results/disease_gene_prioritization.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Disease gene prioritization\n",
    "\n",
    "## Contents\n",
    "* [Load gene and disease network.](#Load-gene-and-disease-network)\n",
    "* [Load the training data.](#Load-the-training-data.)\n",
    "* [Define-the-training process.](#Define-the-training-process.)\n",
    "* [Train the DGP model.](#Train-the-GDP-model.)\n",
    "* [Eval the results.](#Eval-the-results.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "import sys\n",
    "import logging\n",
    "import os\n",
    "import os.path as osp\n",
    "import torch\n",
    "import random\n",
    "import time\n",
    "import torch.nn.functional as F\n",
    "import pandas as pd\n",
    "from torch.nn import Linear\n",
    "import numpy as np\n",
    "from IPython.display import display, HTML\n",
    "from torch_geometric.nn import GCNConv, SAGEConv, GraphConv\n",
    "from sklearn.metrics import label_ranking_average_precision_score, label_ranking_loss, roc_auc_score\n",
    "from sklearn.metrics import roc_curve, precision_recall_curve, auc, average_precision_score\n",
    "from torch_geometric.data import InMemoryDataset, Data\n",
    "from sklearn.model_selection import KFold, StratifiedKFold, train_test_split\n",
    "from tqdm import tqdm\n",
    "import pickle\n",
    "import gzip\n",
    "from livelossplot import PlotLosses\n",
    "\n",
    "logging.basicConfig(stream=sys.stdout, level=logging.INFO)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "##### Expreiment hyperparameters\n",
    "\n",
    "#  NEGATIVE_SAMPLES determines how the negative examples for training are created.\n",
    "#  Choose from {'random', 'random_only_disease_gene'}\n",
    "#    * random: Choose a random (gene, disease) pair which is not in the positive set.\n",
    "#    * random_only_disease_genes: Like random but gene must be assigned to at least one disease.\n",
    "NEGATIVE_SAMPLES = 'random'\n",
    "EXPERIMENT_SLUG = 'final'\n",
    "device = torch.device('cuda')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load gene and disease network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Processing...\n",
      "INFO:root:Generating the gene edges.\n",
      "INFO:root:Creating the feature vectors.\n",
      "INFO:root:Creating the feature vectors from `gene_hpo_disease.tsv`\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 77035/77035 [00:00<00:00, 977025.54it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:root:Skipped TruncatedSVD\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:root:Storing the data.\n",
      "INFO:root:Done.\n",
      "Done!\n",
      "Processing...\n",
      "INFO:root:Create disease_id feature_index mapping.\n",
      "INFO:root:Create feature matrix.\n",
      "INFO:root:Creating disease feature vectors.\n",
      "INFO:root:Create publication feature vectors.\n",
      "INFO:root:Create edges.\n",
      "INFO:root:Generate similarity based edges.\n",
      "INFO:root:Storing the data.\n",
      "INFO:root:Done.\n",
      "Done!\n"
     ]
    }
   ],
   "source": [
    "# Load gene and disease network\n",
    "HERE = osp.abspath('')\n",
    "ROOT = osp.join(HERE, '..', '..')\n",
    "DATA_SOURCE_PATH = osp.join(ROOT, 'data_sources')\n",
    "GENE_DATASET_ROOT = osp.join(DATA_SOURCE_PATH, 'gene_net_fn_hpo')\n",
    "DISEASE_DATASET_ROOT = osp.join(DATA_SOURCE_PATH, 'disease_net_pubmed_knn')\n",
    "RESULTS_STORAGE = osp.join(HERE, 'results', EXPERIMENT_SLUG)\n",
    "MODEL_TMP_STORAGE = osp.join('/', 'var', 'tmp', 'dg_tmp')\n",
    "sys.path.insert(0, osp.abspath(ROOT))\n",
    "\n",
    "from GeneNet import GeneNet\n",
    "from DiseaseNet import DiseaseNet\n",
    "from TheModel import TheModel\n",
    "\n",
    "gene_dataset = GeneNet(\n",
    "    root=GENE_DATASET_ROOT,\n",
    "    humannet_version='FN',\n",
    "    features_to_use='hpo',\n",
