tutorials/advanced/modelling/chemprop_models.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get rid of some chemprop warnings\n",
    "import warnings\n",
    "\n",
    "warnings.simplefilter(action='ignore', category=FutureWarning)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training Chemprop models using QSPRpred\n",
    "\n",
    "This notebook demonstrates how to train Chemprop models using QSPRpred.\n",
    "Chemprop is a library for training and evaluating message passing neural networks for molecular property prediction [1, 2, 3].\n",
    "Using QSPRpred, we have wrapped Chemprop to allow for training Chemprop models using QSPRpred.\n",
    "Not all functionality of Chemprop is available through QSPRpred, but we hope to expand this in the future.\n",
    "\n",
    "1. Yang K, et al. Analyzing Learned Molecular Representations for Property Prediction. J Chem Inf Model. 2019;59(8):3370-3388. doi:10.1021/acs.jcim.9b00237\n",
    "2. Heid E, et al. Chemprop: Machine Learning Package for Chemical Property Prediction. ChemRxiv. 2023; doi:10.26434 chemrxiv-2023-3zcfl (preprint)\n",
    "3. https://github.com/chemprop/chemprop"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's start by preparing the data, which is identical to the data preparation in other tutorials.\n",
    "With the exception that we need to specify that we want to use SMILES strings as features.\n",
    "It is possible to use other features in addition to SMILES strings, but the SMILES strings must be included in the features as input for the message passing neural network.\n",
    "\n",
    "Chemprop also provides the option to specify custom atom and bond features, but this is not (yet) supported in QSPRpred."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Failed to find the pandas get_adjustment() function to patch\n",
      "Failed to patch pandas - PandasTools will have limited functionality\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>SMILES</th>\n",
       "      <th>pchembl_value_Mean</th>\n",
       "      <th>Year</th>\n",
       "      <th>QSPRID</th>\n",
       "      <th>pchembl_value_Mean_original</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>QSPRID</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>ChempropTutorialDataset_0000</th>\n",
       "      <td>Cc1cc(C)n(-c2cc(NC(=O)CCN(C)C)nc(-c3ccc(C)o3)n...</td>\n",
       "      <td>8.68</td>\n",
       "      <td>2008.0</td>\n",
       "      <td>ChempropTutorialDataset_0000</td>\n",
       "      <td>8.68</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ChempropTutorialDataset_0001</th>\n",
       "      <td>Nc1c(C(=O)Nc2ccc([N+](=O)[O-])cc2)sc2nc3c(cc12...</td>\n",
       "      <td>4.82</td>\n",
       "      <td>2010.0</td>\n",
       "      <td>ChempropTutorialDataset_0001</td>\n",
       "      <td>4.82</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ChempropTutorialDataset_0002</th>\n",
       "      <td>O=C(Nc1nc2ncccc2n2c(=O)n(-c3ccccc3)nc12)c1ccccc1</td>\n",
       "      <td>5.65</td>\n",
       "      <td>2009.0</td>\n",
       "      <td>ChempropTutorialDataset_0002</td>\n",
       "      <td>5.65</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ChempropTutorialDataset_0003</th>\n",
       "      <td>CNC(=O)C12CC1C(n1cnc3c(NCc4cccc(Cl)c4)nc(C#CCC...</td>\n",
       "      <td>5.45</td>\n",
       "      <td>2009.0</td>\n",
       "      <td>ChempropTutorialDataset_0003</td>\n",
       "      <td>5.45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ChempropTutorialDataset_0004</th>\n",
       "      <td>CCCn1c(=O)c2c(nc3cc(OC)ccn32)n(CCCNC(=O)c2ccc(...</td>\n",
       "      <td>5.20</td>\n",
       "      <td>2019.0</td>\n",
       "      <td>ChempropTutorialDataset_0004</td>\n",
       "      <td>5.20</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                         SMILES  \\\n",
       "QSPRID                                                                            \n",
       "ChempropTutorialDataset_0000  Cc1cc(C)n(-c2cc(NC(=O)CCN(C)C)nc(-c3ccc(C)o3)n...   \n",
       "ChempropTutorialDataset_0001  Nc1c(C(=O)Nc2ccc([N+](=O)[O-])cc2)sc2nc3c(cc12...   \n",
       "ChempropTutorialDataset_0002   O=C(Nc1nc2ncccc2n2c(=O)n(-c3ccccc3)nc12)c1ccccc1   \n",
       "ChempropTutorialDataset_0003  CNC(=O)C12CC1C(n1cnc3c(NCc4cccc(Cl)c4)nc(C#CCC...   \n",
       "ChempropTutorialDataset_0004  CCCn1c(=O)c2c(nc3cc(OC)ccn32)n(CCCNC(=O)c2ccc(...   \n",
       "\n",
       "                              pchembl_value_Mean    Year  \\\n",
       "QSPRID                                                     \n",
       "ChempropTutorialDataset_0000                8.68  2008.0   \n",
       "ChempropTutorialDataset_0001                4.82  2010.0   \n",
