NP-Classifier.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['3', '7', 'o', '[N]', '[NH3+]', '\\\\', '9', '[N@@+]', 'r', 'l', '[O+]', '[s@]', '[NH-]', ')', 'P', '[S-]', '2', '/', '(', '6', '[C-]', '[nH+]', '[O-]', '[C@@]', '[C@]', '[o+]', '[CH-]', '[c-]', '[O]', '1', '[S@@]', '[P@@H]', '[c+]', '%', 'I', 'L', 'R', '=', '[C]', '[NH+]', '8', '[N@+]', '[P@]', '[CH]', 'O', '[C+]', '0', '[s+]', 's', '[C@@H]', '[Si]', '[CH+]', '[P+]', 'S', '[N-]', 'C', '[NH2+]', '[S@]', '5', 'c', '[n-]', '[OH+]', '[C@H]', 'F', '[nH]', '[P@@]', '[n+]', '[P@H]', '[N+]', 'B', '[cH-]', '#', 'N', '4', '[CH2-]', 'n', '[S+]']\n",
      "74759\n",
      "18691\n",
      "2 properties\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "import math\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "from torch.autograd import Variable\n",
    "from torch.utils.data import Dataset, DataLoader\n",
    "import data_structs\n",
    "import re\n",
    "import string\n",
    "from torch.nn.utils.rnn import pack_padded_sequence, pad_packed_sequence\n",
    "import numpy as np\n",
    "from sklearn import metrics\n",
    "import csv\n",
    "import pandas as pd\n",
    "\n",
    "\n",
    "#data processing\n",
    "class ProDataset(Dataset):\n",
    "    \"\"\" Diabetes dataset.\"\"\"\n",
    "\n",
    "    # Initialize your data, download, etc.\n",
    "    def __init__(self, is_train_set=False):\n",
    "        filename = './data/npscore-train.csv' if is_train_set else './data/npscore-test.csv'\n",
    "        with open(filename, \"rt\") as f:\n",
    "            reader = csv.reader(f)\n",
    "            rows = list(reader)\n",
    "        \n",
    "        self.lines = [row[0] for row in rows]\n",
    "        self.properties = [row[1] for row in rows]\n",
    "        self.len = len(self.properties)\n",
    "\n",
    "        self.property_list = list(sorted(set(self.properties)))\n",
    "\n",
    "    def __getitem__(self, index):\n",
    "        return self.lines[index], self.properties[index]\n",
    "\n",
    "    def __len__(self):\n",
    "        return self.len\n",
    "\n",
    "    def get_properties(self):\n",
    "        return self.property_list\n",
    "\n",
    "    def get_property(self, id):\n",
    "\n",
    "        return self.property_list[id]\n",
    "\n",
    "    def get_property_id(self, property):\n",
    "        return self.property_list.index(property)\n",
    "\n",
    "\n",
    "\n",
    "# Voc loaders\n",
    "with open('./data/Voc-np', 'r') as f:\n",
    "    chars = f.read().split()\n",
    "print(chars)\n",
    "all_letters = chars\n",
    "N_CHARS = len(all_letters)\n",
    "\n",
    "# Parameters and DataLoaders\n",
    "HIDDEN_SIZE = 128\n",
    "N_LAYERS = 2\n",
    "BATCH_SIZE = 256 #should be changed\n",
    "N_EPOCHS = 30\n",
    "\n",
    "\n",
    "test_dataset = ProDataset(is_train_set=False)\n",
    "test_loader = DataLoader(dataset=test_dataset,\n",
    "                         batch_size=BATCH_SIZE, shuffle=True)\n",
    "\n",
    "\n",
    "train_dataset = ProDataset(is_train_set=True)\n",
    "train_loader = DataLoader(dataset=train_dataset,\n",
    "                          batch_size=BATCH_SIZE, shuffle=True)\n",
    "\n",
    "print(len(train_loader.dataset))\n",
    "print(len(test_loader.dataset))\n",
    "\n",
    "N_PROPERTIES = len(train_dataset.get_properties())\n",
    "print(N_PROPERTIES, \"properties\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# Some utility functions\n",
    "def time_since(since):\n",
    "    s = time.time() - since\n",
    "    m = math.floor(s / 60)\n",
    "    s -= m * 60\n",
    "    return '%dm %ds' % (m, s)\n",
    "\n",
    "\n",
    "def create_variable(tensor):\n",
    "    # Do cuda() before wrapping with variable\n",
    "    if torch.cuda.is_available():\n",
    "        return Variable(tensor.cuda())\n",
    "    else:\n",
    "        return Variable(tensor)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "# pad sequences and sort the tensor\n",
    "def pad_sequences(vectorized_seqs, seq_lengths, properties):\n",
    "    seq_tensor = torch.zeros((len(vectorized_seqs), seq_lengths.max())).long()\n",
    "    for idx, (seq, seq_len) in enumerate(zip(vectorized_seqs, seq_lengths)):\n",
    "        seq_tensor[idx, :seq_len] = torch.LongTensor(seq)\n",
    "\n",
    "    # Sort tensors by their length\n",
    "    seq_lengths, perm_idx = seq_lengths.sort(0, descending=True)\n",
    "    seq_tensor = seq_tensor[perm_idx]\n",
    "\n",
    "    # Also sort the target (countries) in the same order\n",
    "    target = properties2tensor(properties)\n",
    "    if len(properties):\n",
    "        target = target[perm_idx]\n",
    "\n",
    "    # Return variables\n",
    "    # DataParallel requires everything to be a Variable\n",
    "    return create_variable(seq_tensor), \\\n",
    "        create_variable(seq_lengths), \\\n",
    "        create_variable(target)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "59\n"
     ]
    }
   ],
   "source": [
    "# Create necessary variables, lengths, and target\n",
    "def make_variables(lines, properties):\n",
    "    sequence_and_length = [line2voc_arr(line) for line in lines]\n",
    "    vectorized_seqs = [sl[0] for sl in sequence_and_length]\n",
    "    seq_lengths = torch.LongTensor([sl[1] for sl in sequence_and_length])\n",
    "    return pad_sequences(vectorized_seqs, seq_lengths, properties)\n",
    "\n",
    "#used to put char into acsii index,ord() will return the index of ascii ie. ord(c) =99\n",
    "def str2ascii_arr(msg):\n",
    "    arr = [ord(c) for c in msg]\n",
    "    return arr, len(arr)\n",
    "\n",
    "def letterToIndex(letter):\n",
    "    return all_letters.index(letter)\n",
    "print(letterToIndex('c'))\n",
    "\n",
    "def line2voc_arr(line):\n",
    "    arr = []\n",
    "    regex = '(\\[[^\\[\\]]{1,10}\\])'\n",
    "    line = data_structs.replace_halogen(line)\n",
    "    char_list = re.split(regex, line)\n",
    "    for li, char in enumerate(char_list):\n",
    "        if char.startswith('['):\n",
    "               arr.append(letterToIndex(char)) \n",
    "        else:\n",
    "            chars = [unit for unit in char]\n",
    "\n",
    "            for i, unit in enumerate(chars):\n",
    "                arr.append(letterToIndex(unit))\n",
    "    return arr, len(arr)\n",
    "\n",
    "#print(make_variables('C1ccccC1',0))\n",
    "\n",
    "def properties2tensor(properties):\n",
    "    property_ids = [train_dataset.get_property_id(\n",
    "        property) for property in properties]\n",
    "    return torch.LongTensor(property_ids)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RNNClassifier(\n",
      "  (embedding): Embedding(77, 128)\n",
      "  (gru): GRU(128, 128, dropout=0.2, bidirectional=True)\n",
      "  (fc): Linear(in_features=128, out_features=2, bias=True)\n",
      ")\n",
      "tensor([[-0.3210,  0.0787]], device='cuda:0')\n",
      "tensor([[ 1]], device='cuda:0')\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/data0/zhengsj/anaconda3/lib/python3.6/site-packages/torch/nn/modules/rnn.py:38: UserWarning: dropout option adds dropout after all but last recurrent layer, so non-zero dropout expects num_layers greater than 1, but got dropout=0.2 and num_layers=1\n",
      "  \"num_layers={}\".format(dropout, num_layers))\n"
     ]
    }
   ],
   "source": [
    "class RNNClassifier(nn.Module):\n",
    "    # Our model\n",
    "\n",
    "    def __init__(self, input_size, hidden_size, output_size, n_layers=1, bidirectional=True):\n",
    "        super(RNNClassifier, self).__init__()\n",
    "        self.hidden_size = hidden_size\n",
    "        self.n_layers = n_layers\n",
    "        self.n_directions = int(bidirectional) + 1\n",
    "\n",
    "        self.embedding = nn.Embedding(input_size, hidden_size)\n",
    "        self.gru = nn.GRU(hidden_size, hidden_size, n_layers,dropout = 0.2,\n",
    "                          bidirectional=bidirectional)\n",
    "        self.fc = nn.Linear(hidden_size, output_size)\n",
    "\n",
    "    def forward(self, input, seq_lengths):\n",
    "        # Note: we run this all at once (over the whole input sequence)\n",
    "        # input shape: B x S (input size)\n",
    "        # transpose to make S(sequence) x B (batch)\n",
    "        input = input.t()\n",
    "        batch_size = input.size(1)\n",
    "\n",
    "        # Make a hidden\n",
    "        hidden = self._init_hidden(batch_size)\n",
    "\n",
    "        # Embedding S x B -> S x B x I (embedding size)\n",
    "        embedded = self.embedding(input)\n",
    "\n",
    "        # Pack them up nicely\n",
    "        gru_input = pack_padded_sequence(\n",
    "            embedded, seq_lengths.data.cpu().numpy())\n",
    "\n",
    "        # To compact weights again call flatten_parameters().\n",
    "        self.gru.flatten_parameters()\n",
    "        output, hidden = self.gru(gru_input, hidden)\n",
    "        #print(hidden)\n",
    "        # Use the last layer output as FC's input\n",
    "        # No need to unpack, since we are going to use hidden\n",
    "        fc_output = self.fc(hidden[-1])\n",
    "        return fc_output\n",
    "\n",
    "    def _init_hidden(self, batch_size):\n",
    "        hidden = torch.zeros(self.n_layers * self.n_directions,\n",
    "                             batch_size, self.hidden_size)\n",
    "        return create_variable(hidden)\n",
    "    \n",
    "\n",
    "classifier = RNNClassifier(N_CHARS, HIDDEN_SIZE, N_PROPERTIES, 1)\n",
    "print(classifier)\n",
    "classifier.cuda()\n",
    "input, seq_lengths, target = make_variables([\"NC(=O)CC(NC(=O)c1ccco1)C(=O)O\"],['NP'])\n",
    "output= classifier(input,seq_lengths)\n",
    "print(output)\n",
    "output = output.data.max(1, keepdim = True)[1]\n",
    "print(output)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Train cycle\n",
    "def train():\n",
    "    total_loss = 0\n",
    "    \n",
    "    correct = 0\n",
    "    train_data_size = len(train_loader.dataset)\n",
    "    print(train_data_size)\n",
    "    \n",
    "    for i,(lines, properties) in enumerate (train_loader, 1):\n",
    "        input, seq_lengths, target = make_variables(lines, properties)\n",
    "        #print(seq_lengths)\n",
    "        output = classifier(input, seq_lengths)\n",
    "        #print(output)\n",
    "        loss = criterion(output, target)\n",