    "    skip_truncated_svd=True\n",
    ")\n",
    "\n",
    "disease_dataset = DiseaseNet(\n",
    "    root=DISEASE_DATASET_ROOT,\n",
    "    hpo_count_freq_cutoff=40,\n",
    "    edge_source='feature_similarity',\n",
    "    feature_source='disease_publications',\n",
    "    skip_truncated_svd=True,\n",
    "    svd_components=2048,\n",
    "    svd_n_iter=12\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data(edge_attr=[371502, 1], edge_index=[2, 371502], x=[17247, 6209])\n",
      "Data(edge_attr=[282464], edge_index=[2, 282464], x=[8827, 11460])\n"
     ]
    }
   ],
   "source": [
    "gene_net_data = gene_dataset[0]\n",
    "disease_net_data = disease_dataset[0]\n",
    "\n",
    "# Print topology details.\n",
    "print(gene_net_data)\n",
    "print(disease_net_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "gene_net_data = gene_net_data.to(device)\n",
    "disease_net_data = disease_net_data.to(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the training data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Generate training data.\n",
    "disease_genes = pd.read_table(\n",
    "    osp.join(DATA_SOURCE_PATH, 'genes_diseases.tsv'),\n",
    "    names=['EntrezGene ID', 'OMIM ID'],\n",
    "    sep='\\t', \n",
    "    low_memory=False, \n",
    "    dtype={'EntrezGene ID': pd.Int64Dtype()}\n",
    ")\n",
    "\n",
    "disease_id_index_feature_mapping = disease_dataset.load_disease_index_feature_mapping()\n",
    "gene_id_index_feature_mapping = gene_dataset.load_node_index_mapping()\n",
    "\n",
    "all_genes = list(gene_id_index_feature_mapping.keys())\n",
    "all_diseases = list(disease_id_index_feature_mapping.keys())\n",
    "\n",
    "# 1. generate positive pairs.\n",
    "# Filter the pairs to only include the ones where the corresponding nodes are available.\n",
    "# i.e. gene_id should be in all_genes and disease_id should be in all_diseases.\n",
    "positives = disease_genes[\n",
    "    disease_genes[\"OMIM ID\"].isin(all_diseases) & disease_genes[\"EntrezGene ID\"].isin(all_genes)\n",
    "]\n",
    "covered_diseases = list(set(positives['OMIM ID']))\n",
    "covered_genes = list(set(positives['EntrezGene ID']))\n",
    "\n",
    "# 2. Generate negatives.\n",
    "# Pick equal amount of pairs not in the positives.\n",
    "negatives_list = []\n",
    "while len(negatives_list) < len(positives):\n",
    "    if NEGATIVE_SAMPLES == 'random_only_disease_genes' :\n",
    "        gene_id = covered_genes[np.random.randint(0, len(covered_genes))]\n",
    "    else:\n",
    "        gene_id = all_genes[np.random.randint(0, len(all_genes))]\n",
    "    disease_id = covered_diseases[np.random.randint(0, len(covered_diseases))]\n",
    "    if not ((positives['OMIM ID'] == disease_id) & (positives['EntrezGene ID'] == gene_id)).any():\n",
    "        negatives_list.append([disease_id, gene_id])\n",
    "negatives = pd.DataFrame(np.array(negatives_list), columns=['OMIM ID', 'EntrezGene ID'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Define the training process."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_training_data_from_indexes(indexes, monogenetic_disease_only=False, multigenetic_diseases_only=False):\n",
    "    train_tuples = set()\n",
    "    for idx in indexes:\n",
    "        pos = positives[positives['OMIM ID'] == covered_diseases[idx]]\n",
    "        neg = negatives[negatives['OMIM ID'] == covered_diseases[idx]]\n",
    "        if monogenetic_disease_only and len(pos) != 1:\n",
    "            continue\n",
    "        if multigenetic_diseases_only and len(pos) == 1:\n",
    "            continue\n",
    "        for index, row in pos.iterrows():\n",
    "            train_tuples.add((row['OMIM ID'], row['EntrezGene ID'], 1))\n",
    "        for index, row in neg.iterrows():\n",
    "            train_tuples.add((row['OMIM ID'], row['EntrezGene ID'], 0))\n",