       "ChempropTutorialDataset_0002                5.65  2009.0   \n",
       "ChempropTutorialDataset_0003                5.45  2009.0   \n",
       "ChempropTutorialDataset_0004                5.20  2019.0   \n",
       "\n",
       "                                                    QSPRID  \\\n",
       "QSPRID                                                       \n",
       "ChempropTutorialDataset_0000  ChempropTutorialDataset_0000   \n",
       "ChempropTutorialDataset_0001  ChempropTutorialDataset_0001   \n",
       "ChempropTutorialDataset_0002  ChempropTutorialDataset_0002   \n",
       "ChempropTutorialDataset_0003  ChempropTutorialDataset_0003   \n",
       "ChempropTutorialDataset_0004  ChempropTutorialDataset_0004   \n",
       "\n",
       "                              pchembl_value_Mean_original  \n",
       "QSPRID                                                     \n",
       "ChempropTutorialDataset_0000                         8.68  \n",
       "ChempropTutorialDataset_0001                         4.82  \n",
       "ChempropTutorialDataset_0002                         5.65  \n",
       "ChempropTutorialDataset_0003                         5.45  \n",
       "ChempropTutorialDataset_0004                         5.20  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import os\n",
    "\n",
    "from qsprpred.data import QSPRDataset, RandomSplit\n",
    "from qsprpred.data.descriptors.fingerprints import MorganFP\n",
    "from qsprpred.data.descriptors.sets import SmilesDesc\n",
    "\n",
    "os.makedirs(\"../../tutorial_output/data\", exist_ok=True)\n",
    "\n",
    "# Create dataset\n",
    "dataset = QSPRDataset.fromTableFile(\n",
    "    filename=\"../../tutorial_data/A2A_LIGANDS.tsv\",\n",
    "    store_dir=\"../../tutorial_output/data\",\n",
    "    name=\"ChempropTutorialDataset\",\n",
    "    target_props=[{\"name\": \"pchembl_value_Mean\", \"task\": \"REGRESSION\"}],\n",
    "    random_state=42\n",
    ")\n",
    "\n",
    "# calculate compound features and split dataset into train and test\n",
    "feature_calculators = [SmilesDesc(), MorganFP(radius=3, nBits=2048)]\n",
    "dataset.prepareDataset(\n",
    "    split=RandomSplit(test_fraction=0.2, dataset=dataset),\n",
    "    feature_calculators=feature_calculators,\n",
    "    recalculate_features=True,\n",
    ")\n",
    "\n",
    "dataset.getDF().head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "After preparing the data, we can train a model.\n",
    "Initizalizing a `ChempropModel` is identical to initializing other `QSPRModel` models.\n",
    "We can specify set `quiet_logger` to False to get the debug output from Chemprop.\n",
    "\n",
    "As not all Chemprop functionality is available through QSPRpred, some parameters are not used in QSPRpred.\n",
    "Therefore, a convienience function `getAvailableParameters` is provided to get a dictionary of the arguments that are used in QSPRpred.\n",
    "Definitions and default values may be found on the Chemprop github (https://github.com/chemprop/chemprop/blob/master/chemprop/args.py)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'no_cuda': 'Turn off cuda (i.e., use CPU instead of GPU).',\n",
       " 'gpu': 'Which GPU to use.',\n",
       " 'num_workers': 'Number of workers for the parallel data loading (0 means sequential).',\n",
       " 'batch_size': 'Batch size.',\n",
       " 'no_cache_mol': 'Whether to not cache the RDKit molecule for each SMILES string to reduce memory usage (cached by default).',\n",
       " 'empty_cache': 'Whether to empty all caches before training or predicting. This is necessary if multiple jobs are run within a single script and the atom or bond features change.',\n",
       " 'loss_function': 'Choice of loss function. Loss functions are limited to compatible dataset types.',\n",
       " 'metric': \"Metric to use with the validation set for early stopping. Defaults to 'auc' for classification, 'rmse' for regression. Note. In Chemprop this metric is also used for test-set evaluation, but in QSPRpred this is determined by the scoring parameter in assessment.\",\n",
       " 'bias': 'Whether to add bias to linear layers.',\n",
       " 'hidden_size': 'Dimensionality of hidden layers in MPN.',\n",
       " 'depth': 'Number of message passing steps.',\n",
       " 'mpn_shared': \"Whether to use the same message passing neural network for all input molecule Only relevant if 'number_of_molecules > 1'\",\n",
       " 'dropout': 'Dropout probability.',\n",
       " 'activation': 'Activation function.',\n",
       " 'atom_messages': 'Centers messages on atoms instead of on bonds.',\n",
       " 'undirected': 'Undirected edges (always sum the two relevant bond vectors).',\n",