    "        total_loss += loss.data[0]\n",
    "\n",
    "        classifier.zero_grad()\n",
    "        loss.backward()\n",
    "        optimizer.step()\n",
    "        \n",
    "        pred = output.data.max(1, keepdim = True)[1]\n",
    "        correct += pred.eq(target.data.view_as(pred)).cpu().sum() #view_as 转化成像pred一样的size\n",
    "        #print('pred.size= ',pred.size())\n",
    "        #print(target.data.view_as(pred).size())\n",
    "        if i % 10 == 0:\n",
    "            print('[{}] Train Epoch: {} [{}/{} ({:.0f}%)]\\tLoss: {:.2f}'.format(\n",
    "                time_since(start), epoch,  i *\n",
    "                len(lines), len(train_loader.dataset),\n",
    "                100. * i * len(lines) / len(train_loader.dataset),\n",
    "                total_loss / i * len(lines)))\n",
    "\n",
    "    print('\\nTrain set: Accuracy: {}/{} ({:.0f}%) '.format(\n",
    "        correct, train_data_size, 100. * correct / train_data_size))   \n",
    "#     print('output.data = ',output.data.size())\n",
    "#     print('target.data = ',target.data.size())\n",
    "#     print('pred.size= ',pred.size())\n",
    "#     print(target.data.view_as(pred).size())    \n",
    "    return total_loss\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    " #Testing cycle\n",
    "def test(line=None):\n",
    "    # Predict for a given name\n",
    "    if line:\n",
    "        input, seq_lengths, target = make_variables([line], [])\n",
    "        output = classifier(input, seq_lengths)\n",
    "        pred = output.data.max(1, keepdim=True)[1]\n",
    "        property_id = pred.cpu().numpy()[0][0]\n",
    "        print(line, \"is\", train_dataset.get_property(property_id))\n",
    "        return\n",
    "\n",
    "    print(\"evaluating trained model ...\")\n",
    "    correct = 0\n",
    "    train_data_size = len(test_loader.dataset)\n",
    "    totalpred = 0\n",
    "    recall = 0\n",
    "    postive_pred = np.array([0])\n",
    "    all_target = np.array([0])\n",
    "    AUC = 0\n",
    "    for i,(lines, properties) in enumerate (test_loader, 1):\n",
    "        input, seq_lengths, target = make_variables(lines, properties)\n",
    "        #print(seq_lengths)\n",
    "        output = classifier(input, seq_lengths)\n",
    "        pred = output.data.max(1, keepdim = True)[1]\n",
    "        correct += pred.eq(target.data.view_as(pred)).cpu().sum()\n",
    "        totalpred = totalpred + pred.sum()\n",
    "        recall += metrics.recall_score(target.data.view_as(pred), pred)\n",
    "        #pred = output.data.max(1, keepdim=True)[1]\n",
    "        pos = output[:,0]  #get the stable        \n",
    "        postive_pred=np.concatenate((postive_pred,pos.data.cpu().numpy()),axis = 0)\n",
    "        all_target=np.concatenate((all_target,target.data.cpu().numpy()),axis = 0)\n",
    "\n",
    "        #if pred :\n",
    "        #   total_pred = total_pred+1\n",
    "        \n",
    "\n",
    "        \n",
    "#         fpr, tpr, thresholds = metrics.roc_curve(target.data.cpu().numpy(), pos.data.cpu().numpy(), pos_label=0)\n",
    "#         AUC = metrics.auc(fpr, tpr)\n",
    "#         print(AUC)\n",
    "        \n",
    "\n",
    "        #recall = metrics.recall_score(target.data.view_as(pred), pred)\n",
    "        #roc_score = roc\n",
    "\n",
    "    fpr, tpr, thresholds = metrics.roc_curve(all_target, postive_pred, pos_label=0)\n",
    "    AUC = metrics.auc(fpr, tpr)\n",
    "    print('AUC = ',AUC)\n",
    "    print('\\nTest set: Accuracy: {}/{} ({:.0f}%) Recall: ({:.0f}%)\\n'.format(\n",
    "        correct, train_data_size, 100. * correct / train_data_size,100. * recall / train_data_size*64))\n",
    "    \n",
    "    return 100. * correct / train_data_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training for 30 epochs...\n",
      "74759\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/data0/zhengsj/anaconda3/lib/python3.6/site-packages/ipykernel_launcher.py:15: UserWarning: invalid index of a 0-dim tensor. This will be an error in PyTorch 0.5. Use tensor.item() to convert a 0-dim tensor to a Python number\n",
      "  from ipykernel import kernelapp as app\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0m 0s] Train Epoch: 1 [2560/74759 (3%)]\tLoss: 177.26\n",
      "[0m 1s] Train Epoch: 1 [5120/74759 (7%)]\tLoss: 173.04\n",
      "[0m 2s] Train Epoch: 1 [7680/74759 (10%)]\tLoss: 170.91\n",
      "[0m 3s] Train Epoch: 1 [10240/74759 (14%)]\tLoss: 168.56\n",
      "[0m 4s] Train Epoch: 1 [12800/74759 (17%)]\tLoss: 165.43\n",
      "[0m 5s] Train Epoch: 1 [15360/74759 (21%)]\tLoss: 161.61\n",
      "[0m 6s] Train Epoch: 1 [17920/74759 (24%)]\tLoss: 157.43\n",
      "[0m 7s] Train Epoch: 1 [20480/74759 (27%)]\tLoss: 153.50\n",
      "[0m 8s] Train Epoch: 1 [23040/74759 (31%)]\tLoss: 149.02\n",
      "[0m 9s] Train Epoch: 1 [25600/74759 (34%)]\tLoss: 144.92\n",
      "[0m 9s] Train Epoch: 1 [28160/74759 (38%)]\tLoss: 141.02\n",
      "[0m 10s] Train Epoch: 1 [30720/74759 (41%)]\tLoss: 137.89\n",
      "[0m 11s] Train Epoch: 1 [33280/74759 (45%)]\tLoss: 135.20\n",
      "[0m 12s] Train Epoch: 1 [35840/74759 (48%)]\tLoss: 132.60\n",
      "[0m 13s] Train Epoch: 1 [38400/74759 (51%)]\tLoss: 130.22\n",
      "[0m 14s] Train Epoch: 1 [40960/74759 (55%)]\tLoss: 128.23\n",
      "[0m 15s] Train Epoch: 1 [43520/74759 (58%)]\tLoss: 126.30\n",
      "[0m 16s] Train Epoch: 1 [46080/74759 (62%)]\tLoss: 124.44\n",
      "[0m 17s] Train Epoch: 1 [48640/74759 (65%)]\tLoss: 122.69\n",
      "[0m 17s] Train Epoch: 1 [51200/74759 (68%)]\tLoss: 121.18\n",
      "[0m 18s] Train Epoch: 1 [53760/74759 (72%)]\tLoss: 119.88\n",
      "[0m 19s] Train Epoch: 1 [56320/74759 (75%)]\tLoss: 118.62\n",
      "[0m 20s] Train Epoch: 1 [58880/74759 (79%)]\tLoss: 117.49\n",
      "[0m 21s] Train Epoch: 1 [61440/74759 (82%)]\tLoss: 116.32\n",
      "[0m 22s] Train Epoch: 1 [64000/74759 (86%)]\tLoss: 115.26\n",
      "[0m 22s] Train Epoch: 1 [66560/74759 (89%)]\tLoss: 114.34\n",
      "[0m 23s] Train Epoch: 1 [69120/74759 (92%)]\tLoss: 113.33\n",
      "[0m 24s] Train Epoch: 1 [71680/74759 (96%)]\tLoss: 112.46\n",
      "[0m 25s] Train Epoch: 1 [74240/74759 (99%)]\tLoss: 111.31\n",
      "\n",
      "Train set: Accuracy: 58452/74759 (78%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.922927630082178\n",
      "\n",
      "Test set: Accuracy: 15328/18691 (82%) Recall: (19%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[0m 29s] Train Epoch: 2 [2560/74759 (3%)]\tLoss: 89.16\n",
      "[0m 30s] Train Epoch: 2 [5120/74759 (7%)]\tLoss: 85.90\n",
      "[0m 31s] Train Epoch: 2 [7680/74759 (10%)]\tLoss: 85.16\n",
      "[0m 32s] Train Epoch: 2 [10240/74759 (14%)]\tLoss: 84.28\n",
      "[0m 33s] Train Epoch: 2 [12800/74759 (17%)]\tLoss: 84.53\n",
      "[0m 33s] Train Epoch: 2 [15360/74759 (21%)]\tLoss: 83.54\n",
      "[0m 34s] Train Epoch: 2 [17920/74759 (24%)]\tLoss: 83.02\n",
      "[0m 35s] Train Epoch: 2 [20480/74759 (27%)]\tLoss: 82.64\n",
      "[0m 36s] Train Epoch: 2 [23040/74759 (31%)]\tLoss: 82.54\n",
      "[0m 37s] Train Epoch: 2 [25600/74759 (34%)]\tLoss: 82.26\n",
      "[0m 38s] Train Epoch: 2 [28160/74759 (38%)]\tLoss: 81.76\n",
      "[0m 39s] Train Epoch: 2 [30720/74759 (41%)]\tLoss: 81.45\n",
      "[0m 40s] Train Epoch: 2 [33280/74759 (45%)]\tLoss: 80.88\n",
      "[0m 40s] Train Epoch: 2 [35840/74759 (48%)]\tLoss: 80.49\n",
      "[0m 41s] Train Epoch: 2 [38400/74759 (51%)]\tLoss: 80.10\n",
      "[0m 42s] Train Epoch: 2 [40960/74759 (55%)]\tLoss: 79.75\n",
      "[0m 43s] Train Epoch: 2 [43520/74759 (58%)]\tLoss: 79.72\n",
      "[0m 44s] Train Epoch: 2 [46080/74759 (62%)]\tLoss: 79.67\n",
      "[0m 45s] Train Epoch: 2 [48640/74759 (65%)]\tLoss: 79.42\n",
      "[0m 46s] Train Epoch: 2 [51200/74759 (68%)]\tLoss: 79.32\n",
      "[0m 47s] Train Epoch: 2 [53760/74759 (72%)]\tLoss: 79.00\n",
      "[0m 48s] Train Epoch: 2 [56320/74759 (75%)]\tLoss: 78.75\n",
      "[0m 49s] Train Epoch: 2 [58880/74759 (79%)]\tLoss: 78.71\n",
      "[0m 50s] Train Epoch: 2 [61440/74759 (82%)]\tLoss: 78.52\n",
      "[0m 51s] Train Epoch: 2 [64000/74759 (86%)]\tLoss: 78.33\n",
      "[0m 51s] Train Epoch: 2 [66560/74759 (89%)]\tLoss: 78.17\n",
      "[0m 52s] Train Epoch: 2 [69120/74759 (92%)]\tLoss: 77.93\n",
      "[0m 53s] Train Epoch: 2 [71680/74759 (96%)]\tLoss: 77.64\n",
      "[0m 54s] Train Epoch: 2 [74240/74759 (99%)]\tLoss: 77.45\n",
      "\n",
      "Train set: Accuracy: 64499/74759 (86%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9389226867790345\n",
      "\n",
      "Test set: Accuracy: 15805/18691 (84%) Recall: (20%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[0m 59s] Train Epoch: 3 [2560/74759 (3%)]\tLoss: 76.47\n",
      "[0m 59s] Train Epoch: 3 [5120/74759 (7%)]\tLoss: 77.64\n",
      "[1m 0s] Train Epoch: 3 [7680/74759 (10%)]\tLoss: 76.50\n",
      "[1m 1s] Train Epoch: 3 [10240/74759 (14%)]\tLoss: 75.12\n",
      "[1m 2s] Train Epoch: 3 [12800/74759 (17%)]\tLoss: 74.09\n",
      "[1m 2s] Train Epoch: 3 [15360/74759 (21%)]\tLoss: 73.12\n",
      "[1m 3s] Train Epoch: 3 [17920/74759 (24%)]\tLoss: 72.40\n",
      "[1m 4s] Train Epoch: 3 [20480/74759 (27%)]\tLoss: 71.85\n",
      "[1m 5s] Train Epoch: 3 [23040/74759 (31%)]\tLoss: 71.95\n",
      "[1m 6s] Train Epoch: 3 [25600/74759 (34%)]\tLoss: 71.50\n",
      "[1m 7s] Train Epoch: 3 [28160/74759 (38%)]\tLoss: 71.16\n",
      "[1m 7s] Train Epoch: 3 [30720/74759 (41%)]\tLoss: 70.74\n",
      "[1m 8s] Train Epoch: 3 [33280/74759 (45%)]\tLoss: 70.23\n",
      "[1m 9s] Train Epoch: 3 [35840/74759 (48%)]\tLoss: 70.30\n",
      "[1m 10s] Train Epoch: 3 [38400/74759 (51%)]\tLoss: 70.09\n",
      "[1m 11s] Train Epoch: 3 [40960/74759 (55%)]\tLoss: 69.78\n",
      "[1m 12s] Train Epoch: 3 [43520/74759 (58%)]\tLoss: 69.69\n",
      "[1m 13s] Train Epoch: 3 [46080/74759 (62%)]\tLoss: 69.69\n",
      "[1m 13s] Train Epoch: 3 [48640/74759 (65%)]\tLoss: 69.39\n",
      "[1m 14s] Train Epoch: 3 [51200/74759 (68%)]\tLoss: 69.20\n",
      "[1m 15s] Train Epoch: 3 [53760/74759 (72%)]\tLoss: 68.95\n",
      "[1m 16s] Train Epoch: 3 [56320/74759 (75%)]\tLoss: 68.78\n",
      "[1m 17s] Train Epoch: 3 [58880/74759 (79%)]\tLoss: 68.55\n",
      "[1m 18s] Train Epoch: 3 [61440/74759 (82%)]\tLoss: 68.41\n",
      "[1m 18s] Train Epoch: 3 [64000/74759 (86%)]\tLoss: 68.17\n",
      "[1m 19s] Train Epoch: 3 [66560/74759 (89%)]\tLoss: 68.00\n",
      "[1m 20s] Train Epoch: 3 [69120/74759 (92%)]\tLoss: 67.88\n",