    "    ## 2. Concat data.\n",
    "    n = len(train_tuples)\n",
    "    x_out = np.ones((n, 2))  # will contain (gene_idx, disease_idx) tuples.\n",
    "    y_out = torch.ones((n,), dtype=torch.long)\n",
    "    for i, (omim_id, gene_id, y) in enumerate(train_tuples):\n",
    "        x_out[i] = (gene_id_index_feature_mapping[int(gene_id)], disease_id_index_feature_mapping[omim_id])\n",
    "        y_out[i] = y\n",
    "    return x_out, y_out\n",
    "\n",
    "def train(\n",
    "    max_epochs, \n",
    "    early_stopping_window=5, \n",
    "    info_each_epoch=1, \n",
    "    folds=5, \n",
    "    lr=0.0005, \n",
    "    weight_decay=5e-4,\n",
    "    fc_hidden_dim=2048,\n",
    "    gene_net_hidden_dim=512,\n",
    "    disease_net_hidden_dim=512\n",
    "):\n",
    "    criterion = torch.nn.CrossEntropyLoss(weight=torch.tensor([0.2, 0.8]).to(device))\n",
    "    metrics = []\n",
    "    dis_dict = {}\n",
    "    fold = 0\n",
    "    start_time = time.time()\n",
    "    liveplot = PlotLosses(groups={'training': ['loss', 'val']})\n",
    "    kf = KFold(n_splits=folds, shuffle=True, random_state=42)\n",
    "    for train_index, test_index in kf.split(covered_diseases):\n",
    "        fold += 1\n",
    "        print(f'Generate training data for fold {fold}.')\n",
    "        all_train_x, all_train_y = get_training_data_from_indexes(train_index)\n",
    "\n",
    "        # Split into train and validation set.\n",
    "        id_tr, id_val = train_test_split(range(len(all_train_x)), test_size=0.1, random_state=42)\n",
    "        train_x = all_train_x[id_tr]\n",
    "        train_y = all_train_y[id_tr].to(device)\n",
    "        val_x = all_train_x[id_val]\n",
    "        val_y = all_train_y[id_val].to(device)\n",
    "\n",
    "        # Generate the test data for mono and multigenetic diseases.\n",
    "        ## 1. Collect data.\n",
    "        print(f'Generate test data for fold {fold}.')\n",
    "        test_x = dict()\n",
    "        test_y = dict()\n",
    "        test_x['mono'], test_y['mono'] = get_training_data_from_indexes(test_index, monogenetic_disease_only=True)\n",
    "        test_y['mono'] =  test_y['mono'].to(device)\n",
    "        test_x['multi'], test_y['multi'] = get_training_data_from_indexes(test_index, multigenetic_diseases_only=True)\n",
    "        test_y['multi'] = test_y['multi'].to(device)\n",
    "\n",
    "        # Create the model\n",
    "        model = TheModel(\n",
    "            gene_feature_dim=gene_net_data.x.shape[1],\n",
    "            disease_feature_dim=disease_net_data.x.shape[1],\n",
    "            fc_hidden_dim=fc_hidden_dim,\n",
    "            gene_net_hidden_dim=gene_net_hidden_dim,\n",
    "            disease_net_hidden_dim=disease_net_hidden_dim,\n",
    "            mode='DGP'\n",
    "        ).to(device)\n",
    "        optimizer = torch.optim.Adam(model.parameters(), lr=lr, weight_decay=weight_decay)\n",
    "        print(f'Stat training fold {fold}/{folds}:')\n",
    "        \n",
    "        losses = dict()\n",
    "        losses['train'] = list()\n",
    "        losses['val'] = list()\n",
    "        \n",
    "        losses['mono'] = {\n",
    "                'AUC': 0,\n",
    "                'TPR': None,\n",
    "                'FPR': None\n",
    "        }\n",
    "        losses['multi'] = {\n",
    "                'AUC': 0,\n",
    "                'TPR': None,\n",
    "                'FPR': None\n",
    "        }\n",
    "        \n",
    "        best_val_loss = 1e80\n",
    "        for epoch in range(max_epochs):\n",
    "            # Train model.\n",
    "            model.train()\n",
    "            optimizer.zero_grad()\n",
    "            out = model(gene_net_data, disease_net_data, train_x)\n",
    "            loss = criterion(out, train_y)\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            losses['train'].append(loss.item())\n",