       " 'ffn_hidden_size': 'Hidden dim for higher-capacity FFN (defaults to hidden_size).',\n",
       " 'ffn_num_layers': 'Number of layers in FFN after MPN encoding.',\n",
       " 'epochs': 'Number of epochs to run.',\n",
       " 'warmup_epochs': \"Number of epochs during which learning rate increases linearly from 'init_lr' to 'max_lr'. Afterwards, learning rate decreases exponentially from 'max_lr' to 'final_lr'.\",\n",
       " 'init_lr': 'Initial learning rate.',\n",
       " 'max_lr': 'Maximum learning rate.',\n",
       " 'final_lr': 'Final learning rate.',\n",
       " 'grad_clip': 'Maximum magnitude of gradient during training.',\n",
       " 'class_balance': 'Trains with an equal number of positives and negatives in each batch.',\n",
       " 'evidential_regularization': 'Value used in regularization for evidential loss function. The default value recommended by Soleimany et al.(2021) is 0.2. Optimal value is dataset-dependent; it is recommended that users test different values to find the best value for their model.'}"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from qsprpred.extra.gpu.models.chemprop import ChempropModel\n",
    "\n",
    "ChempropModel.getAvailableParameters()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will use default parameters, except for the number of epochs, which we will set to 5 for demonstration purposes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create model\n",
    "os.makedirs(\"../../tutorial_output/models\", exist_ok=True)\n",
    "model = ChempropModel(\n",
    "    base_dir='../../tutorial_output/models',\n",
    "    name='ChempropTutorialModel',\n",
    "    parameters={\"epochs\": 5},\n",
    "    quiet_logger=False\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Training a `ChempropModel` works the same as training another `QSPRModel` that supports early stopping.\n",
    "Have a look at the [deep learning tutorial](../../basics/data/data_splitting.ipynb) for more information on early stopping."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "qsprpred - WARNING - Random state supplied, but alg <class 'qsprpred.extra.gpu.models.chemprop.ChempropMoleculeModel'> does not support it. Ignoring this setting.\n",
      "train size = 2,350 | val size = 262\n",
      "Total size = 2,612\n",
      "Fitting scaler\n",
      "Number of parameters = 355,201\n",
      "Moving model to cuda\n",
      "  0%|          | 0/5 [00:00<?, ?it/s]Epoch 0\n",
      "Loss = 9.8159e-01, PNorm = 33.9640, GNorm = 1.1326, lr_0 = 2.0532e-04\n",
      "Loss = 9.9579e-01, PNorm = 33.9681, GNorm = 0.7663, lr_0 = 3.0106e-04\n",
      "Loss = 9.2394e-01, PNorm = 33.9736, GNorm = 3.8675, lr_0 = 3.9681e-04\n",
      "Loss = 1.0338e+00, PNorm = 33.9792, GNorm = 3.0841, lr_0 = 4.9255e-04\n",
      "Validation rmse = 1.197342\n",
      "Model best validation rmse = 1.197342 on epoch                     0\n",
      " 20%|██        | 1/5 [00:17<01:09, 17.35s/it]Epoch 1\n",
      "Loss = 9.1158e-01, PNorm = 33.9879, GNorm = 2.1077, lr_0 = 5.8830e-04\n",
      "Loss = 8.7446e-01, PNorm = 34.0052, GNorm = 1.5421, lr_0 = 6.8404e-04\n",
      "Loss = 9.2941e-01, PNorm = 34.0316, GNorm = 2.3650, lr_0 = 7.7979e-04\n",
      "Loss = 1.0071e+00, PNorm = 34.0677, GNorm = 1.2049, lr_0 = 8.7553e-04\n",
      "Loss = 8.1503e-01, PNorm = 34.1191, GNorm = 0.5373, lr_0 = 9.7128e-04\n",
      "Validation rmse = 1.152012\n",
      "Model best validation rmse = 1.152012 on epoch                     1\n",
      " 40%|████      | 2/5 [00:21<00:27,  9.33s/it]Epoch 2\n",
      "Loss = 9.2997e-01, PNorm = 34.1583, GNorm = 1.0126, lr_0 = 8.9198e-04\n",
      "Loss = 7.7931e-01, PNorm = 34.2090, GNorm = 1.6003, lr_0 = 7.5759e-04\n",
      "Loss = 8.7472e-01, PNorm = 34.2593, GNorm = 0.7230, lr_0 = 6.4344e-04\n",
      "Loss = 9.0038e-01, PNorm = 34.2979, GNorm = 4.3859, lr_0 = 5.4650e-04\n",
      "Loss = 7.8004e-01, PNorm = 34.3264, GNorm = 2.6959, lr_0 = 4.6416e-04\n",
      "Validation rmse = 1.061348\n",
      "Model best validation rmse = 1.061348 on epoch                     2\n",
      " 60%|██████    | 3/5 [00:25<00:14,  7.02s/it]Epoch 3\n",
      "Loss = 7.6340e-01, PNorm = 34.3565, GNorm = 0.6348, lr_0 = 3.9423e-04\n",
      "Loss = 8.3930e-01, PNorm = 34.3816, GNorm = 0.8542, lr_0 = 3.3483e-04\n",
      "Loss = 7.5338e-01, PNorm = 34.4039, GNorm = 2.1820, lr_0 = 2.8438e-04\n",
      "Loss = 7.9965e-01, PNorm = 34.4240, GNorm = 4.0464, lr_0 = 2.4153e-04\n",
      "Validation rmse = 1.007880\n",
      "Model best validation rmse = 1.007880 on epoch                     3\n",
      " 80%|████████  | 4/5 [00:29<00:05,  5.71s/it]Epoch 4\n",
      "Loss = 9.4934e-01, PNorm = 34.4442, GNorm = 2.7529, lr_0 = 2.0514e-04\n",
      "Loss = 7.1463e-01, PNorm = 34.4629, GNorm = 1.7879, lr_0 = 1.7423e-04\n",