      "[1m 21s] Train Epoch: 3 [71680/74759 (96%)]\tLoss: 67.76\n",
      "[1m 22s] Train Epoch: 3 [74240/74759 (99%)]\tLoss: 67.74\n",
      "\n",
      "Train set: Accuracy: 65729/74759 (87%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9577621654701685\n",
      "\n",
      "Test set: Accuracy: 16500/18691 (88%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[1m 26s] Train Epoch: 4 [2560/74759 (3%)]\tLoss: 68.62\n",
      "[1m 27s] Train Epoch: 4 [5120/74759 (7%)]\tLoss: 68.30\n",
      "[1m 28s] Train Epoch: 4 [7680/74759 (10%)]\tLoss: 68.46\n",
      "[1m 29s] Train Epoch: 4 [10240/74759 (14%)]\tLoss: 66.93\n",
      "[1m 29s] Train Epoch: 4 [12800/74759 (17%)]\tLoss: 66.20\n",
      "[1m 30s] Train Epoch: 4 [15360/74759 (21%)]\tLoss: 65.19\n",
      "[1m 31s] Train Epoch: 4 [17920/74759 (24%)]\tLoss: 64.47\n",
      "[1m 32s] Train Epoch: 4 [20480/74759 (27%)]\tLoss: 64.78\n",
      "[1m 33s] Train Epoch: 4 [23040/74759 (31%)]\tLoss: 64.24\n",
      "[1m 34s] Train Epoch: 4 [25600/74759 (34%)]\tLoss: 63.52\n",
      "[1m 35s] Train Epoch: 4 [28160/74759 (38%)]\tLoss: 63.20\n",
      "[1m 35s] Train Epoch: 4 [30720/74759 (41%)]\tLoss: 62.67\n",
      "[1m 36s] Train Epoch: 4 [33280/74759 (45%)]\tLoss: 62.69\n",
      "[1m 37s] Train Epoch: 4 [35840/74759 (48%)]\tLoss: 62.67\n",
      "[1m 38s] Train Epoch: 4 [38400/74759 (51%)]\tLoss: 62.92\n",
      "[1m 39s] Train Epoch: 4 [40960/74759 (55%)]\tLoss: 62.89\n",
      "[1m 40s] Train Epoch: 4 [43520/74759 (58%)]\tLoss: 63.00\n",
      "[1m 40s] Train Epoch: 4 [46080/74759 (62%)]\tLoss: 62.93\n",
      "[1m 41s] Train Epoch: 4 [48640/74759 (65%)]\tLoss: 63.04\n",
      "[1m 42s] Train Epoch: 4 [51200/74759 (68%)]\tLoss: 62.92\n",
      "[1m 43s] Train Epoch: 4 [53760/74759 (72%)]\tLoss: 62.72\n",
      "[1m 44s] Train Epoch: 4 [56320/74759 (75%)]\tLoss: 62.77\n",
      "[1m 45s] Train Epoch: 4 [58880/74759 (79%)]\tLoss: 62.37\n",
      "[1m 45s] Train Epoch: 4 [61440/74759 (82%)]\tLoss: 62.31\n",
      "[1m 46s] Train Epoch: 4 [64000/74759 (86%)]\tLoss: 62.23\n",
      "[1m 47s] Train Epoch: 4 [66560/74759 (89%)]\tLoss: 62.23\n",
      "[1m 48s] Train Epoch: 4 [69120/74759 (92%)]\tLoss: 62.08\n",
      "[1m 49s] Train Epoch: 4 [71680/74759 (96%)]\tLoss: 61.97\n",
      "[1m 50s] Train Epoch: 4 [74240/74759 (99%)]\tLoss: 61.91\n",
      "\n",
      "Train set: Accuracy: 66368/74759 (88%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9642760946833178\n",
      "\n",
      "Test set: Accuracy: 16642/18691 (89%) Recall: (24%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[1m 54s] Train Epoch: 5 [2560/74759 (3%)]\tLoss: 59.18\n",
      "[1m 54s] Train Epoch: 5 [5120/74759 (7%)]\tLoss: 59.38\n",
      "[1m 56s] Train Epoch: 5 [7680/74759 (10%)]\tLoss: 57.79\n",
      "[1m 56s] Train Epoch: 5 [10240/74759 (14%)]\tLoss: 57.51\n",
      "[1m 57s] Train Epoch: 5 [12800/74759 (17%)]\tLoss: 57.22\n",
      "[1m 58s] Train Epoch: 5 [15360/74759 (21%)]\tLoss: 57.29\n",
      "[1m 59s] Train Epoch: 5 [17920/74759 (24%)]\tLoss: 57.37\n",
      "[2m 0s] Train Epoch: 5 [20480/74759 (27%)]\tLoss: 57.44\n",
      "[2m 1s] Train Epoch: 5 [23040/74759 (31%)]\tLoss: 57.25\n",
      "[2m 2s] Train Epoch: 5 [25600/74759 (34%)]\tLoss: 56.95\n",
      "[2m 3s] Train Epoch: 5 [28160/74759 (38%)]\tLoss: 56.93\n",
      "[2m 3s] Train Epoch: 5 [30720/74759 (41%)]\tLoss: 56.79\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[2m 4s] Train Epoch: 5 [33280/74759 (45%)]\tLoss: 56.85\n",
      "[2m 5s] Train Epoch: 5 [35840/74759 (48%)]\tLoss: 56.74\n",
      "[2m 6s] Train Epoch: 5 [38400/74759 (51%)]\tLoss: 56.85\n",
      "[2m 7s] Train Epoch: 5 [40960/74759 (55%)]\tLoss: 57.00\n",
      "[2m 8s] Train Epoch: 5 [43520/74759 (58%)]\tLoss: 56.71\n",
      "[2m 9s] Train Epoch: 5 [46080/74759 (62%)]\tLoss: 56.60\n",
      "[2m 10s] Train Epoch: 5 [48640/74759 (65%)]\tLoss: 56.62\n",
      "[2m 11s] Train Epoch: 5 [51200/74759 (68%)]\tLoss: 56.71\n",
      "[2m 11s] Train Epoch: 5 [53760/74759 (72%)]\tLoss: 56.74\n",
      "[2m 12s] Train Epoch: 5 [56320/74759 (75%)]\tLoss: 56.77\n",
      "[2m 13s] Train Epoch: 5 [58880/74759 (79%)]\tLoss: 56.77\n",
      "[2m 14s] Train Epoch: 5 [61440/74759 (82%)]\tLoss: 56.65\n",
      "[2m 15s] Train Epoch: 5 [64000/74759 (86%)]\tLoss: 56.74\n",
      "[2m 15s] Train Epoch: 5 [66560/74759 (89%)]\tLoss: 56.64\n",
      "[2m 16s] Train Epoch: 5 [69120/74759 (92%)]\tLoss: 56.74\n",
      "[2m 17s] Train Epoch: 5 [71680/74759 (96%)]\tLoss: 56.67\n",
      "[2m 18s] Train Epoch: 5 [74240/74759 (99%)]\tLoss: 56.64\n",
      "\n",
      "Train set: Accuracy: 67066/74759 (89%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9658269559499145\n",
      "\n",
      "Test set: Accuracy: 16602/18691 (88%) Recall: (24%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[2m 22s] Train Epoch: 6 [2560/74759 (3%)]\tLoss: 59.04\n",
      "[2m 23s] Train Epoch: 6 [5120/74759 (7%)]\tLoss: 58.53\n",
      "[2m 24s] Train Epoch: 6 [7680/74759 (10%)]\tLoss: 57.90\n",
      "[2m 25s] Train Epoch: 6 [10240/74759 (14%)]\tLoss: 56.99\n",
      "[2m 26s] Train Epoch: 6 [12800/74759 (17%)]\tLoss: 55.92\n",
      "[2m 26s] Train Epoch: 6 [15360/74759 (21%)]\tLoss: 55.95\n",
      "[2m 27s] Train Epoch: 6 [17920/74759 (24%)]\tLoss: 55.69\n",
      "[2m 28s] Train Epoch: 6 [20480/74759 (27%)]\tLoss: 55.49\n",
      "[2m 29s] Train Epoch: 6 [23040/74759 (31%)]\tLoss: 55.06\n",
      "[2m 30s] Train Epoch: 6 [25600/74759 (34%)]\tLoss: 55.57\n",
      "[2m 31s] Train Epoch: 6 [28160/74759 (38%)]\tLoss: 55.55\n",
      "[2m 32s] Train Epoch: 6 [30720/74759 (41%)]\tLoss: 55.20\n",
      "[2m 33s] Train Epoch: 6 [33280/74759 (45%)]\tLoss: 54.90\n",
      "[2m 34s] Train Epoch: 6 [35840/74759 (48%)]\tLoss: 54.70\n",
      "[2m 35s] Train Epoch: 6 [38400/74759 (51%)]\tLoss: 54.80\n",
      "[2m 36s] Train Epoch: 6 [40960/74759 (55%)]\tLoss: 54.75\n",
      "[2m 36s] Train Epoch: 6 [43520/74759 (58%)]\tLoss: 54.94\n",
      "[2m 37s] Train Epoch: 6 [46080/74759 (62%)]\tLoss: 54.96\n",
      "[2m 38s] Train Epoch: 6 [48640/74759 (65%)]\tLoss: 54.91\n",
      "[2m 39s] Train Epoch: 6 [51200/74759 (68%)]\tLoss: 54.73\n",
      "[2m 40s] Train Epoch: 6 [53760/74759 (72%)]\tLoss: 54.79\n",
      "[2m 40s] Train Epoch: 6 [56320/74759 (75%)]\tLoss: 54.63\n",
      "[2m 41s] Train Epoch: 6 [58880/74759 (79%)]\tLoss: 54.75\n",
      "[2m 42s] Train Epoch: 6 [61440/74759 (82%)]\tLoss: 54.76\n",
      "[2m 43s] Train Epoch: 6 [64000/74759 (86%)]\tLoss: 54.56\n",
      "[2m 43s] Train Epoch: 6 [66560/74759 (89%)]\tLoss: 54.55\n",
      "[2m 44s] Train Epoch: 6 [69120/74759 (92%)]\tLoss: 54.59\n",
      "[2m 45s] Train Epoch: 6 [71680/74759 (96%)]\tLoss: 54.45\n",
      "[2m 46s] Train Epoch: 6 [74240/74759 (99%)]\tLoss: 54.40\n",
      "\n",
      "Train set: Accuracy: 67360/74759 (90%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9657517665269189\n",
      "\n",
      "Test set: Accuracy: 16688/18691 (89%) Recall: (22%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[2m 50s] Train Epoch: 7 [2560/74759 (3%)]\tLoss: 47.70\n",
      "[2m 51s] Train Epoch: 7 [5120/74759 (7%)]\tLoss: 49.35\n",
      "[2m 52s] Train Epoch: 7 [7680/74759 (10%)]\tLoss: 49.02\n",
      "[2m 53s] Train Epoch: 7 [10240/74759 (14%)]\tLoss: 48.93\n",
      "[2m 53s] Train Epoch: 7 [12800/74759 (17%)]\tLoss: 49.18\n",
      "[2m 54s] Train Epoch: 7 [15360/74759 (21%)]\tLoss: 49.43\n",
      "[2m 55s] Train Epoch: 7 [17920/74759 (24%)]\tLoss: 50.07\n",
      "[2m 56s] Train Epoch: 7 [20480/74759 (27%)]\tLoss: 49.96\n",
      "[2m 57s] Train Epoch: 7 [23040/74759 (31%)]\tLoss: 50.57\n",
      "[2m 58s] Train Epoch: 7 [25600/74759 (34%)]\tLoss: 50.67\n",
      "[2m 58s] Train Epoch: 7 [28160/74759 (38%)]\tLoss: 50.41\n",
      "[2m 59s] Train Epoch: 7 [30720/74759 (41%)]\tLoss: 50.39\n",
      "[3m 0s] Train Epoch: 7 [33280/74759 (45%)]\tLoss: 50.18\n",
      "[3m 1s] Train Epoch: 7 [35840/74759 (48%)]\tLoss: 50.10\n",
      "[3m 2s] Train Epoch: 7 [38400/74759 (51%)]\tLoss: 50.18\n",
      "[3m 3s] Train Epoch: 7 [40960/74759 (55%)]\tLoss: 50.12\n",
      "[3m 3s] Train Epoch: 7 [43520/74759 (58%)]\tLoss: 50.35\n",
      "[3m 4s] Train Epoch: 7 [46080/74759 (62%)]\tLoss: 50.42\n",
      "[3m 5s] Train Epoch: 7 [48640/74759 (65%)]\tLoss: 50.32\n",
      "[3m 6s] Train Epoch: 7 [51200/74759 (68%)]\tLoss: 50.37\n",
      "[3m 7s] Train Epoch: 7 [53760/74759 (72%)]\tLoss: 50.47\n",
      "[3m 7s] Train Epoch: 7 [56320/74759 (75%)]\tLoss: 50.39\n",
      "[3m 8s] Train Epoch: 7 [58880/74759 (79%)]\tLoss: 50.55\n",
      "[3m 9s] Train Epoch: 7 [61440/74759 (82%)]\tLoss: 50.66\n",
      "[3m 10s] Train Epoch: 7 [64000/74759 (86%)]\tLoss: 50.69\n",
      "[3m 11s] Train Epoch: 7 [66560/74759 (89%)]\tLoss: 50.78\n",
      "[3m 12s] Train Epoch: 7 [69120/74759 (92%)]\tLoss: 50.71\n",
      "[3m 13s] Train Epoch: 7 [71680/74759 (96%)]\tLoss: 50.49\n",
      "[3m 14s] Train Epoch: 7 [74240/74759 (99%)]\tLoss: 50.36\n",
      "\n",
      "Train set: Accuracy: 67861/74759 (90%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9685239158794737\n",
      "\n",
      "Test set: Accuracy: 16660/18691 (89%) Recall: (24%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[3m 18s] Train Epoch: 8 [2560/74759 (3%)]\tLoss: 48.35\n",
      "[3m 19s] Train Epoch: 8 [5120/74759 (7%)]\tLoss: 48.25\n",
      "[3m 19s] Train Epoch: 8 [7680/74759 (10%)]\tLoss: 48.46\n",
      "[3m 20s] Train Epoch: 8 [10240/74759 (14%)]\tLoss: 49.84\n",
      "[3m 21s] Train Epoch: 8 [12800/74759 (17%)]\tLoss: 49.72\n",
      "[3m 22s] Train Epoch: 8 [15360/74759 (21%)]\tLoss: 49.57\n",
      "[3m 23s] Train Epoch: 8 [17920/74759 (24%)]\tLoss: 49.60\n",
      "[3m 23s] Train Epoch: 8 [20480/74759 (27%)]\tLoss: 49.51\n",
      "[3m 24s] Train Epoch: 8 [23040/74759 (31%)]\tLoss: 48.83\n",
      "[3m 25s] Train Epoch: 8 [25600/74759 (34%)]\tLoss: 48.85\n",
      "[3m 26s] Train Epoch: 8 [28160/74759 (38%)]\tLoss: 48.43\n",
      "[3m 27s] Train Epoch: 8 [30720/74759 (41%)]\tLoss: 48.14\n",
      "[3m 27s] Train Epoch: 8 [33280/74759 (45%)]\tLoss: 48.38\n",
      "[3m 28s] Train Epoch: 8 [35840/74759 (48%)]\tLoss: 48.50\n",
      "[3m 29s] Train Epoch: 8 [38400/74759 (51%)]\tLoss: 48.88\n",
      "[3m 30s] Train Epoch: 8 [40960/74759 (55%)]\tLoss: 49.00\n",
      "[3m 31s] Train Epoch: 8 [43520/74759 (58%)]\tLoss: 48.98\n",
      "[3m 32s] Train Epoch: 8 [46080/74759 (62%)]\tLoss: 49.16\n",