    "\n",
    "            # Validation.\n",
    "            with torch.no_grad():\n",
    "                model.eval()\n",
    "                out = model(gene_net_data, disease_net_data, val_x)\n",
    "                loss = criterion(out, val_y)\n",
    "                current_val_loss = loss.item()\n",
    "                losses['val'].append(current_val_loss)\n",
    "\n",
    "                if epoch % info_each_epoch == 0:\n",
    "                    print(\n",
    "                        'Epoch {}, train_loss: {:.4f}, val_loss: {:.4f}'.format(\n",
    "                            epoch, losses['train'][epoch], losses['val'][epoch]\n",
    "                        )\n",
    "                    )\n",
    "                if current_val_loss < best_val_loss:\n",
    "                    best_val_loss = current_val_loss\n",
    "                    torch.save(model.state_dict(), osp.join(MODEL_TMP_STORAGE, f'best_model_fold_{fold}.ptm'))\n",
    "            \n",
    "            # Early stopping\n",
    "            if epoch > early_stopping_window:\n",
    "                # Stop if validation error did not decrease \n",
    "                # w.r.t. the past early_stopping_window consecutive epochs.\n",
    "                last_window_losses = losses['val'][epoch - early_stopping_window:epoch]\n",
    "                if losses['val'][-1] > max(last_window_losses):\n",
    "                    print('Early Stopping!')\n",
    "                    break\n",
    "\n",
    "        # Test the model for the current fold.\n",
    "        model.load_state_dict(torch.load(osp.join(MODEL_TMP_STORAGE, f'best_model_fold_{fold}.ptm')))\n",
    "        with torch.no_grad():\n",
    "            for modus in ['multi', 'mono']:\n",
    "                predicted_probs = F.log_softmax(\n",
    "                    model(gene_net_data, disease_net_data, test_x[modus]).clone().detach(), dim=1\n",
    "                )\n",
    "                true_y = test_y[modus]\n",
    "                fpr, tpr, _ = roc_curve(true_y.cpu().detach().numpy(), predicted_probs[:,1].cpu().detach().numpy(), pos_label=1)\n",
    "                roc_auc = auc(fpr, tpr)\n",
    "                losses[modus]['TEST_Y'] = true_y.cpu().detach().numpy()\n",
    "                losses[modus]['TEST_PREDICT'] = predicted_probs.cpu().numpy()\n",
    "                losses[modus]['AUC'] = roc_auc\n",
    "                losses[modus]['TPR'] = tpr\n",
    "                losses[modus]['FPR'] = fpr\n",
    "                print(f'\"{modus}\" auc for fold: {fold}: {roc_auc}')            \n",
    "        metrics.append(losses)\n",
    "\n",
    "    print('Done!')\n",
    "    return metrics, dis_dict, model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define helpers for evaluation\n",
    "def negcum(rank_vec):\n",
    "    rank_vec_cum = []\n",
    "    prev = 0\n",
    "    for x in rank_vec:\n",
    "        if x == 0:\n",
    "            prev += 1\n",
    "            rank_vec_cum.append(prev)\n",
    "        else:\n",
    "            rank_vec_cum.append(prev)\n",
    "    rank_vec_cum = np.array(rank_vec_cum)\n",
    "    return rank_vec_cum\n",
    "\n",
    "disease_idx_to_omim_mapping = dict()\n",
    "for omim_id, disease_idx in disease_id_index_feature_mapping.items():\n",
    "    disease_idx_to_omim_mapping[disease_idx] = omim_id\n",
    "\n",
    "gene_idx_entrez_id_mapping = dict()\n",
    "for entrez_id, gene_idx in gene_id_index_feature_mapping.items():\n",
    "    gene_idx_entrez_id_mapping[gene_idx] = entrez_id\n",
    "\n",
    "def get_genes_assoc_to_omim_disease(omim_id):\n",
    "    return positives[positives[\"OMIM ID\"].isin([omim_id])][\"EntrezGene ID\"].values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train the GDP model.\n",