      "Loss = 7.0122e-01, PNorm = 34.4785, GNorm = 0.7353, lr_0 = 1.4798e-04\n",
      "Loss = 7.2596e-01, PNorm = 34.4897, GNorm = 2.7668, lr_0 = 1.2569e-04\n",
      "Loss = 6.8799e-01, PNorm = 34.4982, GNorm = 2.2652, lr_0 = 1.0675e-04\n",
      "Validation rmse = 0.988709\n",
      "Model best validation rmse = 0.988709 on epoch                     4\n",
      "100%|██████████| 5/5 [00:32<00:00,  6.53s/it]\n",
      "train size = 2,350 | val size = 262\n",
      "Total size = 2,612\n",
      "Fitting scaler\n",
      "Number of parameters = 355,201\n",
      "Moving model to cuda\n",
      "  0%|          | 0/5 [00:00<?, ?it/s]Epoch 0\n",
      "Loss = 1.0248e+00, PNorm = 33.9641, GNorm = 2.8310, lr_0 = 2.0532e-04\n",
      "Loss = 9.8230e-01, PNorm = 33.9672, GNorm = 2.7039, lr_0 = 3.0106e-04\n",
      "Loss = 8.6326e-01, PNorm = 33.9752, GNorm = 6.2004, lr_0 = 3.9681e-04\n",
      "Loss = 9.9771e-01, PNorm = 33.9858, GNorm = 17.7899, lr_0 = 4.9255e-04\n",
      "Validation rmse = 1.187259\n",
      "Model best validation rmse = 1.187259 on epoch                     0\n",
      " 20%|██        | 1/5 [00:04<00:18,  4.64s/it]Epoch 1\n",
      "Loss = 1.0094e+00, PNorm = 33.9941, GNorm = 7.3664, lr_0 = 5.8830e-04\n",
      "Loss = 1.0443e+00, PNorm = 34.0092, GNorm = 0.8962, lr_0 = 6.8404e-04\n",
      "Loss = 9.4559e-01, PNorm = 34.0362, GNorm = 2.8125, lr_0 = 7.7979e-04\n",
      "Loss = 8.1075e-01, PNorm = 34.0716, GNorm = 1.2867, lr_0 = 8.7553e-04\n",
      "Loss = 9.4318e-01, PNorm = 34.1086, GNorm = 1.1371, lr_0 = 9.7128e-04\n",
      "Validation rmse = 1.066612\n",
      "Model best validation rmse = 1.066612 on epoch                     1\n",
      " 40%|████      | 2/5 [00:08<00:12,  4.08s/it]Epoch 2\n",
      "Loss = 7.3999e-01, PNorm = 34.1535, GNorm = 0.4346, lr_0 = 8.9198e-04\n",
      "Loss = 7.9198e-01, PNorm = 34.1983, GNorm = 5.6844, lr_0 = 7.5759e-04\n",
      "Loss = 7.3291e-01, PNorm = 34.2473, GNorm = 0.6201, lr_0 = 6.4344e-04\n",
      "Loss = 7.7808e-01, PNorm = 34.2874, GNorm = 2.7372, lr_0 = 5.4650e-04\n",
      "Loss = 8.9753e-01, PNorm = 34.3210, GNorm = 2.4806, lr_0 = 4.6416e-04\n",
      "Validation rmse = 1.013272\n",
      "Model best validation rmse = 1.013272 on epoch                     2\n",
      " 60%|██████    | 3/5 [00:12<00:08,  4.13s/it]Epoch 3\n",
      "Loss = 8.0214e-01, PNorm = 34.3530, GNorm = 1.0864, lr_0 = 3.9423e-04\n",
      "Loss = 7.2100e-01, PNorm = 34.3792, GNorm = 5.2348, lr_0 = 3.3483e-04\n",
      "Loss = 7.2450e-01, PNorm = 34.4022, GNorm = 1.3032, lr_0 = 2.8438e-04\n",
      "Loss = 6.7477e-01, PNorm = 34.4189, GNorm = 0.6692, lr_0 = 2.4153e-04\n",
      "Validation rmse = 0.984997\n",
      "Model best validation rmse = 0.984997 on epoch                     3\n",
      " 80%|████████  | 4/5 [00:16<00:03,  3.91s/it]Epoch 4\n",
      "Loss = 6.5337e-01, PNorm = 34.4337, GNorm = 1.4941, lr_0 = 2.0514e-04\n",
      "Loss = 7.2274e-01, PNorm = 34.4473, GNorm = 4.5168, lr_0 = 1.7423e-04\n",
      "Loss = 6.9917e-01, PNorm = 34.4585, GNorm = 1.9005, lr_0 = 1.4798e-04\n",
      "Loss = 6.9283e-01, PNorm = 34.4689, GNorm = 1.4910, lr_0 = 1.2569e-04\n",
      "Loss = 6.0776e-01, PNorm = 34.4784, GNorm = 2.3038, lr_0 = 1.0675e-04\n",
      "Validation rmse = 0.970690\n",
      "Model best validation rmse = 0.970690 on epoch                     4\n",
      "100%|██████████| 5/5 [00:19<00:00,  3.95s/it]\n",
      "train size = 2,350 | val size = 262\n",
      "Total size = 2,612\n",
      "Fitting scaler\n",
      "Number of parameters = 355,201\n",
      "Moving model to cuda\n",
      "  0%|          | 0/5 [00:00<?, ?it/s]Epoch 0\n",
      "Loss = 1.0159e+00, PNorm = 33.9637, GNorm = 5.4536, lr_0 = 2.0532e-04\n",
      "Loss = 9.6508e-01, PNorm = 33.9656, GNorm = 2.0029, lr_0 = 3.0106e-04\n",
      "Loss = 9.2271e-01, PNorm = 33.9719, GNorm = 1.0822, lr_0 = 3.9681e-04\n",
      "Loss = 1.0040e+00, PNorm = 33.9812, GNorm = 9.5395, lr_0 = 4.9255e-04\n",
      "Validation rmse = 1.219072\n",
      "Model best validation rmse = 1.219072 on epoch                     0\n",
      " 20%|██        | 1/5 [00:03<00:15,  3.75s/it]Epoch 1\n",
      "Loss = 9.1799e-01, PNorm = 33.9910, GNorm = 4.0710, lr_0 = 5.8830e-04\n",
      "Loss = 9.8140e-01, PNorm = 34.0062, GNorm = 1.1808, lr_0 = 6.8404e-04\n",
      "Loss = 7.7771e-01, PNorm = 34.0314, GNorm = 2.9937, lr_0 = 7.7979e-04\n",
      "Loss = 9.9337e-01, PNorm = 34.0557, GNorm = 0.9449, lr_0 = 8.7553e-04\n",
      "Loss = 8.7814e-01, PNorm = 34.1022, GNorm = 1.8016, lr_0 = 9.7128e-04\n",
      "Validation rmse = 1.060404\n",
      "Model best validation rmse = 1.060404 on epoch                     1\n",
      " 40%|████      | 2/5 [00:07<00:11,  3.75s/it]Epoch 2\n",