      "[3m 33s] Train Epoch: 8 [48640/74759 (65%)]\tLoss: 49.12\n",
      "[3m 34s] Train Epoch: 8 [51200/74759 (68%)]\tLoss: 48.95\n",
      "[3m 35s] Train Epoch: 8 [53760/74759 (72%)]\tLoss: 48.85\n",
      "[3m 36s] Train Epoch: 8 [56320/74759 (75%)]\tLoss: 48.92\n",
      "[3m 37s] Train Epoch: 8 [58880/74759 (79%)]\tLoss: 48.81\n",
      "[3m 38s] Train Epoch: 8 [61440/74759 (82%)]\tLoss: 48.79\n",
      "[3m 38s] Train Epoch: 8 [64000/74759 (86%)]\tLoss: 48.81\n",
      "[3m 39s] Train Epoch: 8 [66560/74759 (89%)]\tLoss: 48.93\n",
      "[3m 40s] Train Epoch: 8 [69120/74759 (92%)]\tLoss: 48.96\n",
      "[3m 41s] Train Epoch: 8 [71680/74759 (96%)]\tLoss: 48.86\n",
      "[3m 42s] Train Epoch: 8 [74240/74759 (99%)]\tLoss: 48.98\n",
      "\n",
      "Train set: Accuracy: 68048/74759 (91%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9683306170302345\n",
      "\n",
      "Test set: Accuracy: 16729/18691 (89%) Recall: (22%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[3m 46s] Train Epoch: 9 [2560/74759 (3%)]\tLoss: 44.96\n",
      "[3m 47s] Train Epoch: 9 [5120/74759 (7%)]\tLoss: 46.47\n",
      "[3m 48s] Train Epoch: 9 [7680/74759 (10%)]\tLoss: 45.57\n",
      "[3m 49s] Train Epoch: 9 [10240/74759 (14%)]\tLoss: 44.69\n",
      "[3m 50s] Train Epoch: 9 [12800/74759 (17%)]\tLoss: 44.40\n",
      "[3m 51s] Train Epoch: 9 [15360/74759 (21%)]\tLoss: 44.46\n",
      "[3m 52s] Train Epoch: 9 [17920/74759 (24%)]\tLoss: 44.74\n",
      "[3m 53s] Train Epoch: 9 [20480/74759 (27%)]\tLoss: 44.60\n",
      "[3m 53s] Train Epoch: 9 [23040/74759 (31%)]\tLoss: 45.01\n",
      "[3m 54s] Train Epoch: 9 [25600/74759 (34%)]\tLoss: 44.45\n",
      "[3m 55s] Train Epoch: 9 [28160/74759 (38%)]\tLoss: 44.92\n",
      "[3m 56s] Train Epoch: 9 [30720/74759 (41%)]\tLoss: 45.12\n",
      "[3m 57s] Train Epoch: 9 [33280/74759 (45%)]\tLoss: 45.30\n",
      "[3m 58s] Train Epoch: 9 [35840/74759 (48%)]\tLoss: 45.43\n",
      "[3m 59s] Train Epoch: 9 [38400/74759 (51%)]\tLoss: 45.51\n",
      "[4m 0s] Train Epoch: 9 [40960/74759 (55%)]\tLoss: 45.60\n",
      "[4m 1s] Train Epoch: 9 [43520/74759 (58%)]\tLoss: 45.53\n",
      "[4m 1s] Train Epoch: 9 [46080/74759 (62%)]\tLoss: 45.55\n",
      "[4m 2s] Train Epoch: 9 [48640/74759 (65%)]\tLoss: 45.62\n",
      "[4m 3s] Train Epoch: 9 [51200/74759 (68%)]\tLoss: 45.67\n",
      "[4m 4s] Train Epoch: 9 [53760/74759 (72%)]\tLoss: 45.76\n",
      "[4m 5s] Train Epoch: 9 [56320/74759 (75%)]\tLoss: 45.81\n",
      "[4m 6s] Train Epoch: 9 [58880/74759 (79%)]\tLoss: 45.97\n",
      "[4m 7s] Train Epoch: 9 [61440/74759 (82%)]\tLoss: 46.00\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[4m 8s] Train Epoch: 9 [64000/74759 (86%)]\tLoss: 46.09\n",
      "[4m 8s] Train Epoch: 9 [66560/74759 (89%)]\tLoss: 46.05\n",
      "[4m 9s] Train Epoch: 9 [69120/74759 (92%)]\tLoss: 45.98\n",
      "[4m 10s] Train Epoch: 9 [71680/74759 (96%)]\tLoss: 45.90\n",
      "[4m 11s] Train Epoch: 9 [74240/74759 (99%)]\tLoss: 45.89\n",
      "\n",
      "Train set: Accuracy: 68432/74759 (91%) \n",
      "evaluating trained model ...\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/data0/zhengsj/anaconda3/lib/python3.6/site-packages/sklearn/metrics/classification.py:1137: UndefinedMetricWarning: Recall is ill-defined and being set to 0.0 due to no true samples.\n",
      "  'recall', 'true', average, warn_for)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUC =  0.9685861732761912\n",
      "\n",
      "Test set: Accuracy: 16702/18691 (89%) Recall: (22%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[4m 15s] Train Epoch: 10 [2560/74759 (3%)]\tLoss: 53.97\n",
      "[4m 16s] Train Epoch: 10 [5120/74759 (7%)]\tLoss: 50.25\n",
      "[4m 17s] Train Epoch: 10 [7680/74759 (10%)]\tLoss: 49.44\n",
      "[4m 17s] Train Epoch: 10 [10240/74759 (14%)]\tLoss: 48.03\n",
      "[4m 18s] Train Epoch: 10 [12800/74759 (17%)]\tLoss: 47.24\n",
      "[4m 19s] Train Epoch: 10 [15360/74759 (21%)]\tLoss: 47.43\n",
      "[4m 20s] Train Epoch: 10 [17920/74759 (24%)]\tLoss: 47.18\n",
      "[4m 21s] Train Epoch: 10 [20480/74759 (27%)]\tLoss: 46.85\n",
      "[4m 22s] Train Epoch: 10 [23040/74759 (31%)]\tLoss: 46.54\n",
      "[4m 23s] Train Epoch: 10 [25600/74759 (34%)]\tLoss: 46.51\n",
      "[4m 24s] Train Epoch: 10 [28160/74759 (38%)]\tLoss: 46.11\n",
      "[4m 24s] Train Epoch: 10 [30720/74759 (41%)]\tLoss: 45.93\n",
      "[4m 25s] Train Epoch: 10 [33280/74759 (45%)]\tLoss: 45.78\n",
      "[4m 26s] Train Epoch: 10 [35840/74759 (48%)]\tLoss: 45.47\n",
      "[4m 27s] Train Epoch: 10 [38400/74759 (51%)]\tLoss: 45.63\n",
      "[4m 28s] Train Epoch: 10 [40960/74759 (55%)]\tLoss: 45.61\n",
      "[4m 29s] Train Epoch: 10 [43520/74759 (58%)]\tLoss: 45.40\n",
      "[4m 29s] Train Epoch: 10 [46080/74759 (62%)]\tLoss: 45.00\n",
      "[4m 30s] Train Epoch: 10 [48640/74759 (65%)]\tLoss: 45.12\n",
      "[4m 31s] Train Epoch: 10 [51200/74759 (68%)]\tLoss: 45.20\n",
      "[4m 32s] Train Epoch: 10 [53760/74759 (72%)]\tLoss: 45.25\n",
      "[4m 32s] Train Epoch: 10 [56320/74759 (75%)]\tLoss: 45.29\n",
      "[4m 33s] Train Epoch: 10 [58880/74759 (79%)]\tLoss: 45.31\n",
      "[4m 34s] Train Epoch: 10 [61440/74759 (82%)]\tLoss: 45.27\n",
      "[4m 35s] Train Epoch: 10 [64000/74759 (86%)]\tLoss: 45.25\n",
      "[4m 36s] Train Epoch: 10 [66560/74759 (89%)]\tLoss: 45.18\n",
      "[4m 37s] Train Epoch: 10 [69120/74759 (92%)]\tLoss: 45.03\n",
      "[4m 37s] Train Epoch: 10 [71680/74759 (96%)]\tLoss: 45.02\n",
      "[4m 38s] Train Epoch: 10 [74240/74759 (99%)]\tLoss: 45.15\n",
      "\n",
      "Train set: Accuracy: 68572/74759 (91%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9710976547958058\n",
      "\n",
      "Test set: Accuracy: 16848/18691 (90%) Recall: (24%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[4m 43s] Train Epoch: 11 [2560/74759 (3%)]\tLoss: 39.76\n",
      "[4m 44s] Train Epoch: 11 [5120/74759 (7%)]\tLoss: 39.15\n",
      "[4m 45s] Train Epoch: 11 [7680/74759 (10%)]\tLoss: 39.90\n",
      "[4m 46s] Train Epoch: 11 [10240/74759 (14%)]\tLoss: 39.95\n",
      "[4m 46s] Train Epoch: 11 [12800/74759 (17%)]\tLoss: 40.37\n",
      "[4m 47s] Train Epoch: 11 [15360/74759 (21%)]\tLoss: 40.59\n",
      "[4m 48s] Train Epoch: 11 [17920/74759 (24%)]\tLoss: 40.41\n",
      "[4m 49s] Train Epoch: 11 [20480/74759 (27%)]\tLoss: 40.52\n",
      "[4m 50s] Train Epoch: 11 [23040/74759 (31%)]\tLoss: 40.48\n",
      "[4m 51s] Train Epoch: 11 [25600/74759 (34%)]\tLoss: 40.82\n",
      "[4m 51s] Train Epoch: 11 [28160/74759 (38%)]\tLoss: 41.04\n",
      "[4m 52s] Train Epoch: 11 [30720/74759 (41%)]\tLoss: 41.11\n",
      "[4m 53s] Train Epoch: 11 [33280/74759 (45%)]\tLoss: 41.28\n",
      "[4m 54s] Train Epoch: 11 [35840/74759 (48%)]\tLoss: 41.26\n",
      "[4m 55s] Train Epoch: 11 [38400/74759 (51%)]\tLoss: 41.55\n",
      "[4m 56s] Train Epoch: 11 [40960/74759 (55%)]\tLoss: 41.52\n",
      "[4m 57s] Train Epoch: 11 [43520/74759 (58%)]\tLoss: 41.70\n",
      "[4m 58s] Train Epoch: 11 [46080/74759 (62%)]\tLoss: 41.90\n",
      "[4m 59s] Train Epoch: 11 [48640/74759 (65%)]\tLoss: 41.90\n",
      "[4m 59s] Train Epoch: 11 [51200/74759 (68%)]\tLoss: 41.72\n",
      "[5m 0s] Train Epoch: 11 [53760/74759 (72%)]\tLoss: 41.88\n",
      "[5m 1s] Train Epoch: 11 [56320/74759 (75%)]\tLoss: 41.98\n",
      "[5m 2s] Train Epoch: 11 [58880/74759 (79%)]\tLoss: 41.88\n",
      "[5m 3s] Train Epoch: 11 [61440/74759 (82%)]\tLoss: 41.96\n",
      "[5m 4s] Train Epoch: 11 [64000/74759 (86%)]\tLoss: 42.16\n",
      "[5m 5s] Train Epoch: 11 [66560/74759 (89%)]\tLoss: 42.25\n",
      "[5m 5s] Train Epoch: 11 [69120/74759 (92%)]\tLoss: 42.39\n",
      "[5m 6s] Train Epoch: 11 [71680/74759 (96%)]\tLoss: 42.33\n",
      "[5m 7s] Train Epoch: 11 [74240/74759 (99%)]\tLoss: 42.43\n",
      "\n",
      "Train set: Accuracy: 68924/74759 (92%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.971222215795677\n",
      "\n",
      "Test set: Accuracy: 16861/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[5m 11s] Train Epoch: 12 [2560/74759 (3%)]\tLoss: 40.95\n",
      "[5m 12s] Train Epoch: 12 [5120/74759 (7%)]\tLoss: 38.42\n",
      "[5m 13s] Train Epoch: 12 [7680/74759 (10%)]\tLoss: 39.45\n",
      "[5m 13s] Train Epoch: 12 [10240/74759 (14%)]\tLoss: 38.95\n",
      "[5m 14s] Train Epoch: 12 [12800/74759 (17%)]\tLoss: 39.00\n",
      "[5m 15s] Train Epoch: 12 [15360/74759 (21%)]\tLoss: 39.05\n",
      "[5m 16s] Train Epoch: 12 [17920/74759 (24%)]\tLoss: 40.19\n",
      "[5m 17s] Train Epoch: 12 [20480/74759 (27%)]\tLoss: 40.71\n",
      "[5m 17s] Train Epoch: 12 [23040/74759 (31%)]\tLoss: 40.39\n",
      "[5m 18s] Train Epoch: 12 [25600/74759 (34%)]\tLoss: 40.30\n",
      "[5m 19s] Train Epoch: 12 [28160/74759 (38%)]\tLoss: 40.12\n",
      "[5m 20s] Train Epoch: 12 [30720/74759 (41%)]\tLoss: 40.00\n",
      "[5m 21s] Train Epoch: 12 [33280/74759 (45%)]\tLoss: 40.25\n",
      "[5m 22s] Train Epoch: 12 [35840/74759 (48%)]\tLoss: 40.33\n",
      "[5m 23s] Train Epoch: 12 [38400/74759 (51%)]\tLoss: 40.37\n",
      "[5m 23s] Train Epoch: 12 [40960/74759 (55%)]\tLoss: 40.28\n",
      "[5m 24s] Train Epoch: 12 [43520/74759 (58%)]\tLoss: 40.34\n",
      "[5m 25s] Train Epoch: 12 [46080/74759 (62%)]\tLoss: 40.09\n",
      "[5m 26s] Train Epoch: 12 [48640/74759 (65%)]\tLoss: 40.11\n",
      "[5m 26s] Train Epoch: 12 [51200/74759 (68%)]\tLoss: 40.16\n",
      "[5m 27s] Train Epoch: 12 [53760/74759 (72%)]\tLoss: 40.18\n",
      "[5m 28s] Train Epoch: 12 [56320/74759 (75%)]\tLoss: 40.28\n",
      "[5m 29s] Train Epoch: 12 [58880/74759 (79%)]\tLoss: 40.39\n",
      "[5m 29s] Train Epoch: 12 [61440/74759 (82%)]\tLoss: 40.40\n",
      "[5m 30s] Train Epoch: 12 [64000/74759 (86%)]\tLoss: 40.41\n",
      "[5m 31s] Train Epoch: 12 [66560/74759 (89%)]\tLoss: 40.47\n",
      "[5m 32s] Train Epoch: 12 [69120/74759 (92%)]\tLoss: 40.73\n",
      "[5m 32s] Train Epoch: 12 [71680/74759 (96%)]\tLoss: 40.69\n",
      "[5m 33s] Train Epoch: 12 [74240/74759 (99%)]\tLoss: 40.70\n",
      "\n",
      "Train set: Accuracy: 69128/74759 (92%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.968636307259261\n",
      "\n",
      "Test set: Accuracy: 16800/18691 (89%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[5m 37s] Train Epoch: 13 [2560/74759 (3%)]\tLoss: 65.58\n",
      "[5m 38s] Train Epoch: 13 [5120/74759 (7%)]\tLoss: 61.23\n",
      "[5m 39s] Train Epoch: 13 [7680/74759 (10%)]\tLoss: 57.82\n",
      "[5m 40s] Train Epoch: 13 [10240/74759 (14%)]\tLoss: 55.56\n",