    "* Store trained models in `best_model_fold_{fold}.ptm`\n",
    "* Store results in `{EXPERIMENT_SLUG}_metrics.pickle.gz` and `{EXPERIMENT_SLUG}_dis_dict.pickle.gz`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generate training data for fold 1.\n",
      "Generate test data for fold 1.\n",
      "Stat training fold 1/5:\n",
      "Epoch 0, train_loss: 0.7132, val_loss: 0.6790\n",
      "Epoch 1, train_loss: 0.6892, val_loss: 0.6761\n",
      "Epoch 2, train_loss: 0.6698, val_loss: 0.6735\n",
      "Epoch 3, train_loss: 0.6533, val_loss: 0.6710\n",
      "Epoch 4, train_loss: 0.6382, val_loss: 0.6685\n",
      "Epoch 5, train_loss: 0.6254, val_loss: 0.6661\n",
      "Epoch 6, train_loss: 0.6138, val_loss: 0.6638\n",
      "Epoch 7, train_loss: 0.6029, val_loss: 0.6616\n",
      "Epoch 8, train_loss: 0.5933, val_loss: 0.6594\n",
      "Epoch 9, train_loss: 0.5856, val_loss: 0.6573\n",
      "Epoch 10, train_loss: 0.5777, val_loss: 0.6553\n",
      "Epoch 11, train_loss: 0.5710, val_loss: 0.6533\n",
      "Epoch 12, train_loss: 0.5645, val_loss: 0.6514\n",
      "Epoch 13, train_loss: 0.5580, val_loss: 0.6495\n",
      "Epoch 14, train_loss: 0.5536, val_loss: 0.6477\n",
      "Epoch 15, train_loss: 0.5483, val_loss: 0.6459\n",
      "Epoch 16, train_loss: 0.5432, val_loss: 0.6442\n",
      "Epoch 17, train_loss: 0.5397, val_loss: 0.6425\n",
      "Epoch 18, train_loss: 0.5349, val_loss: 0.6408\n",
      "Epoch 19, train_loss: 0.5320, val_loss: 0.6391\n",
      "Epoch 20, train_loss: 0.5286, val_loss: 0.6375\n",
      "Epoch 21, train_loss: 0.5260, val_loss: 0.6359\n",
      "Epoch 22, train_loss: 0.5229, val_loss: 0.6342\n",
      "Epoch 23, train_loss: 0.5202, val_loss: 0.6326\n",
      "Epoch 24, train_loss: 0.5176, val_loss: 0.6311\n",
      "Epoch 25, train_loss: 0.5152, val_loss: 0.6295\n",
      "Epoch 26, train_loss: 0.5136, val_loss: 0.6279\n",
      "Epoch 27, train_loss: 0.5110, val_loss: 0.6264\n",
      "Epoch 28, train_loss: 0.5089, val_loss: 0.6248\n",
      "Epoch 29, train_loss: 0.5078, val_loss: 0.6233\n",
      "Epoch 30, train_loss: 0.5060, val_loss: 0.6217\n",
      "Epoch 31, train_loss: 0.5046, val_loss: 0.6202\n",
      "Epoch 32, train_loss: 0.5030, val_loss: 0.6186\n",
      "Epoch 33, train_loss: 0.5018, val_loss: 0.6170\n",
      "Epoch 34, train_loss: 0.4995, val_loss: 0.6155\n",
      "Epoch 35, train_loss: 0.4986, val_loss: 0.6139\n",
      "Epoch 36, train_loss: 0.4974, val_loss: 0.6122\n",
      "Epoch 37, train_loss: 0.4959, val_loss: 0.6106\n",
      "Epoch 38, train_loss: 0.4944, val_loss: 0.6090\n",
      "Epoch 39, train_loss: 0.4931, val_loss: 0.6073\n",
      "Epoch 40, train_loss: 0.4921, val_loss: 0.6056\n",
      "Epoch 41, train_loss: 0.4905, val_loss: 0.6038\n",
      "Epoch 42, train_loss: 0.4899, val_loss: 0.6020\n",
      "Epoch 43, train_loss: 0.4878, val_loss: 0.6002\n",
      "Epoch 44, train_loss: 0.4866, val_loss: 0.5983\n",
      "Epoch 45, train_loss: 0.4853, val_loss: 0.5964\n",
      "Epoch 46, train_loss: 0.4840, val_loss: 0.5944\n",
      "Epoch 47, train_loss: 0.4830, val_loss: 0.5923\n",
      "Epoch 48, train_loss: 0.4815, val_loss: 0.5903\n",
      "Epoch 49, train_loss: 0.4796, val_loss: 0.5881\n",
      "Epoch 50, train_loss: 0.4787, val_loss: 0.5859\n",
      "Epoch 51, train_loss: 0.4771, val_loss: 0.5837\n",
      "Epoch 52, train_loss: 0.4760, val_loss: 0.5814\n",
      "Epoch 53, train_loss: 0.4742, val_loss: 0.5791\n",
      "Epoch 54, train_loss: 0.4735, val_loss: 0.5767\n",
      "Epoch 55, train_loss: 0.4720, val_loss: 0.5744\n",
      "Epoch 56, train_loss: 0.4702, val_loss: 0.5719\n",
      "Epoch 57, train_loss: 0.4691, val_loss: 0.5694\n",
      "Epoch 58, train_loss: 0.4672, val_loss: 0.5669\n",
      "Epoch 59, train_loss: 0.4666, val_loss: 0.5644\n",
      "Epoch 60, train_loss: 0.4654, val_loss: 0.5618\n",
      "Epoch 61, train_loss: 0.4637, val_loss: 0.5592\n",