      "Loss = 8.7579e-01, PNorm = 34.1555, GNorm = 0.6882, lr_0 = 8.9198e-04\n",
      "Loss = 7.9147e-01, PNorm = 34.2025, GNorm = 4.0001, lr_0 = 7.5759e-04\n",
      "Loss = 9.1854e-01, PNorm = 34.2502, GNorm = 1.1780, lr_0 = 6.4344e-04\n",
      "Loss = 8.7270e-01, PNorm = 34.2919, GNorm = 3.2980, lr_0 = 5.4650e-04\n",
      "Loss = 7.4622e-01, PNorm = 34.3230, GNorm = 1.2139, lr_0 = 4.6416e-04\n",
      "Validation rmse = 1.028341\n",
      "Model best validation rmse = 1.028341 on epoch                     2\n",
      " 60%|██████    | 3/5 [00:11<00:07,  3.71s/it]Epoch 3\n",
      "Loss = 8.6318e-01, PNorm = 34.3555, GNorm = 3.3662, lr_0 = 3.9423e-04\n",
      "Loss = 8.0665e-01, PNorm = 34.3837, GNorm = 2.2241, lr_0 = 3.3483e-04\n",
      "Loss = 7.4221e-01, PNorm = 34.4080, GNorm = 2.6576, lr_0 = 2.8438e-04\n",
      "Loss = 7.2206e-01, PNorm = 34.4293, GNorm = 1.9117, lr_0 = 2.4153e-04\n",
      "Validation rmse = 0.991911\n",
      "Model best validation rmse = 0.991911 on epoch                     3\n",
      " 80%|████████  | 4/5 [00:15<00:03,  3.76s/it]Epoch 4\n",
      "Loss = 5.3762e-01, PNorm = 34.4453, GNorm = 1.6824, lr_0 = 2.0514e-04\n",
      "Loss = 7.4128e-01, PNorm = 34.4595, GNorm = 3.5782, lr_0 = 1.7423e-04\n",
      "Loss = 6.6491e-01, PNorm = 34.4722, GNorm = 1.5103, lr_0 = 1.4798e-04\n",
      "Loss = 7.1596e-01, PNorm = 34.4831, GNorm = 2.4865, lr_0 = 1.2569e-04\n",
      "Loss = 7.5453e-01, PNorm = 34.4920, GNorm = 3.0035, lr_0 = 1.0675e-04\n",
      "Validation rmse = 0.982648\n",
      "Model best validation rmse = 0.982648 on epoch                     4\n",
      "100%|██████████| 5/5 [00:19<00:00,  3.85s/it]\n",
      "train size = 2,350 | val size = 262\n",
      "Total size = 2,612\n",
      "Fitting scaler\n",
      "Number of parameters = 355,201\n",
      "Moving model to cuda\n",
      "  0%|          | 0/5 [00:00<?, ?it/s]Epoch 0\n",
      "Loss = 9.6161e-01, PNorm = 33.9641, GNorm = 1.6660, lr_0 = 2.0532e-04\n",
      "Loss = 9.6434e-01, PNorm = 33.9682, GNorm = 2.2360, lr_0 = 3.0106e-04\n",
      "Loss = 9.1577e-01, PNorm = 33.9766, GNorm = 1.2693, lr_0 = 3.9681e-04\n",
      "Loss = 8.7676e-01, PNorm = 33.9862, GNorm = 2.5774, lr_0 = 4.9255e-04\n",
      "Validation rmse = 1.108638\n",
      "Model best validation rmse = 1.108638 on epoch                     0\n",
      " 20%|██        | 1/5 [00:04<00:17,  4.49s/it]Epoch 1\n",
      "Loss = 8.4993e-01, PNorm = 34.0001, GNorm = 0.8386, lr_0 = 5.8830e-04\n",
      "Loss = 9.8678e-01, PNorm = 34.0131, GNorm = 2.7058, lr_0 = 6.8404e-04\n",
      "Loss = 8.8839e-01, PNorm = 34.0378, GNorm = 1.1372, lr_0 = 7.7979e-04\n",
      "Loss = 7.7432e-01, PNorm = 34.0686, GNorm = 1.2038, lr_0 = 8.7553e-04\n",
      "Loss = 1.1474e+00, PNorm = 34.1035, GNorm = 3.3968, lr_0 = 9.7128e-04\n",
      "Validation rmse = 1.177985\n",
      "Model best validation rmse = 1.108638 on epoch                     0\n",
      " 40%|████      | 2/5 [00:08<00:12,  4.09s/it]Epoch 2\n",
      "Loss = 8.3051e-01, PNorm = 34.1485, GNorm = 1.0922, lr_0 = 8.9198e-04\n",
      "Loss = 8.9588e-01, PNorm = 34.1976, GNorm = 5.2871, lr_0 = 7.5759e-04\n",
      "Loss = 9.2366e-01, PNorm = 34.2279, GNorm = 3.3864, lr_0 = 6.4344e-04\n",
      "Loss = 8.5504e-01, PNorm = 34.2606, GNorm = 2.3551, lr_0 = 5.4650e-04\n",
      "Loss = 7.7818e-01, PNorm = 34.2963, GNorm = 0.9439, lr_0 = 4.6416e-04\n",
      "Validation rmse = 1.037084\n",
      "Model best validation rmse = 1.037084 on epoch                     2\n",
      " 60%|██████    | 3/5 [00:12<00:07,  3.94s/it]Epoch 3\n",
      "Loss = 7.9421e-01, PNorm = 34.3247, GNorm = 4.1104, lr_0 = 3.9423e-04\n",
      "Loss = 7.8451e-01, PNorm = 34.3525, GNorm = 1.2666, lr_0 = 3.3483e-04\n",
      "Loss = 6.9671e-01, PNorm = 34.3715, GNorm = 2.9573, lr_0 = 2.8438e-04\n",
      "Loss = 7.4009e-01, PNorm = 34.3886, GNorm = 0.6557, lr_0 = 2.4153e-04\n",
      "Validation rmse = 0.981765\n",
      "Model best validation rmse = 0.981765 on epoch                     3\n",
      " 80%|████████  | 4/5 [00:15<00:03,  3.89s/it]Epoch 4\n",
      "Loss = 7.7988e-01, PNorm = 34.4067, GNorm = 4.3733, lr_0 = 2.0514e-04\n",
      "Loss = 7.9419e-01, PNorm = 34.4241, GNorm = 1.3760, lr_0 = 1.7423e-04\n",
      "Loss = 7.1167e-01, PNorm = 34.4357, GNorm = 2.8875, lr_0 = 1.4798e-04\n",
      "Loss = 7.0467e-01, PNorm = 34.4454, GNorm = 3.5834, lr_0 = 1.2569e-04\n",
      "Loss = 6.8785e-01, PNorm = 34.4530, GNorm = 2.9794, lr_0 = 1.0675e-04\n",
      "Validation rmse = 0.975451\n",
      "Model best validation rmse = 0.975451 on epoch                     4\n",
      "100%|██████████| 5/5 [00:19<00:00,  3.92s/it]\n",
      "train size = 2,350 | val size = 262\n",
      "Total size = 2,612\n",
      "Fitting scaler\n",