      "[5m 41s] Train Epoch: 13 [12800/74759 (17%)]\tLoss: 53.83\n",
      "[5m 42s] Train Epoch: 13 [15360/74759 (21%)]\tLoss: 51.59\n",
      "[5m 43s] Train Epoch: 13 [17920/74759 (24%)]\tLoss: 50.05\n",
      "[5m 44s] Train Epoch: 13 [20480/74759 (27%)]\tLoss: 49.08\n",
      "[5m 45s] Train Epoch: 13 [23040/74759 (31%)]\tLoss: 48.13\n",
      "[5m 45s] Train Epoch: 13 [25600/74759 (34%)]\tLoss: 47.09\n",
      "[5m 46s] Train Epoch: 13 [28160/74759 (38%)]\tLoss: 46.53\n",
      "[5m 47s] Train Epoch: 13 [30720/74759 (41%)]\tLoss: 46.01\n",
      "[5m 48s] Train Epoch: 13 [33280/74759 (45%)]\tLoss: 45.62\n",
      "[5m 49s] Train Epoch: 13 [35840/74759 (48%)]\tLoss: 45.06\n",
      "[5m 50s] Train Epoch: 13 [38400/74759 (51%)]\tLoss: 44.97\n",
      "[5m 51s] Train Epoch: 13 [40960/74759 (55%)]\tLoss: 44.70\n",
      "[5m 52s] Train Epoch: 13 [43520/74759 (58%)]\tLoss: 44.63\n",
      "[5m 53s] Train Epoch: 13 [46080/74759 (62%)]\tLoss: 44.45\n",
      "[5m 53s] Train Epoch: 13 [48640/74759 (65%)]\tLoss: 44.22\n",
      "[5m 54s] Train Epoch: 13 [51200/74759 (68%)]\tLoss: 43.99\n",
      "[5m 55s] Train Epoch: 13 [53760/74759 (72%)]\tLoss: 43.89\n",
      "[5m 56s] Train Epoch: 13 [56320/74759 (75%)]\tLoss: 43.74\n",
      "[5m 56s] Train Epoch: 13 [58880/74759 (79%)]\tLoss: 43.72\n",
      "[5m 57s] Train Epoch: 13 [61440/74759 (82%)]\tLoss: 43.58\n",
      "[5m 58s] Train Epoch: 13 [64000/74759 (86%)]\tLoss: 43.37\n",
      "[5m 59s] Train Epoch: 13 [66560/74759 (89%)]\tLoss: 43.19\n",
      "[6m 0s] Train Epoch: 13 [69120/74759 (92%)]\tLoss: 43.09\n",
      "[6m 1s] Train Epoch: 13 [71680/74759 (96%)]\tLoss: 42.94\n",
      "[6m 2s] Train Epoch: 13 [74240/74759 (99%)]\tLoss: 42.80\n",
      "\n",
      "Train set: Accuracy: 68909/74759 (92%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9714383348485002\n",
      "\n",
      "Test set: Accuracy: 16859/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[6m 6s] Train Epoch: 14 [2560/74759 (3%)]\tLoss: 38.01\n",
      "[6m 7s] Train Epoch: 14 [5120/74759 (7%)]\tLoss: 39.46\n",
      "[6m 7s] Train Epoch: 14 [7680/74759 (10%)]\tLoss: 39.03\n",
      "[6m 8s] Train Epoch: 14 [10240/74759 (14%)]\tLoss: 38.36\n",
      "[6m 10s] Train Epoch: 14 [12800/74759 (17%)]\tLoss: 38.10\n",
      "[6m 10s] Train Epoch: 14 [15360/74759 (21%)]\tLoss: 38.25\n",
      "[6m 11s] Train Epoch: 14 [17920/74759 (24%)]\tLoss: 38.14\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[6m 12s] Train Epoch: 14 [20480/74759 (27%)]\tLoss: 37.94\n",
      "[6m 13s] Train Epoch: 14 [23040/74759 (31%)]\tLoss: 37.89\n",
      "[6m 14s] Train Epoch: 14 [25600/74759 (34%)]\tLoss: 38.03\n",
      "[6m 14s] Train Epoch: 14 [28160/74759 (38%)]\tLoss: 37.60\n",
      "[6m 15s] Train Epoch: 14 [30720/74759 (41%)]\tLoss: 37.59\n",
      "[6m 16s] Train Epoch: 14 [33280/74759 (45%)]\tLoss: 37.57\n",
      "[6m 17s] Train Epoch: 14 [35840/74759 (48%)]\tLoss: 37.58\n",
      "[6m 18s] Train Epoch: 14 [38400/74759 (51%)]\tLoss: 37.66\n",
      "[6m 19s] Train Epoch: 14 [40960/74759 (55%)]\tLoss: 37.82\n",
      "[6m 19s] Train Epoch: 14 [43520/74759 (58%)]\tLoss: 37.97\n",
      "[6m 20s] Train Epoch: 14 [46080/74759 (62%)]\tLoss: 38.07\n",
      "[6m 21s] Train Epoch: 14 [48640/74759 (65%)]\tLoss: 38.01\n",
      "[6m 22s] Train Epoch: 14 [51200/74759 (68%)]\tLoss: 38.23\n",
      "[6m 23s] Train Epoch: 14 [53760/74759 (72%)]\tLoss: 38.15\n",
      "[6m 24s] Train Epoch: 14 [56320/74759 (75%)]\tLoss: 38.16\n",
      "[6m 25s] Train Epoch: 14 [58880/74759 (79%)]\tLoss: 38.23\n",
      "[6m 25s] Train Epoch: 14 [61440/74759 (82%)]\tLoss: 38.35\n",
      "[6m 26s] Train Epoch: 14 [64000/74759 (86%)]\tLoss: 38.32\n",
      "[6m 27s] Train Epoch: 14 [66560/74759 (89%)]\tLoss: 38.22\n",
      "[6m 28s] Train Epoch: 14 [69120/74759 (92%)]\tLoss: 38.21\n",
      "[6m 29s] Train Epoch: 14 [71680/74759 (96%)]\tLoss: 38.21\n",
      "[6m 30s] Train Epoch: 14 [74240/74759 (99%)]\tLoss: 38.21\n",
      "\n",
      "Train set: Accuracy: 69491/74759 (92%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9714492973254434\n",
      "\n",
      "Test set: Accuracy: 16873/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[6m 36s] Train Epoch: 15 [2560/74759 (3%)]\tLoss: 32.43\n",
      "[6m 36s] Train Epoch: 15 [5120/74759 (7%)]\tLoss: 34.94\n",
      "[6m 37s] Train Epoch: 15 [7680/74759 (10%)]\tLoss: 34.11\n",
      "[6m 38s] Train Epoch: 15 [10240/74759 (14%)]\tLoss: 34.87\n",
      "[6m 39s] Train Epoch: 15 [12800/74759 (17%)]\tLoss: 35.46\n",
      "[6m 40s] Train Epoch: 15 [15360/74759 (21%)]\tLoss: 35.61\n",
      "[6m 41s] Train Epoch: 15 [17920/74759 (24%)]\tLoss: 36.00\n",
      "[6m 41s] Train Epoch: 15 [20480/74759 (27%)]\tLoss: 35.62\n",
      "[6m 42s] Train Epoch: 15 [23040/74759 (31%)]\tLoss: 35.75\n",
      "[6m 43s] Train Epoch: 15 [25600/74759 (34%)]\tLoss: 36.18\n",
      "[6m 44s] Train Epoch: 15 [28160/74759 (38%)]\tLoss: 35.98\n",
      "[6m 45s] Train Epoch: 15 [30720/74759 (41%)]\tLoss: 36.33\n",
      "[6m 46s] Train Epoch: 15 [33280/74759 (45%)]\tLoss: 36.58\n",
      "[6m 47s] Train Epoch: 15 [35840/74759 (48%)]\tLoss: 36.36\n",
      "[6m 48s] Train Epoch: 15 [38400/74759 (51%)]\tLoss: 36.48\n",
      "[6m 49s] Train Epoch: 15 [40960/74759 (55%)]\tLoss: 36.57\n",
      "[6m 50s] Train Epoch: 15 [43520/74759 (58%)]\tLoss: 36.53\n",
      "[6m 51s] Train Epoch: 15 [46080/74759 (62%)]\tLoss: 36.46\n",
      "[6m 52s] Train Epoch: 15 [48640/74759 (65%)]\tLoss: 36.52\n",
      "[6m 52s] Train Epoch: 15 [51200/74759 (68%)]\tLoss: 36.36\n",
      "[6m 53s] Train Epoch: 15 [53760/74759 (72%)]\tLoss: 36.43\n",
      "[6m 54s] Train Epoch: 15 [56320/74759 (75%)]\tLoss: 36.38\n",
      "[6m 55s] Train Epoch: 15 [58880/74759 (79%)]\tLoss: 36.59\n",
      "[6m 55s] Train Epoch: 15 [61440/74759 (82%)]\tLoss: 36.59\n",
      "[6m 56s] Train Epoch: 15 [64000/74759 (86%)]\tLoss: 36.70\n",
      "[6m 57s] Train Epoch: 15 [66560/74759 (89%)]\tLoss: 36.80\n",
      "[6m 58s] Train Epoch: 15 [69120/74759 (92%)]\tLoss: 36.78\n",
      "[6m 59s] Train Epoch: 15 [71680/74759 (96%)]\tLoss: 36.81\n",
      "[7m 0s] Train Epoch: 15 [74240/74759 (99%)]\tLoss: 36.75\n",
      "\n",
      "Train set: Accuracy: 69690/74759 (93%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9712159663752069\n",
      "\n",
      "Test set: Accuracy: 16837/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is NP\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[7m 4s] Train Epoch: 16 [2560/74759 (3%)]\tLoss: 37.65\n",
      "[7m 5s] Train Epoch: 16 [5120/74759 (7%)]\tLoss: 36.30\n",
      "[7m 6s] Train Epoch: 16 [7680/74759 (10%)]\tLoss: 36.59\n",
      "[7m 7s] Train Epoch: 16 [10240/74759 (14%)]\tLoss: 36.72\n",
      "[7m 8s] Train Epoch: 16 [12800/74759 (17%)]\tLoss: 36.34\n",
      "[7m 9s] Train Epoch: 16 [15360/74759 (21%)]\tLoss: 36.10\n",
      "[7m 10s] Train Epoch: 16 [17920/74759 (24%)]\tLoss: 35.93\n",
      "[7m 11s] Train Epoch: 16 [20480/74759 (27%)]\tLoss: 36.17\n",
      "[7m 11s] Train Epoch: 16 [23040/74759 (31%)]\tLoss: 36.06\n",
      "[7m 12s] Train Epoch: 16 [25600/74759 (34%)]\tLoss: 35.89\n",
      "[7m 13s] Train Epoch: 16 [28160/74759 (38%)]\tLoss: 36.35\n",
      "[7m 14s] Train Epoch: 16 [30720/74759 (41%)]\tLoss: 36.10\n",
      "[7m 15s] Train Epoch: 16 [33280/74759 (45%)]\tLoss: 36.02\n",
      "[7m 16s] Train Epoch: 16 [35840/74759 (48%)]\tLoss: 35.99\n",
      "[7m 16s] Train Epoch: 16 [38400/74759 (51%)]\tLoss: 35.97\n",
      "[7m 17s] Train Epoch: 16 [40960/74759 (55%)]\tLoss: 36.03\n",
      "[7m 18s] Train Epoch: 16 [43520/74759 (58%)]\tLoss: 35.99\n",
      "[7m 19s] Train Epoch: 16 [46080/74759 (62%)]\tLoss: 36.04\n",
      "[7m 20s] Train Epoch: 16 [48640/74759 (65%)]\tLoss: 35.97\n",
      "[7m 21s] Train Epoch: 16 [51200/74759 (68%)]\tLoss: 35.77\n",
      "[7m 22s] Train Epoch: 16 [53760/74759 (72%)]\tLoss: 35.87\n",
      "[7m 23s] Train Epoch: 16 [56320/74759 (75%)]\tLoss: 35.88\n",
      "[7m 23s] Train Epoch: 16 [58880/74759 (79%)]\tLoss: 35.86\n",
      "[7m 24s] Train Epoch: 16 [61440/74759 (82%)]\tLoss: 35.80\n",
      "[7m 25s] Train Epoch: 16 [64000/74759 (86%)]\tLoss: 35.90\n",
      "[7m 26s] Train Epoch: 16 [66560/74759 (89%)]\tLoss: 35.82\n",
      "[7m 27s] Train Epoch: 16 [69120/74759 (92%)]\tLoss: 35.92\n",
      "[7m 27s] Train Epoch: 16 [71680/74759 (96%)]\tLoss: 36.00\n",
      "[7m 28s] Train Epoch: 16 [74240/74759 (99%)]\tLoss: 36.03\n",
      "\n",
      "Train set: Accuracy: 69779/74759 (93%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9714708988342775\n",
      "\n",
      "Test set: Accuracy: 16855/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[7m 33s] Train Epoch: 17 [2560/74759 (3%)]\tLoss: 28.92\n",
      "[7m 34s] Train Epoch: 17 [5120/74759 (7%)]\tLoss: 30.70\n",
      "[7m 35s] Train Epoch: 17 [7680/74759 (10%)]\tLoss: 31.48\n",
      "[7m 35s] Train Epoch: 17 [10240/74759 (14%)]\tLoss: 32.81\n",
      "[7m 36s] Train Epoch: 17 [12800/74759 (17%)]\tLoss: 33.72\n",
      "[7m 37s] Train Epoch: 17 [15360/74759 (21%)]\tLoss: 34.08\n",
      "[7m 38s] Train Epoch: 17 [17920/74759 (24%)]\tLoss: 33.74\n",
      "[7m 39s] Train Epoch: 17 [20480/74759 (27%)]\tLoss: 33.76\n",
      "[7m 40s] Train Epoch: 17 [23040/74759 (31%)]\tLoss: 33.86\n",
      "[7m 41s] Train Epoch: 17 [25600/74759 (34%)]\tLoss: 33.77\n",
      "[7m 42s] Train Epoch: 17 [28160/74759 (38%)]\tLoss: 33.89\n",
      "[7m 43s] Train Epoch: 17 [30720/74759 (41%)]\tLoss: 34.43\n",
      "[7m 43s] Train Epoch: 17 [33280/74759 (45%)]\tLoss: 34.36\n",
      "[7m 44s] Train Epoch: 17 [35840/74759 (48%)]\tLoss: 34.49\n",
      "[7m 45s] Train Epoch: 17 [38400/74759 (51%)]\tLoss: 34.35\n",
      "[7m 46s] Train Epoch: 17 [40960/74759 (55%)]\tLoss: 34.34\n",
      "[7m 47s] Train Epoch: 17 [43520/74759 (58%)]\tLoss: 34.32\n",
      "[7m 48s] Train Epoch: 17 [46080/74759 (62%)]\tLoss: 34.46\n",
      "[7m 48s] Train Epoch: 17 [48640/74759 (65%)]\tLoss: 34.46\n",
      "[7m 49s] Train Epoch: 17 [51200/74759 (68%)]\tLoss: 34.57\n",
      "[7m 50s] Train Epoch: 17 [53760/74759 (72%)]\tLoss: 34.60\n",
      "[7m 51s] Train Epoch: 17 [56320/74759 (75%)]\tLoss: 34.66\n",
      "[7m 52s] Train Epoch: 17 [58880/74759 (79%)]\tLoss: 34.65\n",
      "[7m 52s] Train Epoch: 17 [61440/74759 (82%)]\tLoss: 34.75\n",
      "[7m 53s] Train Epoch: 17 [64000/74759 (86%)]\tLoss: 34.94\n",
      "[7m 54s] Train Epoch: 17 [66560/74759 (89%)]\tLoss: 35.08\n",