      "Epoch 62, train_loss: 0.4623, val_loss: 0.5566\n",
      "Epoch 63, train_loss: 0.4616, val_loss: 0.5540\n",
      "Epoch 64, train_loss: 0.4601, val_loss: 0.5513\n",
      "Epoch 65, train_loss: 0.4587, val_loss: 0.5487\n",
      "Epoch 66, train_loss: 0.4573, val_loss: 0.5460\n",
      "Epoch 67, train_loss: 0.4563, val_loss: 0.5433\n",
      "Epoch 68, train_loss: 0.4548, val_loss: 0.5406\n",
      "Epoch 69, train_loss: 0.4535, val_loss: 0.5378\n",
      "Epoch 70, train_loss: 0.4524, val_loss: 0.5351\n",
      "Epoch 71, train_loss: 0.4512, val_loss: 0.5323\n",
      "Epoch 72, train_loss: 0.4502, val_loss: 0.5296\n",
      "Epoch 73, train_loss: 0.4487, val_loss: 0.5268\n",
      "Epoch 74, train_loss: 0.4481, val_loss: 0.5240\n",
      "Epoch 75, train_loss: 0.4468, val_loss: 0.5211\n",
      "Epoch 76, train_loss: 0.4453, val_loss: 0.5183\n",
      "Epoch 77, train_loss: 0.4445, val_loss: 0.5155\n",
      "Epoch 78, train_loss: 0.4431, val_loss: 0.5127\n",
      "Epoch 79, train_loss: 0.4423, val_loss: 0.5099\n",
      "Epoch 80, train_loss: 0.4414, val_loss: 0.5071\n",
      "Epoch 81, train_loss: 0.4398, val_loss: 0.5043\n",
      "Epoch 82, train_loss: 0.4382, val_loss: 0.5015\n",
      "Epoch 83, train_loss: 0.4375, val_loss: 0.4988\n",
      "Epoch 84, train_loss: 0.4365, val_loss: 0.4961\n",
      "Epoch 85, train_loss: 0.4354, val_loss: 0.4934\n",
      "Epoch 86, train_loss: 0.4344, val_loss: 0.4907\n",
      "Epoch 87, train_loss: 0.4336, val_loss: 0.4882\n",
      "Epoch 88, train_loss: 0.4324, val_loss: 0.4856\n",
      "Epoch 89, train_loss: 0.4313, val_loss: 0.4831\n",
      "Epoch 90, train_loss: 0.4303, val_loss: 0.4806\n",
      "Epoch 91, train_loss: 0.4286, val_loss: 0.4782\n",
      "Epoch 92, train_loss: 0.4276, val_loss: 0.4759\n",
      "Epoch 93, train_loss: 0.4262, val_loss: 0.4735\n",
      "Epoch 94, train_loss: 0.4261, val_loss: 0.4713\n",
      "Epoch 95, train_loss: 0.4244, val_loss: 0.4691\n",
      "Epoch 96, train_loss: 0.4239, val_loss: 0.4669\n",
      "Epoch 97, train_loss: 0.4232, val_loss: 0.4648\n",
      "Epoch 98, train_loss: 0.4215, val_loss: 0.4627\n",
      "Epoch 99, train_loss: 0.4204, val_loss: 0.4607\n",
      "Epoch 100, train_loss: 0.4191, val_loss: 0.4588\n",
      "Epoch 101, train_loss: 0.4185, val_loss: 0.4569\n",
      "Epoch 102, train_loss: 0.4183, val_loss: 0.4551\n",
      "Epoch 103, train_loss: 0.4166, val_loss: 0.4533\n",
      "Epoch 104, train_loss: 0.4157, val_loss: 0.4516\n",
      "Epoch 105, train_loss: 0.4142, val_loss: 0.4499\n",
      "Epoch 106, train_loss: 0.4136, val_loss: 0.4483\n",
      "Epoch 107, train_loss: 0.4123, val_loss: 0.4467\n",
      "Epoch 108, train_loss: 0.4114, val_loss: 0.4451\n",
      "Epoch 109, train_loss: 0.4106, val_loss: 0.4436\n",
      "Epoch 110, train_loss: 0.4100, val_loss: 0.4421\n",
      "Epoch 111, train_loss: 0.4083, val_loss: 0.4407\n",
      "Epoch 112, train_loss: 0.4079, val_loss: 0.4393\n",
      "Epoch 113, train_loss: 0.4072, val_loss: 0.4380\n",
      "Epoch 114, train_loss: 0.4059, val_loss: 0.4367\n",
      "Epoch 115, train_loss: 0.4050, val_loss: 0.4355\n",
      "Epoch 116, train_loss: 0.4042, val_loss: 0.4343\n",
      "Epoch 117, train_loss: 0.4035, val_loss: 0.4331\n",
      "Epoch 118, train_loss: 0.4022, val_loss: 0.4319\n",
      "Epoch 119, train_loss: 0.4016, val_loss: 0.4307\n",
      "Epoch 120, train_loss: 0.4001, val_loss: 0.4295\n",
      "Epoch 121, train_loss: 0.3999, val_loss: 0.4284\n",
      "Epoch 122, train_loss: 0.3983, val_loss: 0.4272\n",
      "Epoch 123, train_loss: 0.3971, val_loss: 0.4260\n",
      "Epoch 124, train_loss: 0.3961, val_loss: 0.4248\n",
      "Epoch 125, train_loss: 0.3952, val_loss: 0.4236\n",
      "Epoch 126, train_loss: 0.3941, val_loss: 0.4225\n",
      "Epoch 127, train_loss: 0.3936, val_loss: 0.4214\n",
      "Epoch 128, train_loss: 0.3924, val_loss: 0.4203\n",