      "Number of parameters = 355,201\n",
      "Moving model to cuda\n",
      "  0%|          | 0/5 [00:00<?, ?it/s]Epoch 0\n",
      "Loss = 9.9853e-01, PNorm = 33.9638, GNorm = 0.7330, lr_0 = 2.0532e-04\n",
      "Loss = 9.0244e-01, PNorm = 33.9673, GNorm = 1.3094, lr_0 = 3.0106e-04\n",
      "Loss = 9.1198e-01, PNorm = 33.9744, GNorm = 7.3457, lr_0 = 3.9681e-04\n",
      "Loss = 8.4574e-01, PNorm = 33.9827, GNorm = 0.6453, lr_0 = 4.9255e-04\n",
      "Validation rmse = 1.126597\n",
      "Model best validation rmse = 1.126597 on epoch                     0\n",
      " 20%|██        | 1/5 [00:03<00:14,  3.68s/it]Epoch 1\n",
      "Loss = 7.2963e-01, PNorm = 33.9978, GNorm = 6.8186, lr_0 = 5.8830e-04\n",
      "Loss = 8.5872e-01, PNorm = 34.0202, GNorm = 4.4879, lr_0 = 6.8404e-04\n",
      "Loss = 9.3638e-01, PNorm = 34.0516, GNorm = 9.6551, lr_0 = 7.7979e-04\n",
      "Loss = 8.3926e-01, PNorm = 34.0857, GNorm = 1.9290, lr_0 = 8.7553e-04\n",
      "Loss = 8.7462e-01, PNorm = 34.1292, GNorm = 5.5581, lr_0 = 9.7128e-04\n",
      "Validation rmse = 1.097221\n",
      "Model best validation rmse = 1.097221 on epoch                     1\n",
      " 40%|████      | 2/5 [00:07<00:11,  3.72s/it]Epoch 2\n",
      "Loss = 7.0086e-01, PNorm = 34.1800, GNorm = 1.1642, lr_0 = 8.9198e-04\n",
      "Loss = 8.1189e-01, PNorm = 34.2347, GNorm = 1.9810, lr_0 = 7.5759e-04\n",
      "Loss = 8.2363e-01, PNorm = 34.2844, GNorm = 1.2317, lr_0 = 6.4344e-04\n",
      "Loss = 7.0713e-01, PNorm = 34.3177, GNorm = 2.4536, lr_0 = 5.4650e-04\n",
      "Loss = 7.7721e-01, PNorm = 34.3508, GNorm = 1.6281, lr_0 = 4.6416e-04\n",
      "Validation rmse = 1.073977\n",
      "Model best validation rmse = 1.073977 on epoch                     2\n",
      " 60%|██████    | 3/5 [00:11<00:07,  3.99s/it]Epoch 3\n",
      "Loss = 7.5941e-01, PNorm = 34.3801, GNorm = 3.1207, lr_0 = 3.9423e-04\n",
      "Loss = 7.2962e-01, PNorm = 34.4056, GNorm = 3.0835, lr_0 = 3.3483e-04\n",
      "Loss = 7.1125e-01, PNorm = 34.4243, GNorm = 0.9759, lr_0 = 2.8438e-04\n",
      "Loss = 7.5164e-01, PNorm = 34.4441, GNorm = 0.8565, lr_0 = 2.4153e-04\n",
      "Validation rmse = 1.025282\n",
      "Model best validation rmse = 1.025282 on epoch                     3\n",
      " 80%|████████  | 4/5 [00:15<00:03,  3.91s/it]Epoch 4\n",
      "Loss = 6.7920e-01, PNorm = 34.4617, GNorm = 1.9947, lr_0 = 2.0514e-04\n",
      "Loss = 6.7991e-01, PNorm = 34.4746, GNorm = 1.0946, lr_0 = 1.7423e-04\n",
      "Loss = 6.8062e-01, PNorm = 34.4850, GNorm = 4.5501, lr_0 = 1.4798e-04\n",
      "Loss = 6.0686e-01, PNorm = 34.4954, GNorm = 0.9835, lr_0 = 1.2569e-04\n",
      "Loss = 6.7101e-01, PNorm = 34.5059, GNorm = 3.1859, lr_0 = 1.0675e-04\n",
      "Validation rmse = 0.977678\n",
      "Model best validation rmse = 0.977678 on epoch                     4\n",
      "100%|██████████| 5/5 [00:19<00:00,  3.84s/it]\n",
      "train size = 2,938 | val size = 327\n",
      "Total size = 3,265\n",
      "Fitting scaler\n",
      "Number of parameters = 355,201\n",
      "Moving model to cuda\n",
      "  0%|          | 0/5 [00:00<?, ?it/s]Epoch 0\n",
      "Loss = 1.0170e+00, PNorm = 33.9636, GNorm = 11.4309, lr_0 = 1.8534e-04\n",
      "Loss = 8.9154e-01, PNorm = 33.9655, GNorm = 2.1009, lr_0 = 2.6293e-04\n",
      "Loss = 9.9012e-01, PNorm = 33.9699, GNorm = 5.2935, lr_0 = 3.4052e-04\n",
      "Loss = 1.1180e+00, PNorm = 33.9757, GNorm = 7.3783, lr_0 = 4.1810e-04\n",
      "Loss = 9.7764e-01, PNorm = 33.9835, GNorm = 1.6488, lr_0 = 4.9569e-04\n",
      "Validation rmse = 1.021993\n",
      "Model best validation rmse = 1.021993 on epoch                     0\n",
      " 20%|██        | 1/5 [00:04<00:19,  4.80s/it]Epoch 1\n",
      "Loss = 7.5355e-01, PNorm = 34.0035, GNorm = 2.5964, lr_0 = 5.8103e-04\n",
      "Loss = 8.5921e-01, PNorm = 34.0240, GNorm = 4.7273, lr_0 = 6.5862e-04\n",
      "Loss = 9.1011e-01, PNorm = 34.0527, GNorm = 6.8103, lr_0 = 7.3621e-04\n",
      "Loss = 9.2334e-01, PNorm = 34.0865, GNorm = 1.4559, lr_0 = 8.1379e-04\n",
      "Loss = 8.8782e-01, PNorm = 34.1316, GNorm = 0.8523, lr_0 = 8.9138e-04\n",
      "Loss = 8.6712e-01, PNorm = 34.1926, GNorm = 4.2133, lr_0 = 9.6897e-04\n",
      "Validation rmse = 0.933249\n",
      "Model best validation rmse = 0.933249 on epoch                     1\n",
      " 40%|████      | 2/5 [00:09<00:14,  4.78s/it]Epoch 2\n",
      "Loss = 8.9621e-01, PNorm = 34.2660, GNorm = 3.3422, lr_0 = 9.2367e-04\n",
      "Loss = 8.1042e-01, PNorm = 34.3290, GNorm = 4.6028, lr_0 = 8.0918e-04\n",
      "Loss = 6.9660e-01, PNorm = 34.3812, GNorm = 1.2373, lr_0 = 7.0888e-04\n",
      "Loss = 8.4112e-01, PNorm = 34.4247, GNorm = 3.4195, lr_0 = 6.2102e-04\n",
      "Loss = 7.4547e-01, PNorm = 34.4695, GNorm = 2.2296, lr_0 = 5.4404e-04\n",