      "[7m 55s] Train Epoch: 17 [69120/74759 (92%)]\tLoss: 35.08\n",
      "[7m 56s] Train Epoch: 17 [71680/74759 (96%)]\tLoss: 35.09\n",
      "[7m 56s] Train Epoch: 17 [74240/74759 (99%)]\tLoss: 35.20\n",
      "\n",
      "Train set: Accuracy: 69848/74759 (93%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9730498709413812\n",
      "\n",
      "Test set: Accuracy: 16945/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[8m 1s] Train Epoch: 18 [2560/74759 (3%)]\tLoss: 32.21\n",
      "[8m 1s] Train Epoch: 18 [5120/74759 (7%)]\tLoss: 33.08\n",
      "[8m 2s] Train Epoch: 18 [7680/74759 (10%)]\tLoss: 33.46\n",
      "[8m 3s] Train Epoch: 18 [10240/74759 (14%)]\tLoss: 34.06\n",
      "[8m 4s] Train Epoch: 18 [12800/74759 (17%)]\tLoss: 33.57\n",
      "[8m 5s] Train Epoch: 18 [15360/74759 (21%)]\tLoss: 33.77\n",
      "[8m 5s] Train Epoch: 18 [17920/74759 (24%)]\tLoss: 33.89\n",
      "[8m 6s] Train Epoch: 18 [20480/74759 (27%)]\tLoss: 34.20\n",
      "[8m 7s] Train Epoch: 18 [23040/74759 (31%)]\tLoss: 34.17\n",
      "[8m 8s] Train Epoch: 18 [25600/74759 (34%)]\tLoss: 33.99\n",
      "[8m 9s] Train Epoch: 18 [28160/74759 (38%)]\tLoss: 34.06\n",
      "[8m 10s] Train Epoch: 18 [30720/74759 (41%)]\tLoss: 34.11\n",
      "[8m 11s] Train Epoch: 18 [33280/74759 (45%)]\tLoss: 34.22\n",
      "[8m 12s] Train Epoch: 18 [35840/74759 (48%)]\tLoss: 34.20\n",
      "[8m 12s] Train Epoch: 18 [38400/74759 (51%)]\tLoss: 34.42\n",
      "[8m 13s] Train Epoch: 18 [40960/74759 (55%)]\tLoss: 34.30\n",
      "[8m 14s] Train Epoch: 18 [43520/74759 (58%)]\tLoss: 34.46\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[8m 15s] Train Epoch: 18 [46080/74759 (62%)]\tLoss: 34.28\n",
      "[8m 16s] Train Epoch: 18 [48640/74759 (65%)]\tLoss: 34.36\n",
      "[8m 17s] Train Epoch: 18 [51200/74759 (68%)]\tLoss: 34.10\n",
      "[8m 18s] Train Epoch: 18 [53760/74759 (72%)]\tLoss: 34.30\n",
      "[8m 19s] Train Epoch: 18 [56320/74759 (75%)]\tLoss: 34.21\n",
      "[8m 20s] Train Epoch: 18 [58880/74759 (79%)]\tLoss: 34.25\n",
      "[8m 21s] Train Epoch: 18 [61440/74759 (82%)]\tLoss: 34.42\n",
      "[8m 21s] Train Epoch: 18 [64000/74759 (86%)]\tLoss: 34.42\n",
      "[8m 22s] Train Epoch: 18 [66560/74759 (89%)]\tLoss: 34.44\n",
      "[8m 23s] Train Epoch: 18 [69120/74759 (92%)]\tLoss: 34.52\n",
      "[8m 24s] Train Epoch: 18 [71680/74759 (96%)]\tLoss: 34.50\n",
      "[8m 25s] Train Epoch: 18 [74240/74759 (99%)]\tLoss: 34.51\n",
      "\n",
      "Train set: Accuracy: 69961/74759 (93%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9718996368023958\n",
      "\n",
      "Test set: Accuracy: 16856/18691 (90%) Recall: (22%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[8m 29s] Train Epoch: 19 [2560/74759 (3%)]\tLoss: 36.25\n",
      "[8m 30s] Train Epoch: 19 [5120/74759 (7%)]\tLoss: 36.47\n",
      "[8m 31s] Train Epoch: 19 [7680/74759 (10%)]\tLoss: 36.35\n",
      "[8m 32s] Train Epoch: 19 [10240/74759 (14%)]\tLoss: 37.22\n",
      "[8m 32s] Train Epoch: 19 [12800/74759 (17%)]\tLoss: 36.66\n",
      "[8m 33s] Train Epoch: 19 [15360/74759 (21%)]\tLoss: 36.03\n",
      "[8m 34s] Train Epoch: 19 [17920/74759 (24%)]\tLoss: 36.07\n",
      "[8m 35s] Train Epoch: 19 [20480/74759 (27%)]\tLoss: 35.89\n",
      "[8m 36s] Train Epoch: 19 [23040/74759 (31%)]\tLoss: 35.11\n",
      "[8m 36s] Train Epoch: 19 [25600/74759 (34%)]\tLoss: 34.79\n",
      "[8m 37s] Train Epoch: 19 [28160/74759 (38%)]\tLoss: 34.91\n",
      "[8m 38s] Train Epoch: 19 [30720/74759 (41%)]\tLoss: 35.10\n",
      "[8m 39s] Train Epoch: 19 [33280/74759 (45%)]\tLoss: 34.98\n",
      "[8m 40s] Train Epoch: 19 [35840/74759 (48%)]\tLoss: 35.24\n",
      "[8m 41s] Train Epoch: 19 [38400/74759 (51%)]\tLoss: 35.18\n",
      "[8m 42s] Train Epoch: 19 [40960/74759 (55%)]\tLoss: 34.92\n",
      "[8m 43s] Train Epoch: 19 [43520/74759 (58%)]\tLoss: 34.83\n",
      "[8m 43s] Train Epoch: 19 [46080/74759 (62%)]\tLoss: 34.68\n",
      "[8m 44s] Train Epoch: 19 [48640/74759 (65%)]\tLoss: 34.79\n",
      "[8m 45s] Train Epoch: 19 [51200/74759 (68%)]\tLoss: 34.81\n",
      "[8m 46s] Train Epoch: 19 [53760/74759 (72%)]\tLoss: 34.76\n",
      "[8m 47s] Train Epoch: 19 [56320/74759 (75%)]\tLoss: 34.78\n",
      "[8m 48s] Train Epoch: 19 [58880/74759 (79%)]\tLoss: 34.77\n",
      "[8m 48s] Train Epoch: 19 [61440/74759 (82%)]\tLoss: 34.73\n",
      "[8m 49s] Train Epoch: 19 [64000/74759 (86%)]\tLoss: 34.55\n",
      "[8m 50s] Train Epoch: 19 [66560/74759 (89%)]\tLoss: 34.60\n",
      "[8m 51s] Train Epoch: 19 [69120/74759 (92%)]\tLoss: 34.52\n",
      "[8m 52s] Train Epoch: 19 [71680/74759 (96%)]\tLoss: 34.59\n",
      "[8m 53s] Train Epoch: 19 [74240/74759 (99%)]\tLoss: 34.63\n",
      "\n",
      "Train set: Accuracy: 70013/74759 (93%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9727615947629257\n",
      "\n",
      "Test set: Accuracy: 16834/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[8m 57s] Train Epoch: 20 [2560/74759 (3%)]\tLoss: 33.01\n",
      "[8m 58s] Train Epoch: 20 [5120/74759 (7%)]\tLoss: 32.76\n",
      "[8m 59s] Train Epoch: 20 [7680/74759 (10%)]\tLoss: 32.05\n",
      "[9m 0s] Train Epoch: 20 [10240/74759 (14%)]\tLoss: 32.91\n",
      "[9m 1s] Train Epoch: 20 [12800/74759 (17%)]\tLoss: 32.68\n",
      "[9m 2s] Train Epoch: 20 [15360/74759 (21%)]\tLoss: 31.81\n",
      "[9m 2s] Train Epoch: 20 [17920/74759 (24%)]\tLoss: 31.63\n",
      "[9m 3s] Train Epoch: 20 [20480/74759 (27%)]\tLoss: 31.58\n",
      "[9m 4s] Train Epoch: 20 [23040/74759 (31%)]\tLoss: 31.29\n",
      "[9m 5s] Train Epoch: 20 [25600/74759 (34%)]\tLoss: 31.30\n",
      "[9m 6s] Train Epoch: 20 [28160/74759 (38%)]\tLoss: 30.97\n",
      "[9m 7s] Train Epoch: 20 [30720/74759 (41%)]\tLoss: 31.30\n",
      "[9m 7s] Train Epoch: 20 [33280/74759 (45%)]\tLoss: 31.64\n",
      "[9m 8s] Train Epoch: 20 [35840/74759 (48%)]\tLoss: 31.93\n",
      "[9m 9s] Train Epoch: 20 [38400/74759 (51%)]\tLoss: 32.03\n",
      "[9m 10s] Train Epoch: 20 [40960/74759 (55%)]\tLoss: 31.89\n",
      "[9m 10s] Train Epoch: 20 [43520/74759 (58%)]\tLoss: 32.00\n",
      "[9m 11s] Train Epoch: 20 [46080/74759 (62%)]\tLoss: 31.88\n",
      "[9m 12s] Train Epoch: 20 [48640/74759 (65%)]\tLoss: 31.97\n",
      "[9m 13s] Train Epoch: 20 [51200/74759 (68%)]\tLoss: 32.09\n",
      "[9m 14s] Train Epoch: 20 [53760/74759 (72%)]\tLoss: 32.04\n",
      "[9m 15s] Train Epoch: 20 [56320/74759 (75%)]\tLoss: 32.11\n",
      "[9m 15s] Train Epoch: 20 [58880/74759 (79%)]\tLoss: 32.09\n",
      "[9m 16s] Train Epoch: 20 [61440/74759 (82%)]\tLoss: 32.24\n",
      "[9m 17s] Train Epoch: 20 [64000/74759 (86%)]\tLoss: 32.22\n",
      "[9m 18s] Train Epoch: 20 [66560/74759 (89%)]\tLoss: 32.26\n",
      "[9m 19s] Train Epoch: 20 [69120/74759 (92%)]\tLoss: 32.42\n",
      "[9m 20s] Train Epoch: 20 [71680/74759 (96%)]\tLoss: 32.52\n",
      "[9m 21s] Train Epoch: 20 [74240/74759 (99%)]\tLoss: 32.52\n",
      "\n",
      "Train set: Accuracy: 70202/74759 (93%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9728273003149304\n",
      "\n",
      "Test set: Accuracy: 16899/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is NP\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[9m 25s] Train Epoch: 21 [2560/74759 (3%)]\tLoss: 34.47\n",
      "[9m 26s] Train Epoch: 21 [5120/74759 (7%)]\tLoss: 36.26\n",
      "[9m 27s] Train Epoch: 21 [7680/74759 (10%)]\tLoss: 35.85\n",
      "[9m 27s] Train Epoch: 21 [10240/74759 (14%)]\tLoss: 34.78\n",
      "[9m 28s] Train Epoch: 21 [12800/74759 (17%)]\tLoss: 35.02\n",
      "[9m 29s] Train Epoch: 21 [15360/74759 (21%)]\tLoss: 34.39\n",
      "[9m 30s] Train Epoch: 21 [17920/74759 (24%)]\tLoss: 33.98\n",
      "[9m 30s] Train Epoch: 21 [20480/74759 (27%)]\tLoss: 33.76\n",
      "[9m 31s] Train Epoch: 21 [23040/74759 (31%)]\tLoss: 33.34\n",
      "[9m 32s] Train Epoch: 21 [25600/74759 (34%)]\tLoss: 33.37\n",
      "[9m 33s] Train Epoch: 21 [28160/74759 (38%)]\tLoss: 32.95\n",
      "[9m 34s] Train Epoch: 21 [30720/74759 (41%)]\tLoss: 32.82\n",
      "[9m 35s] Train Epoch: 21 [33280/74759 (45%)]\tLoss: 32.71\n",
      "[9m 36s] Train Epoch: 21 [35840/74759 (48%)]\tLoss: 32.91\n",
      "[9m 37s] Train Epoch: 21 [38400/74759 (51%)]\tLoss: 33.04\n",
      "[9m 38s] Train Epoch: 21 [40960/74759 (55%)]\tLoss: 32.79\n",
      "[9m 38s] Train Epoch: 21 [43520/74759 (58%)]\tLoss: 32.71\n",
      "[9m 39s] Train Epoch: 21 [46080/74759 (62%)]\tLoss: 32.62\n",
      "[9m 40s] Train Epoch: 21 [48640/74759 (65%)]\tLoss: 32.85\n",
      "[9m 41s] Train Epoch: 21 [51200/74759 (68%)]\tLoss: 32.87\n",
      "[9m 42s] Train Epoch: 21 [53760/74759 (72%)]\tLoss: 32.90\n",
      "[9m 43s] Train Epoch: 21 [56320/74759 (75%)]\tLoss: 32.68\n",
      "[9m 44s] Train Epoch: 21 [58880/74759 (79%)]\tLoss: 32.77\n",
      "[9m 45s] Train Epoch: 21 [61440/74759 (82%)]\tLoss: 32.68\n",
      "[9m 45s] Train Epoch: 21 [64000/74759 (86%)]\tLoss: 32.60\n",
      "[9m 46s] Train Epoch: 21 [66560/74759 (89%)]\tLoss: 32.53\n",
      "[9m 47s] Train Epoch: 21 [69120/74759 (92%)]\tLoss: 32.50\n",
      "[9m 48s] Train Epoch: 21 [71680/74759 (96%)]\tLoss: 32.56\n",
      "[9m 48s] Train Epoch: 21 [74240/74759 (99%)]\tLoss: 32.66\n",
      "\n",
      "Train set: Accuracy: 70188/74759 (93%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9723850989465198\n",
      "\n",
      "Test set: Accuracy: 16848/18691 (90%) Recall: (24%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[9m 52s] Train Epoch: 22 [2560/74759 (3%)]\tLoss: 31.50\n",
      "[9m 53s] Train Epoch: 22 [5120/74759 (7%)]\tLoss: 32.30\n",
      "[9m 54s] Train Epoch: 22 [7680/74759 (10%)]\tLoss: 31.48\n",
      "[9m 55s] Train Epoch: 22 [10240/74759 (14%)]\tLoss: 31.23\n",
      "[9m 56s] Train Epoch: 22 [12800/74759 (17%)]\tLoss: 31.20\n",
      "[9m 56s] Train Epoch: 22 [15360/74759 (21%)]\tLoss: 31.45\n",
      "[9m 57s] Train Epoch: 22 [17920/74759 (24%)]\tLoss: 31.97\n",
      "[9m 58s] Train Epoch: 22 [20480/74759 (27%)]\tLoss: 31.78\n",
      "[9m 59s] Train Epoch: 22 [23040/74759 (31%)]\tLoss: 31.77\n",
      "[10m 0s] Train Epoch: 22 [25600/74759 (34%)]\tLoss: 31.43\n",
      "[10m 0s] Train Epoch: 22 [28160/74759 (38%)]\tLoss: 31.67\n",
      "[10m 1s] Train Epoch: 22 [30720/74759 (41%)]\tLoss: 31.62\n",
      "[10m 2s] Train Epoch: 22 [33280/74759 (45%)]\tLoss: 31.54\n",
      "[10m 3s] Train Epoch: 22 [35840/74759 (48%)]\tLoss: 31.60\n",
      "[10m 4s] Train Epoch: 22 [38400/74759 (51%)]\tLoss: 31.71\n",
      "[10m 5s] Train Epoch: 22 [40960/74759 (55%)]\tLoss: 31.66\n",