      "Epoch 129, train_loss: 0.3917, val_loss: 0.4192\n",
      "Epoch 130, train_loss: 0.3907, val_loss: 0.4181\n",
      "Epoch 131, train_loss: 0.3898, val_loss: 0.4171\n",
      "Epoch 132, train_loss: 0.3884, val_loss: 0.4160\n",
      "Epoch 133, train_loss: 0.3882, val_loss: 0.4150\n",
      "Epoch 134, train_loss: 0.3874, val_loss: 0.4140\n",
      "Epoch 135, train_loss: 0.3859, val_loss: 0.4130\n",
      "Epoch 136, train_loss: 0.3856, val_loss: 0.4120\n",
      "Epoch 137, train_loss: 0.3851, val_loss: 0.4110\n",
      "Epoch 138, train_loss: 0.3839, val_loss: 0.4100\n",
      "Epoch 139, train_loss: 0.3831, val_loss: 0.4090\n",
      "Epoch 140, train_loss: 0.3823, val_loss: 0.4079\n",
      "Epoch 141, train_loss: 0.3817, val_loss: 0.4069\n",
      "Epoch 142, train_loss: 0.3804, val_loss: 0.4059\n",
      "Epoch 143, train_loss: 0.3794, val_loss: 0.4049\n",
      "Epoch 144, train_loss: 0.3784, val_loss: 0.4040\n",
      "Epoch 145, train_loss: 0.3778, val_loss: 0.4031\n",
      "Epoch 146, train_loss: 0.3767, val_loss: 0.4022\n",
      "Epoch 147, train_loss: 0.3764, val_loss: 0.4013\n",
      "Epoch 148, train_loss: 0.3753, val_loss: 0.4004\n",
      "Epoch 149, train_loss: 0.3738, val_loss: 0.3995\n",
      "Epoch 150, train_loss: 0.3733, val_loss: 0.3986\n",
      "Epoch 151, train_loss: 0.3725, val_loss: 0.3977\n",
      "Epoch 152, train_loss: 0.3717, val_loss: 0.3969\n",
      "Epoch 153, train_loss: 0.3709, val_loss: 0.3959\n",
      "Epoch 154, train_loss: 0.3706, val_loss: 0.3950\n",
      "Epoch 155, train_loss: 0.3692, val_loss: 0.3942\n",
      "Epoch 156, train_loss: 0.3678, val_loss: 0.3933\n",
      "Epoch 157, train_loss: 0.3677, val_loss: 0.3924\n",
      "Epoch 158, train_loss: 0.3667, val_loss: 0.3916\n",
      "Epoch 159, train_loss: 0.3664, val_loss: 0.3908\n",
      "Epoch 160, train_loss: 0.3652, val_loss: 0.3899\n",
      "Epoch 161, train_loss: 0.3644, val_loss: 0.3890\n",
      "Epoch 162, train_loss: 0.3635, val_loss: 0.3881\n",
      "Epoch 163, train_loss: 0.3628, val_loss: 0.3873\n",
      "Epoch 164, train_loss: 0.3622, val_loss: 0.3865\n",
      "Epoch 165, train_loss: 0.3607, val_loss: 0.3857\n",
      "Epoch 166, train_loss: 0.3602, val_loss: 0.3850\n",
      "Epoch 167, train_loss: 0.3594, val_loss: 0.3843\n",
      "Epoch 168, train_loss: 0.3587, val_loss: 0.3835\n",
      "Epoch 169, train_loss: 0.3579, val_loss: 0.3827\n",
      "Epoch 170, train_loss: 0.3568, val_loss: 0.3820\n",
      "Epoch 171, train_loss: 0.3562, val_loss: 0.3812\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 172, train_loss: 0.3553, val_loss: 0.3804\n",
      "Epoch 173, train_loss: 0.3549, val_loss: 0.3795\n",
      "Epoch 174, train_loss: 0.3542, val_loss: 0.3787\n",
      "Epoch 175, train_loss: 0.3536, val_loss: 0.3779\n",
      "Epoch 176, train_loss: 0.3522, val_loss: 0.3771\n",
      "Epoch 177, train_loss: 0.3520, val_loss: 0.3764\n",
      "Epoch 178, train_loss: 0.3511, val_loss: 0.3757\n",
      "Epoch 179, train_loss: 0.3504, val_loss: 0.3750\n",
      "Epoch 180, train_loss: 0.3497, val_loss: 0.3743\n",
      "Epoch 181, train_loss: 0.3487, val_loss: 0.3736\n",
      "Epoch 182, train_loss: 0.3486, val_loss: 0.3728\n",
      "Epoch 183, train_loss: 0.3475, val_loss: 0.3721\n",
      "Epoch 184, train_loss: 0.3462, val_loss: 0.3713\n",
      "Epoch 185, train_loss: 0.3463, val_loss: 0.3706\n",
      "Epoch 186, train_loss: 0.3449, val_loss: 0.3699\n",
      "Epoch 187, train_loss: 0.3445, val_loss: 0.3693\n",
      "Epoch 188, train_loss: 0.3442, val_loss: 0.3686\n",
      "Epoch 189, train_loss: 0.3428, val_loss: 0.3679\n",
      "Epoch 190, train_loss: 0.3424, val_loss: 0.3672\n",
      "Epoch 191, train_loss: 0.3416, val_loss: 0.3664\n",
      "Epoch 192, train_loss: 0.3410, val_loss: 0.3658\n",
      "Epoch 193, train_loss: 0.3403, val_loss: 0.3652\n",
