = test_set[args.sum_col_name].iloc[idx]\n",
    "\n",
    "        n = min(len(txt), len(summaries))\n",
    "\n",
    "        while n < len(txt) and txt[n].isalnum():\n",
    "            n += 1\n",
    "\n",
    "        txt = txt[:n]\n",
    "\n",
    "        scores = scorer.score(summaries, txt)\n",
    "        arr_rouge1.append(scores[\"rouge1\"].fmeasure)\n",
    "        arr_rouge2.append(scores[\"rouge2\"].fmeasure)\n",
    "        arr_rougeL.append(scores[\"rougeL\"].fmeasure)\n",
    "\n",
    "        idx += 1\n",
    "\n",
    "        tepoch.set_postfix(rouge1 = np.mean(arr_rouge1), rouge2 = np.mean(arr_rouge2), rougeL = np.mean(arr_rougeL))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "e4f4796c",
   "metadata": {},
   "outputs": [
    {
     "ename": "SystemExit",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "An exception has occurred, use %tb to see the full traceback.\n",
      "\u001b[0;31mSystemExit\u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/raoufdine/anaconda3/lib/python3.9/site-packages/IPython/core/interactiveshell.py:3465: UserWarning: To exit: use 'exit', 'quit', or Ctrl-D.\n",
      "  warn(\"To exit: use 'exit', 'quit', or Ctrl-D.\", stacklevel=1)\n"
     ]
    }
   ],
   "source": [
    "# Does not excute this cell if you want to execute the following cells.\n",
    "import sys\n",
    "sys.exit()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "base",
   "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": "3.9.13"
  },
  "vscode": {
   "interpreter": {
    "hash": "4f810f0bde7d7b5dc90f9e6832822547e13e234c53731a97fc91254323942f40"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}