      "Loss = 7.6372e-01, PNorm = 34.5140, GNorm = 3.5317, lr_0 = 4.7661e-04\n",
      "Validation rmse = 0.905393\n",
      "Model best validation rmse = 0.905393 on epoch                     2\n",
      " 60%|██████    | 3/5 [00:14<00:09,  4.79s/it]Epoch 3\n",
      "Loss = 9.1666e-01, PNorm = 34.5557, GNorm = 1.8569, lr_0 = 4.1753e-04\n",
      "Loss = 6.0185e-01, PNorm = 34.5969, GNorm = 0.9774, lr_0 = 3.6578e-04\n",
      "Loss = 6.7341e-01, PNorm = 34.6261, GNorm = 2.3029, lr_0 = 3.2044e-04\n",
      "Loss = 6.5004e-01, PNorm = 34.6498, GNorm = 6.2139, lr_0 = 2.8072e-04\n",
      "Loss = 6.4573e-01, PNorm = 34.6664, GNorm = 1.7850, lr_0 = 2.4593e-04\n",
      "Loss = 6.8536e-01, PNorm = 34.6824, GNorm = 1.1269, lr_0 = 2.1544e-04\n",
      "Validation rmse = 0.860619\n",
      "Model best validation rmse = 0.860619 on epoch                     3\n",
      " 80%|████████  | 4/5 [00:19<00:04,  4.73s/it]Epoch 4\n",
      "Loss = 6.7648e-01, PNorm = 34.7008, GNorm = 5.0548, lr_0 = 1.8874e-04\n",
      "Loss = 7.0349e-01, PNorm = 34.7170, GNorm = 5.6085, lr_0 = 1.6534e-04\n",
      "Loss = 5.8797e-01, PNorm = 34.7339, GNorm = 1.0818, lr_0 = 1.4485e-04\n",
      "Loss = 6.9574e-01, PNorm = 34.7465, GNorm = 1.0938, lr_0 = 1.2690e-04\n",
      "Loss = 5.5851e-01, PNorm = 34.7565, GNorm = 4.3766, lr_0 = 1.1117e-04\n",
      "Loss = 5.7325e-01, PNorm = 34.7642, GNorm = 3.3861, lr_0 = 1.0000e-04\n",
      "Validation rmse = 0.820985\n",
      "Model best validation rmse = 0.820985 on epoch                     4\n",
      "100%|██████████| 5/5 [00:24<00:00,  4.83s/it]\n",
      "Total size = 4,082\n",
      "Fitting scaler\n",
      "Number of parameters = 355,201\n",
      "Moving model to cuda\n",
      "  0%|          | 0/5 [00:00<?, ?it/s]Epoch 0\n",
      "Loss = 9.4430e-01, PNorm = 33.9636, GNorm = 4.8548, lr_0 = 1.6111e-04\n",
      "Loss = 1.0211e+00, PNorm = 33.9653, GNorm = 4.1721, lr_0 = 2.1667e-04\n",
      "Loss = 9.6043e-01, PNorm = 33.9695, GNorm = 0.6979, lr_0 = 2.7222e-04\n",
      "Loss = 8.4431e-01, PNorm = 33.9765, GNorm = 1.1135, lr_0 = 3.2778e-04\n",
      "Loss = 9.5611e-01, PNorm = 33.9833, GNorm = 11.0795, lr_0 = 3.8333e-04\n",
      "Loss = 8.6006e-01, PNorm = 33.9941, GNorm = 3.5641, lr_0 = 4.3889e-04\n",
      "Loss = 8.9175e-01, PNorm = 34.0097, GNorm = 8.6917, lr_0 = 4.9444e-04\n",
      "Loss = 9.0584e-01, PNorm = 34.0293, GNorm = 7.4949, lr_0 = 5.5000e-04\n",
      " 20%|██        | 1/5 [00:07<00:29,  7.26s/it]Epoch 1\n",
      "Loss = 8.6427e-01, PNorm = 34.0564, GNorm = 12.6435, lr_0 = 6.1111e-04\n",
      "Loss = 8.6439e-01, PNorm = 34.0854, GNorm = 2.9525, lr_0 = 6.6667e-04\n",
      "Loss = 8.7216e-01, PNorm = 34.1201, GNorm = 0.6086, lr_0 = 7.2222e-04\n",
      "Loss = 8.5656e-01, PNorm = 34.1564, GNorm = 7.9748, lr_0 = 7.7778e-04\n",
      "Loss = 8.6763e-01, PNorm = 34.2061, GNorm = 1.3233, lr_0 = 8.3333e-04\n",
      "Loss = 8.2510e-01, PNorm = 34.2576, GNorm = 1.2597, lr_0 = 8.8889e-04\n",
      "Loss = 8.5615e-01, PNorm = 34.3283, GNorm = 1.8525, lr_0 = 9.4444e-04\n",
      "Loss = 7.7722e-01, PNorm = 34.3971, GNorm = 6.2036, lr_0 = 1.0000e-03\n",
      " 40%|████      | 2/5 [00:13<00:20,  6.90s/it]Epoch 2\n",
      "Loss = 8.2419e-01, PNorm = 34.4623, GNorm = 1.2522, lr_0 = 9.0959e-04\n",
      "Loss = 6.8186e-01, PNorm = 34.5338, GNorm = 3.2453, lr_0 = 8.2736e-04\n",
      "Loss = 7.5162e-01, PNorm = 34.5945, GNorm = 4.5183, lr_0 = 7.5256e-04\n",
      "Loss = 7.4107e-01, PNorm = 34.6451, GNorm = 4.3620, lr_0 = 6.8453e-04\n",
      "Loss = 6.9100e-01, PNorm = 34.6894, GNorm = 6.9087, lr_0 = 6.2264e-04\n",
      "Loss = 7.0512e-01, PNorm = 34.7340, GNorm = 2.4326, lr_0 = 5.6635e-04\n",
      "Loss = 7.0569e-01, PNorm = 34.7760, GNorm = 0.9125, lr_0 = 5.1515e-04\n",
      "Loss = 6.1139e-01, PNorm = 34.8094, GNorm = 1.5326, lr_0 = 4.6858e-04\n",
      " 60%|██████    | 3/5 [00:19<00:12,  6.45s/it]Epoch 3\n",
      "Loss = 6.5085e-01, PNorm = 34.8537, GNorm = 3.0652, lr_0 = 4.2220e-04\n",
      "Loss = 5.8856e-01, PNorm = 34.8901, GNorm = 6.8293, lr_0 = 3.8403e-04\n",
      "Loss = 6.4166e-01, PNorm = 34.9201, GNorm = 2.8731, lr_0 = 3.4931e-04\n",
      "Loss = 6.3637e-01, PNorm = 34.9453, GNorm = 3.0439, lr_0 = 3.1773e-04\n",
      "Loss = 5.7868e-01, PNorm = 34.9627, GNorm = 4.8165, lr_0 = 2.8900e-04\n",
      "Loss = 6.6469e-01, PNorm = 34.9863, GNorm = 1.0559, lr_0 = 2.6288e-04\n",
      "Loss = 5.8377e-01, PNorm = 35.0073, GNorm = 1.3895, lr_0 = 2.3911e-04\n",
      "Loss = 5.7520e-01, PNorm = 35.0224, GNorm = 4.6767, lr_0 = 2.1749e-04\n",
      " 80%|████████  | 4/5 [00:25<00:06,  6.25s/it]Epoch 4\n",
      "Loss = 6.2194e-01, PNorm = 35.0325, GNorm = 3.9125, lr_0 = 1.9783e-04\n",