      "[10m 6s] Train Epoch: 22 [43520/74759 (58%)]\tLoss: 31.56\n",
      "[10m 6s] Train Epoch: 22 [46080/74759 (62%)]\tLoss: 31.55\n",
      "[10m 7s] Train Epoch: 22 [48640/74759 (65%)]\tLoss: 31.69\n",
      "[10m 8s] Train Epoch: 22 [51200/74759 (68%)]\tLoss: 31.68\n",
      "[10m 9s] Train Epoch: 22 [53760/74759 (72%)]\tLoss: 31.60\n",
      "[10m 10s] Train Epoch: 22 [56320/74759 (75%)]\tLoss: 31.75\n",
      "[10m 10s] Train Epoch: 22 [58880/74759 (79%)]\tLoss: 31.87\n",
      "[10m 11s] Train Epoch: 22 [61440/74759 (82%)]\tLoss: 31.77\n",
      "[10m 12s] Train Epoch: 22 [64000/74759 (86%)]\tLoss: 31.70\n",
      "[10m 13s] Train Epoch: 22 [66560/74759 (89%)]\tLoss: 31.68\n",
      "[10m 14s] Train Epoch: 22 [69120/74759 (92%)]\tLoss: 31.73\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[10m 15s] Train Epoch: 22 [71680/74759 (96%)]\tLoss: 31.79\n",
      "[10m 15s] Train Epoch: 22 [74240/74759 (99%)]\tLoss: 31.81\n",
      "\n",
      "Train set: Accuracy: 70245/74759 (93%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.972676401646534\n",
      "\n",
      "Test set: Accuracy: 16843/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is NP\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[10m 19s] Train Epoch: 23 [2560/74759 (3%)]\tLoss: 25.88\n",
      "[10m 20s] Train Epoch: 23 [5120/74759 (7%)]\tLoss: 27.82\n",
      "[10m 21s] Train Epoch: 23 [7680/74759 (10%)]\tLoss: 28.82\n",
      "[10m 22s] Train Epoch: 23 [10240/74759 (14%)]\tLoss: 29.17\n",
      "[10m 22s] Train Epoch: 23 [12800/74759 (17%)]\tLoss: 29.42\n",
      "[10m 23s] Train Epoch: 23 [15360/74759 (21%)]\tLoss: 29.50\n",
      "[10m 24s] Train Epoch: 23 [17920/74759 (24%)]\tLoss: 28.88\n",
      "[10m 25s] Train Epoch: 23 [20480/74759 (27%)]\tLoss: 29.05\n",
      "[10m 26s] Train Epoch: 23 [23040/74759 (31%)]\tLoss: 29.36\n",
      "[10m 27s] Train Epoch: 23 [25600/74759 (34%)]\tLoss: 29.54\n",
      "[10m 28s] Train Epoch: 23 [28160/74759 (38%)]\tLoss: 30.06\n",
      "[10m 29s] Train Epoch: 23 [30720/74759 (41%)]\tLoss: 30.15\n",
      "[10m 30s] Train Epoch: 23 [33280/74759 (45%)]\tLoss: 30.16\n",
      "[10m 30s] Train Epoch: 23 [35840/74759 (48%)]\tLoss: 30.26\n",
      "[10m 31s] Train Epoch: 23 [38400/74759 (51%)]\tLoss: 30.28\n",
      "[10m 32s] Train Epoch: 23 [40960/74759 (55%)]\tLoss: 30.27\n",
      "[10m 33s] Train Epoch: 23 [43520/74759 (58%)]\tLoss: 30.31\n",
      "[10m 34s] Train Epoch: 23 [46080/74759 (62%)]\tLoss: 30.43\n",
      "[10m 35s] Train Epoch: 23 [48640/74759 (65%)]\tLoss: 30.25\n",
      "[10m 36s] Train Epoch: 23 [51200/74759 (68%)]\tLoss: 30.26\n",
      "[10m 36s] Train Epoch: 23 [53760/74759 (72%)]\tLoss: 30.33\n",
      "[10m 37s] Train Epoch: 23 [56320/74759 (75%)]\tLoss: 30.42\n",
      "[10m 38s] Train Epoch: 23 [58880/74759 (79%)]\tLoss: 30.55\n",
      "[10m 39s] Train Epoch: 23 [61440/74759 (82%)]\tLoss: 30.56\n",
      "[10m 39s] Train Epoch: 23 [64000/74759 (86%)]\tLoss: 30.60\n",
      "[10m 40s] Train Epoch: 23 [66560/74759 (89%)]\tLoss: 30.66\n",
      "[10m 41s] Train Epoch: 23 [69120/74759 (92%)]\tLoss: 30.64\n",
      "[10m 42s] Train Epoch: 23 [71680/74759 (96%)]\tLoss: 30.65\n",
      "[10m 43s] Train Epoch: 23 [74240/74759 (99%)]\tLoss: 30.78\n",
      "\n",
      "Train set: Accuracy: 70511/74759 (94%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9732435047872814\n",
      "\n",
      "Test set: Accuracy: 16885/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[10m 47s] Train Epoch: 24 [2560/74759 (3%)]\tLoss: 27.77\n",
      "[10m 48s] Train Epoch: 24 [5120/74759 (7%)]\tLoss: 26.06\n",
      "[10m 49s] Train Epoch: 24 [7680/74759 (10%)]\tLoss: 25.89\n",
      "[10m 50s] Train Epoch: 24 [10240/74759 (14%)]\tLoss: 26.16\n",
      "[10m 51s] Train Epoch: 24 [12800/74759 (17%)]\tLoss: 26.56\n",
      "[10m 52s] Train Epoch: 24 [15360/74759 (21%)]\tLoss: 26.87\n",
      "[10m 53s] Train Epoch: 24 [17920/74759 (24%)]\tLoss: 27.59\n",
      "[10m 54s] Train Epoch: 24 [20480/74759 (27%)]\tLoss: 27.32\n",
      "[10m 55s] Train Epoch: 24 [23040/74759 (31%)]\tLoss: 27.51\n",
      "[10m 56s] Train Epoch: 24 [25600/74759 (34%)]\tLoss: 27.36\n",
      "[10m 57s] Train Epoch: 24 [28160/74759 (38%)]\tLoss: 27.50\n",
      "[10m 58s] Train Epoch: 24 [30720/74759 (41%)]\tLoss: 27.55\n",
      "[10m 58s] Train Epoch: 24 [33280/74759 (45%)]\tLoss: 27.86\n",
      "[10m 59s] Train Epoch: 24 [35840/74759 (48%)]\tLoss: 28.07\n",
      "[11m 0s] Train Epoch: 24 [38400/74759 (51%)]\tLoss: 28.09\n",
      "[11m 1s] Train Epoch: 24 [40960/74759 (55%)]\tLoss: 28.11\n",
      "[11m 2s] Train Epoch: 24 [43520/74759 (58%)]\tLoss: 28.27\n",
      "[11m 3s] Train Epoch: 24 [46080/74759 (62%)]\tLoss: 28.13\n",
      "[11m 4s] Train Epoch: 24 [48640/74759 (65%)]\tLoss: 28.38\n",
      "[11m 4s] Train Epoch: 24 [51200/74759 (68%)]\tLoss: 28.52\n",
      "[11m 5s] Train Epoch: 24 [53760/74759 (72%)]\tLoss: 28.77\n",
      "[11m 6s] Train Epoch: 24 [56320/74759 (75%)]\tLoss: 28.98\n",
      "[11m 7s] Train Epoch: 24 [58880/74759 (79%)]\tLoss: 29.09\n",
      "[11m 8s] Train Epoch: 24 [61440/74759 (82%)]\tLoss: 29.24\n",
      "[11m 9s] Train Epoch: 24 [64000/74759 (86%)]\tLoss: 29.23\n",
      "[11m 10s] Train Epoch: 24 [66560/74759 (89%)]\tLoss: 29.17\n",
      "[11m 11s] Train Epoch: 24 [69120/74759 (92%)]\tLoss: 29.20\n",
      "[11m 11s] Train Epoch: 24 [71680/74759 (96%)]\tLoss: 29.23\n",
      "[11m 12s] Train Epoch: 24 [74240/74759 (99%)]\tLoss: 29.21\n",
      "\n",
      "Train set: Accuracy: 70640/74759 (94%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9726174768890136\n",
      "\n",
      "Test set: Accuracy: 16903/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[11m 17s] Train Epoch: 25 [2560/74759 (3%)]\tLoss: 29.06\n",
      "[11m 18s] Train Epoch: 25 [5120/74759 (7%)]\tLoss: 26.21\n",
      "[11m 18s] Train Epoch: 25 [7680/74759 (10%)]\tLoss: 27.70\n",
      "[11m 19s] Train Epoch: 25 [10240/74759 (14%)]\tLoss: 27.24\n",
      "[11m 20s] Train Epoch: 25 [12800/74759 (17%)]\tLoss: 27.54\n",
      "[11m 21s] Train Epoch: 25 [15360/74759 (21%)]\tLoss: 27.54\n",
      "[11m 22s] Train Epoch: 25 [17920/74759 (24%)]\tLoss: 27.49\n",
      "[11m 23s] Train Epoch: 25 [20480/74759 (27%)]\tLoss: 27.82\n",
      "[11m 24s] Train Epoch: 25 [23040/74759 (31%)]\tLoss: 28.23\n",
      "[11m 25s] Train Epoch: 25 [25600/74759 (34%)]\tLoss: 28.24\n",
      "[11m 25s] Train Epoch: 25 [28160/74759 (38%)]\tLoss: 28.27\n",
      "[11m 26s] Train Epoch: 25 [30720/74759 (41%)]\tLoss: 28.57\n",
      "[11m 27s] Train Epoch: 25 [33280/74759 (45%)]\tLoss: 28.62\n",
      "[11m 28s] Train Epoch: 25 [35840/74759 (48%)]\tLoss: 28.70\n",
      "[11m 29s] Train Epoch: 25 [38400/74759 (51%)]\tLoss: 28.86\n",
      "[11m 30s] Train Epoch: 25 [40960/74759 (55%)]\tLoss: 28.73\n",
      "[11m 30s] Train Epoch: 25 [43520/74759 (58%)]\tLoss: 28.65\n",
      "[11m 31s] Train Epoch: 25 [46080/74759 (62%)]\tLoss: 28.65\n",
      "[11m 32s] Train Epoch: 25 [48640/74759 (65%)]\tLoss: 28.69\n",
      "[11m 33s] Train Epoch: 25 [51200/74759 (68%)]\tLoss: 28.72\n",
      "[11m 34s] Train Epoch: 25 [53760/74759 (72%)]\tLoss: 28.80\n",
      "[11m 35s] Train Epoch: 25 [56320/74759 (75%)]\tLoss: 28.79\n",
      "[11m 35s] Train Epoch: 25 [58880/74759 (79%)]\tLoss: 28.81\n",
      "[11m 36s] Train Epoch: 25 [61440/74759 (82%)]\tLoss: 28.90\n",
      "[11m 37s] Train Epoch: 25 [64000/74759 (86%)]\tLoss: 29.09\n",
      "[11m 38s] Train Epoch: 25 [66560/74759 (89%)]\tLoss: 29.08\n",
      "[11m 39s] Train Epoch: 25 [69120/74759 (92%)]\tLoss: 29.16\n",
      "[11m 40s] Train Epoch: 25 [71680/74759 (96%)]\tLoss: 29.17\n",
      "[11m 41s] Train Epoch: 25 [74240/74759 (99%)]\tLoss: 29.28\n",
      "\n",
      "Train set: Accuracy: 70587/74759 (94%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9731007500032316\n",
      "\n",
      "Test set: Accuracy: 16866/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[11m 45s] Train Epoch: 26 [2560/74759 (3%)]\tLoss: 24.81\n",
      "[11m 46s] Train Epoch: 26 [5120/74759 (7%)]\tLoss: 24.94\n",
      "[11m 46s] Train Epoch: 26 [7680/74759 (10%)]\tLoss: 25.81\n",
      "[11m 47s] Train Epoch: 26 [10240/74759 (14%)]\tLoss: 25.73\n",
      "[11m 48s] Train Epoch: 26 [12800/74759 (17%)]\tLoss: 26.51\n",
      "[11m 49s] Train Epoch: 26 [15360/74759 (21%)]\tLoss: 26.68\n",
      "[11m 50s] Train Epoch: 26 [17920/74759 (24%)]\tLoss: 26.83\n",
      "[11m 50s] Train Epoch: 26 [20480/74759 (27%)]\tLoss: 26.66\n",
      "[11m 51s] Train Epoch: 26 [23040/74759 (31%)]\tLoss: 26.50\n",
      "[11m 52s] Train Epoch: 26 [25600/74759 (34%)]\tLoss: 26.61\n",
      "[11m 53s] Train Epoch: 26 [28160/74759 (38%)]\tLoss: 26.83\n",
      "[11m 54s] Train Epoch: 26 [30720/74759 (41%)]\tLoss: 27.16\n",
      "[11m 55s] Train Epoch: 26 [33280/74759 (45%)]\tLoss: 27.28\n",
      "[11m 56s] Train Epoch: 26 [35840/74759 (48%)]\tLoss: 27.51\n",
      "[11m 56s] Train Epoch: 26 [38400/74759 (51%)]\tLoss: 27.52\n",
      "[11m 57s] Train Epoch: 26 [40960/74759 (55%)]\tLoss: 27.43\n",
      "[11m 58s] Train Epoch: 26 [43520/74759 (58%)]\tLoss: 27.53\n",
      "[11m 59s] Train Epoch: 26 [46080/74759 (62%)]\tLoss: 27.45\n",
      "[12m 0s] Train Epoch: 26 [48640/74759 (65%)]\tLoss: 27.49\n",
      "[12m 1s] Train Epoch: 26 [51200/74759 (68%)]\tLoss: 27.46\n",
      "[12m 2s] Train Epoch: 26 [53760/74759 (72%)]\tLoss: 27.69\n",
      "[12m 3s] Train Epoch: 26 [56320/74759 (75%)]\tLoss: 27.77\n",
      "[12m 4s] Train Epoch: 26 [58880/74759 (79%)]\tLoss: 27.95\n",
      "[12m 5s] Train Epoch: 26 [61440/74759 (82%)]\tLoss: 28.05\n",
      "[12m 6s] Train Epoch: 26 [64000/74759 (86%)]\tLoss: 28.09\n",
      "[12m 7s] Train Epoch: 26 [66560/74759 (89%)]\tLoss: 28.22\n",
      "[12m 7s] Train Epoch: 26 [69120/74759 (92%)]\tLoss: 28.34\n",
      "[12m 8s] Train Epoch: 26 [71680/74759 (96%)]\tLoss: 28.47\n",
      "[12m 9s] Train Epoch: 26 [74240/74759 (99%)]\tLoss: 28.58\n",
      "\n",
      "Train set: Accuracy: 70730/74759 (94%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9719613108428588\n",
      "\n",
      "Test set: Accuracy: 16793/18691 (89%) Recall: (24%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[12m 13s] Train Epoch: 27 [2560/74759 (3%)]\tLoss: 30.24\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[12m 14s] Train Epoch: 27 [5120/74759 (7%)]\tLoss: 28.98\n",
      "[12m 15s] Train Epoch: 27 [7680/74759 (10%)]\tLoss: 29.67\n",