      "Loss = 5.7686e-01, PNorm = 35.0459, GNorm = 1.7605, lr_0 = 1.7995e-04\n",
      "Loss = 5.8290e-01, PNorm = 35.0615, GNorm = 2.2839, lr_0 = 1.6368e-04\n",
      "Loss = 5.2093e-01, PNorm = 35.0742, GNorm = 1.8857, lr_0 = 1.4888e-04\n",
      "Loss = 5.2086e-01, PNorm = 35.0838, GNorm = 3.4072, lr_0 = 1.3542e-04\n",
      "Loss = 6.0942e-01, PNorm = 35.0936, GNorm = 2.0931, lr_0 = 1.2318e-04\n",
      "Loss = 6.1745e-01, PNorm = 35.1027, GNorm = 8.4424, lr_0 = 1.1204e-04\n",
      "Loss = 5.0599e-01, PNorm = 35.1089, GNorm = 1.0934, lr_0 = 1.0191e-04\n",
      "100%|██████████| 5/5 [00:31<00:00,  6.35s/it]\n",
      "qsprpred - WARNING - Early stopping did not yield a best model, using last model instead.\n"
     ]
    }
   ],
   "source": [
    "from qsprpred.models import CrossValAssessor, TestSetAssessor\n",
    "from qsprpred.models.early_stopping import EarlyStoppingMode\n",
    "\n",
    "# We can now assess the model performance on the training set using cross validation\n",
    "CrossValAssessor('r2')(model, dataset)\n",
    "\n",
    "# and on the test set\n",
    "TestSetAssessor('r2')(model, dataset)\n",
    "\n",
    "# Finally, we need to fit the model on the complete dataset if we want to use it further\n",
    "# This will save the fitted model and metadata to disk\n",
    "model.earlyStopping.numEpochs = 5\n",
    "model.fitDataset(dataset, mode=EarlyStoppingMode.FIXED)\n",
    "\n",
    "# We can optionally save the model and metadata to disk explicitly as well\n",
    "_ = model.save()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ChempropTutorialModel\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 569.5x400 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from qsprpred.plotting.regression import CorrelationPlot\n",
    "\n",
    "plot = CorrelationPlot([model])\n",
    "axes, summary = plot.make(save=False, show=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Model</th>\n",
       "      <th>Fold</th>\n",
       "      <th>Property</th>\n",
       "      <th>R2</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>Set</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>ChempropTutorialModel</td>\n",
       "      <td>0.0</td>\n",
       "      <td>pchembl_value_Mean</td>\n",
       "      <td>0.364764</td>\n",
       "      <td>0.923366</td>\n",
       "      <td>Cross Validation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>ChempropTutorialModel</td>\n",
       "      <td>1.0</td>\n",
       "      <td>pchembl_value_Mean</td>\n",
       "      <td>0.274390</td>\n",
       "      <td>1.001930</td>\n",
       "      <td>Cross Validation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>ChempropTutorialModel</td>\n",
       "      <td>2.0</td>\n",
       "      <td>pchembl_value_Mean</td>\n",
       "      <td>0.319186</td>\n",
       "      <td>0.946741</td>\n",
       "      <td>Cross Validation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>ChempropTutorialModel</td>\n",
       "      <td>3.0</td>\n",
       "      <td>pchembl_value_Mean</td>\n",
       "      <td>0.278532</td>\n",
       "      <td>0.965912</td>\n",
       "      <td>Cross Validation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ChempropTutorialModel</td>\n",
       "      <td>4.0</td>\n",
       "      <td>pchembl_value_Mean</td>\n",
       "      <td>0.233830</td>\n",
       "      <td>1.025186</td>\n",
       "      <td>Cross Validation</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>ChempropTutorialModel</td>\n",
       "      <td>Independent Test</td>\n",
       "      <td>pchembl_value_Mean</td>\n",
       "      <td>0.362437</td>\n",
       "      <td>0.937360</td>\n",
       "      <td>Independent Test</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   Model              Fold            Property        R2  \\\n",
       "0  ChempropTutorialModel               0.0  pchembl_value_Mean  0.364764   \n",
       "1  ChempropTutorialModel               1.0  pchembl_value_Mean  0.274390   \n",
       "2  ChempropTutorialModel               2.0  pchembl_value_Mean  0.319186   \n",
       "3  ChempropTutorialModel               3.0  pchembl_value_Mean  0.278532   \n",
       "4  ChempropTutorialModel               4.0  pchembl_value_Mean  0.233830   \n",
       "5  ChempropTutorialModel  Independent Test  pchembl_value_Mean  0.362437   \n",
       "\n",
       "       RMSE               Set  \n",
       "0  0.923366  Cross Validation  \n",
       "1  1.001930  Cross Validation  \n",
       "2  0.946741  Cross Validation  \n",
       "3  0.965912  Cross Validation  \n",
       "4  1.025186  Cross Validation  \n",
       "5  0.937360  Independent Test  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "summary"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}