      "[12m 16s] Train Epoch: 27 [10240/74759 (14%)]\tLoss: 29.13\n",
      "[12m 17s] Train Epoch: 27 [12800/74759 (17%)]\tLoss: 29.43\n",
      "[12m 18s] Train Epoch: 27 [15360/74759 (21%)]\tLoss: 29.22\n",
      "[12m 18s] Train Epoch: 27 [17920/74759 (24%)]\tLoss: 29.19\n",
      "[12m 19s] Train Epoch: 27 [20480/74759 (27%)]\tLoss: 29.07\n",
      "[12m 20s] Train Epoch: 27 [23040/74759 (31%)]\tLoss: 28.69\n",
      "[12m 21s] Train Epoch: 27 [25600/74759 (34%)]\tLoss: 28.40\n",
      "[12m 22s] Train Epoch: 27 [28160/74759 (38%)]\tLoss: 28.19\n",
      "[12m 23s] Train Epoch: 27 [30720/74759 (41%)]\tLoss: 28.48\n",
      "[12m 23s] Train Epoch: 27 [33280/74759 (45%)]\tLoss: 28.44\n",
      "[12m 24s] Train Epoch: 27 [35840/74759 (48%)]\tLoss: 28.36\n",
      "[12m 25s] Train Epoch: 27 [38400/74759 (51%)]\tLoss: 28.39\n",
      "[12m 26s] Train Epoch: 27 [40960/74759 (55%)]\tLoss: 28.42\n",
      "[12m 27s] Train Epoch: 27 [43520/74759 (58%)]\tLoss: 28.40\n",
      "[12m 27s] Train Epoch: 27 [46080/74759 (62%)]\tLoss: 28.52\n",
      "[12m 28s] Train Epoch: 27 [48640/74759 (65%)]\tLoss: 28.63\n",
      "[12m 29s] Train Epoch: 27 [51200/74759 (68%)]\tLoss: 28.73\n",
      "[12m 30s] Train Epoch: 27 [53760/74759 (72%)]\tLoss: 28.78\n",
      "[12m 30s] Train Epoch: 27 [56320/74759 (75%)]\tLoss: 28.73\n",
      "[12m 31s] Train Epoch: 27 [58880/74759 (79%)]\tLoss: 28.66\n",
      "[12m 32s] Train Epoch: 27 [61440/74759 (82%)]\tLoss: 28.69\n",
      "[12m 33s] Train Epoch: 27 [64000/74759 (86%)]\tLoss: 28.65\n",
      "[12m 34s] Train Epoch: 27 [66560/74759 (89%)]\tLoss: 28.67\n",
      "[12m 34s] Train Epoch: 27 [69120/74759 (92%)]\tLoss: 28.67\n",
      "[12m 35s] Train Epoch: 27 [71680/74759 (96%)]\tLoss: 28.68\n",
      "[12m 36s] Train Epoch: 27 [74240/74759 (99%)]\tLoss: 28.67\n",
      "\n",
      "Train set: Accuracy: 70747/74759 (94%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9732278061506473\n",
      "\n",
      "Test set: Accuracy: 16948/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[12m 40s] Train Epoch: 28 [2560/74759 (3%)]\tLoss: 26.89\n",
      "[12m 41s] Train Epoch: 28 [5120/74759 (7%)]\tLoss: 27.10\n",
      "[12m 42s] Train Epoch: 28 [7680/74759 (10%)]\tLoss: 27.07\n",
      "[12m 43s] Train Epoch: 28 [10240/74759 (14%)]\tLoss: 26.61\n",
      "[12m 44s] Train Epoch: 28 [12800/74759 (17%)]\tLoss: 26.57\n",
      "[12m 44s] Train Epoch: 28 [15360/74759 (21%)]\tLoss: 26.50\n",
      "[12m 45s] Train Epoch: 28 [17920/74759 (24%)]\tLoss: 26.56\n",
      "[12m 46s] Train Epoch: 28 [20480/74759 (27%)]\tLoss: 26.45\n",
      "[12m 47s] Train Epoch: 28 [23040/74759 (31%)]\tLoss: 26.31\n",
      "[12m 48s] Train Epoch: 28 [25600/74759 (34%)]\tLoss: 26.61\n",
      "[12m 49s] Train Epoch: 28 [28160/74759 (38%)]\tLoss: 26.62\n",
      "[12m 50s] Train Epoch: 28 [30720/74759 (41%)]\tLoss: 26.56\n",
      "[12m 50s] Train Epoch: 28 [33280/74759 (45%)]\tLoss: 26.52\n",
      "[12m 51s] Train Epoch: 28 [35840/74759 (48%)]\tLoss: 26.60\n",
      "[12m 52s] Train Epoch: 28 [38400/74759 (51%)]\tLoss: 26.57\n",
      "[12m 53s] Train Epoch: 28 [40960/74759 (55%)]\tLoss: 26.79\n",
      "[12m 54s] Train Epoch: 28 [43520/74759 (58%)]\tLoss: 26.92\n",
      "[12m 54s] Train Epoch: 28 [46080/74759 (62%)]\tLoss: 27.00\n",
      "[12m 55s] Train Epoch: 28 [48640/74759 (65%)]\tLoss: 27.09\n",
      "[12m 56s] Train Epoch: 28 [51200/74759 (68%)]\tLoss: 27.30\n",
      "[12m 57s] Train Epoch: 28 [53760/74759 (72%)]\tLoss: 27.39\n",
      "[12m 58s] Train Epoch: 28 [56320/74759 (75%)]\tLoss: 27.49\n",
      "[12m 59s] Train Epoch: 28 [58880/74759 (79%)]\tLoss: 27.63\n",
      "[13m 0s] Train Epoch: 28 [61440/74759 (82%)]\tLoss: 27.84\n",
      "[13m 0s] Train Epoch: 28 [64000/74759 (86%)]\tLoss: 27.97\n",
      "[13m 1s] Train Epoch: 28 [66560/74759 (89%)]\tLoss: 28.03\n",
      "[13m 2s] Train Epoch: 28 [69120/74759 (92%)]\tLoss: 28.02\n",
      "[13m 3s] Train Epoch: 28 [71680/74759 (96%)]\tLoss: 28.02\n",
      "[13m 4s] Train Epoch: 28 [74240/74759 (99%)]\tLoss: 27.93\n",
      "\n",
      "Train set: Accuracy: 70839/74759 (94%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9723497221439505\n",
      "\n",
      "Test set: Accuracy: 16859/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is SM\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[13m 8s] Train Epoch: 29 [2560/74759 (3%)]\tLoss: 23.23\n",
      "[13m 9s] Train Epoch: 29 [5120/74759 (7%)]\tLoss: 23.29\n",
      "[13m 10s] Train Epoch: 29 [7680/74759 (10%)]\tLoss: 23.34\n",
      "[13m 11s] Train Epoch: 29 [10240/74759 (14%)]\tLoss: 25.08\n",
      "[13m 12s] Train Epoch: 29 [12800/74759 (17%)]\tLoss: 24.60\n",
      "[13m 13s] Train Epoch: 29 [15360/74759 (21%)]\tLoss: 24.57\n",
      "[13m 14s] Train Epoch: 29 [17920/74759 (24%)]\tLoss: 24.99\n",
      "[13m 15s] Train Epoch: 29 [20480/74759 (27%)]\tLoss: 25.46\n",
      "[13m 16s] Train Epoch: 29 [23040/74759 (31%)]\tLoss: 25.73\n",
      "[13m 17s] Train Epoch: 29 [25600/74759 (34%)]\tLoss: 26.03\n",
      "[13m 17s] Train Epoch: 29 [28160/74759 (38%)]\tLoss: 26.13\n",
      "[13m 18s] Train Epoch: 29 [30720/74759 (41%)]\tLoss: 26.28\n",
      "[13m 19s] Train Epoch: 29 [33280/74759 (45%)]\tLoss: 26.27\n",
      "[13m 20s] Train Epoch: 29 [35840/74759 (48%)]\tLoss: 26.13\n",
      "[13m 21s] Train Epoch: 29 [38400/74759 (51%)]\tLoss: 26.19\n",
      "[13m 22s] Train Epoch: 29 [40960/74759 (55%)]\tLoss: 26.31\n",
      "[13m 23s] Train Epoch: 29 [43520/74759 (58%)]\tLoss: 26.24\n",
      "[13m 24s] Train Epoch: 29 [46080/74759 (62%)]\tLoss: 26.33\n",
      "[13m 25s] Train Epoch: 29 [48640/74759 (65%)]\tLoss: 26.27\n",
      "[13m 25s] Train Epoch: 29 [51200/74759 (68%)]\tLoss: 26.38\n",
      "[13m 26s] Train Epoch: 29 [53760/74759 (72%)]\tLoss: 26.46\n",
      "[13m 27s] Train Epoch: 29 [56320/74759 (75%)]\tLoss: 26.62\n",
      "[13m 28s] Train Epoch: 29 [58880/74759 (79%)]\tLoss: 26.69\n",
      "[13m 29s] Train Epoch: 29 [61440/74759 (82%)]\tLoss: 26.82\n",
      "[13m 30s] Train Epoch: 29 [64000/74759 (86%)]\tLoss: 26.98\n",
      "[13m 30s] Train Epoch: 29 [66560/74759 (89%)]\tLoss: 27.01\n",
      "[13m 31s] Train Epoch: 29 [69120/74759 (92%)]\tLoss: 27.14\n",
      "[13m 32s] Train Epoch: 29 [71680/74759 (96%)]\tLoss: 27.23\n",
      "[13m 33s] Train Epoch: 29 [74240/74759 (99%)]\tLoss: 27.29\n",
      "\n",
      "Train set: Accuracy: 70940/74759 (94%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9719315365706556\n",
      "\n",
      "Test set: Accuracy: 16850/18691 (90%) Recall: (23%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n",
      "74759\n",
      "[13m 37s] Train Epoch: 30 [2560/74759 (3%)]\tLoss: 41.14\n",
      "[13m 38s] Train Epoch: 30 [5120/74759 (7%)]\tLoss: 41.75\n",
      "[13m 39s] Train Epoch: 30 [7680/74759 (10%)]\tLoss: 40.34\n",
      "[13m 40s] Train Epoch: 30 [10240/74759 (14%)]\tLoss: 38.81\n",
      "[13m 41s] Train Epoch: 30 [12800/74759 (17%)]\tLoss: 37.54\n",
      "[13m 42s] Train Epoch: 30 [15360/74759 (21%)]\tLoss: 36.71\n",
      "[13m 42s] Train Epoch: 30 [17920/74759 (24%)]\tLoss: 35.80\n",
      "[13m 43s] Train Epoch: 30 [20480/74759 (27%)]\tLoss: 36.31\n",
      "[13m 44s] Train Epoch: 30 [23040/74759 (31%)]\tLoss: 35.80\n",
      "[13m 45s] Train Epoch: 30 [25600/74759 (34%)]\tLoss: 35.58\n",
      "[13m 46s] Train Epoch: 30 [28160/74759 (38%)]\tLoss: 35.09\n",
      "[13m 47s] Train Epoch: 30 [30720/74759 (41%)]\tLoss: 34.58\n",
      "[13m 47s] Train Epoch: 30 [33280/74759 (45%)]\tLoss: 34.14\n",
      "[13m 48s] Train Epoch: 30 [35840/74759 (48%)]\tLoss: 33.76\n",
      "[13m 49s] Train Epoch: 30 [38400/74759 (51%)]\tLoss: 33.62\n",
      "[13m 50s] Train Epoch: 30 [40960/74759 (55%)]\tLoss: 33.60\n",
      "[13m 51s] Train Epoch: 30 [43520/74759 (58%)]\tLoss: 33.24\n",
      "[13m 52s] Train Epoch: 30 [46080/74759 (62%)]\tLoss: 33.06\n",
      "[13m 52s] Train Epoch: 30 [48640/74759 (65%)]\tLoss: 32.81\n",
      "[13m 53s] Train Epoch: 30 [51200/74759 (68%)]\tLoss: 32.50\n",
      "[13m 54s] Train Epoch: 30 [53760/74759 (72%)]\tLoss: 32.22\n",
      "[13m 55s] Train Epoch: 30 [56320/74759 (75%)]\tLoss: 31.91\n",
      "[13m 55s] Train Epoch: 30 [58880/74759 (79%)]\tLoss: 31.80\n",
      "[13m 56s] Train Epoch: 30 [61440/74759 (82%)]\tLoss: 31.69\n",
      "[13m 57s] Train Epoch: 30 [64000/74759 (86%)]\tLoss: 31.62\n",
      "[13m 58s] Train Epoch: 30 [66560/74759 (89%)]\tLoss: 31.61\n",
      "[13m 59s] Train Epoch: 30 [69120/74759 (92%)]\tLoss: 31.49\n",
      "[14m 0s] Train Epoch: 30 [71680/74759 (96%)]\tLoss: 31.44\n",
      "[14m 0s] Train Epoch: 30 [74240/74759 (99%)]\tLoss: 31.45\n",
      "\n",
      "Train set: Accuracy: 70521/74759 (94%) \n",
      "evaluating trained model ...\n",
      "AUC =  0.9717834507190514\n",
      "\n",
      "Test set: Accuracy: 16879/18691 (90%) Recall: (24%)\n",
      "\n",
      "C1ccccC1 is NP\n",
      "NC(=O)CC(NC(=O)c1ccco1)C(=O)O is NP\n",
      "Cc1ccccc1N1CC(C)C(=O)NC1=S is SM\n",
      "COC(=O)c1c[nH]c2nnsc12 is SM\n"
     ]
    }
   ],
   "source": [
    "classifier = RNNClassifier(N_CHARS, HIDDEN_SIZE, N_PROPERTIES, N_LAYERS)\n",
    "# if torch.cuda.device_count() > 1:\n",
    "#     print(\"Let's use\", torch.cuda.device_count(), \"GPUs!\")\n",
    "#     # dim = 0 [33, xxx] -> [11, ...], [11, ...], [11, ...] on 3 GPUs\n",
    "#     classifier = nn.DataParallel(classifier)\n",
    "\n",
    "if torch.cuda.is_available():\n",
    "    classifier.cuda()\n",
    "\n",
    "optimizer = torch.optim.Adam(classifier.parameters(), lr=0.001)\n",
    "criterion = nn.CrossEntropyLoss()\n",
    "\n",
    "start = time.time()\n",
    "\n",
    "print(\"Training for %d epochs...\" % N_EPOCHS)\n",
    "\n",
    "all_losses = []\n",
    "all_accuracy = []\n",
    "for epoch in range(1, N_EPOCHS + 1):\n",
    "    # Train cycle\n",
    "    all_losses.append(train())\n",
    "    # Testing\n",
    "    all_accuracy.append(test())\n",
    "    \n",
    "\n",
    "    # Testing several samples\n",
    "    test(\"C1ccccC1\")\n",
    "    test(\"NC(=O)CC(NC(=O)c1ccco1)C(=O)O\")\n",
    "    test(\"Cc1ccccc1N1CC(C)C(=O)NC1=S\")\n",
    "    test(\"COC(=O)c1c[nH]c2nnsc12\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f5088831c88>]"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.ticker as ticker\n",
    "\n",
    "\n",
    "plt.plot(all_losses)\n",
    "# plit.plot(all_accuracy)"
   ]
  },
  {
   "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}