      "Epoch 194, train_loss: 0.3394, val_loss: 0.3647\n",
      "Epoch 195, train_loss: 0.3382, val_loss: 0.3640\n",
      "Epoch 196, train_loss: 0.3381, val_loss: 0.3632\n",
      "Epoch 197, train_loss: 0.3371, val_loss: 0.3625\n",
      "Epoch 198, train_loss: 0.3365, val_loss: 0.3618\n",
      "Epoch 199, train_loss: 0.3357, val_loss: 0.3611\n",
      "Epoch 200, train_loss: 0.3356, val_loss: 0.3604\n",
      "Epoch 201, train_loss: 0.3347, val_loss: 0.3598\n",
      "Epoch 202, train_loss: 0.3346, val_loss: 0.3595\n",
      "Epoch 203, train_loss: 0.3333, val_loss: 0.3590\n",
      "Epoch 204, train_loss: 0.3328, val_loss: 0.3585\n",
      "Epoch 205, train_loss: 0.3327, val_loss: 0.3579\n",
      "Epoch 206, train_loss: 0.3319, val_loss: 0.3573\n",
      "Epoch 207, train_loss: 0.3311, val_loss: 0.3566\n",
      "Epoch 208, train_loss: 0.3304, val_loss: 0.3560\n",
      "Epoch 209, train_loss: 0.3291, val_loss: 0.3555\n",
      "Epoch 210, train_loss: 0.3288, val_loss: 0.3552\n",
      "Epoch 211, train_loss: 0.3286, val_loss: 0.3548\n",
      "Epoch 212, train_loss: 0.3274, val_loss: 0.3542\n",
      "Epoch 213, train_loss: 0.3265, val_loss: 0.3535\n",
      "Epoch 214, train_loss: 0.3266, val_loss: 0.3528\n",
      "Epoch 215, train_loss: 0.3259, val_loss: 0.3522\n",
      "Epoch 216, train_loss: 0.3251, val_loss: 0.3515\n",
      "Epoch 217, train_loss: 0.3245, val_loss: 0.3511\n",
      "Epoch 218, train_loss: 0.3240, val_loss: 0.3507\n",
      "Epoch 219, train_loss: 0.3234, val_loss: 0.3502\n",
      "Epoch 220, train_loss: 0.3231, val_loss: 0.3496\n",
      "Epoch 221, train_loss: 0.3226, val_loss: 0.3490\n",
      "Epoch 222, train_loss: 0.3211, val_loss: 0.3486\n",
      "Epoch 223, train_loss: 0.3210, val_loss: 0.3483\n",
      "Epoch 224, train_loss: 0.3203, val_loss: 0.3481\n",
      "Epoch 225, train_loss: 0.3198, val_loss: 0.3477\n",
      "Epoch 226, train_loss: 0.3189, val_loss: 0.3470\n",
      "Epoch 227, train_loss: 0.3191, val_loss: 0.3463\n",
      "Epoch 228, train_loss: 0.3180, val_loss: 0.3457\n",
      "Epoch 229, train_loss: 0.3173, val_loss: 0.3452\n",
      "Epoch 230, train_loss: 0.3174, val_loss: 0.3448\n",
      "Epoch 231, train_loss: 0.3166, val_loss: 0.3445\n",
      "Epoch 232, train_loss: 0.3159, val_loss: 0.3443\n",
      "Epoch 233, train_loss: 0.3156, val_loss: 0.3439\n",
      "Epoch 234, train_loss: 0.3153, val_loss: 0.3434\n",
      "Epoch 235, train_loss: 0.3144, val_loss: 0.3428\n",
      "Epoch 236, train_loss: 0.3138, val_loss: 0.3423\n",
      "Epoch 237, train_loss: 0.3133, val_loss: 0.3419\n",
      "Epoch 238, train_loss: 0.3132, val_loss: 0.3413\n",
      "Epoch 239, train_loss: 0.3133, val_loss: 0.3410\n",
      "Epoch 240, train_loss: 0.3122, val_loss: 0.3407\n",
      "Epoch 241, train_loss: 0.3115, val_loss: 0.3402\n",
      "Epoch 242, train_loss: 0.3111, val_loss: 0.3395\n",
      "Epoch 243, train_loss: 0.3105, val_loss: 0.3390\n",
      "Epoch 244, train_loss: 0.3100, val_loss: 0.3386\n",
      "Epoch 245, train_loss: 0.3093, val_loss: 0.3385\n",
      "Epoch 246, train_loss: 0.3090, val_loss: 0.3383\n",
      "Epoch 247, train_loss: 0.3090, val_loss: 0.3379\n",
      "Epoch 248, train_loss: 0.3083, val_loss: 0.3371\n",
      "Epoch 249, train_loss: 0.3077, val_loss: 0.3366\n",
      "Epoch 250, train_loss: 0.3072, val_loss: 0.3365\n",
      "Epoch 251, train_loss: 0.3071, val_loss: 0.3365\n",
      "Epoch 252, train_loss: 0.3064, val_loss: 0.3363\n",
      "Epoch 253, train_loss: 0.3060, val_loss: 0.3356\n",
      "Epoch 254, train_loss: 0.3049, val_loss: 0.3348\n",
      "Epoch 255, train_loss: 0.3047, val_loss: 0.3344\n",
      "Epoch 256, train_loss: 0.3043, val_loss: 0.3342\n",
      "Epoch 257, train_loss: 0.3042, val_loss: 0.3340\n",
      "Epoch 258, train_loss: 0.3039, val_loss: 0.3337\n",
      "Epoch 259, train_loss: 0.3034, val_loss: 0.3332\n",
      "Epoch 260, train_loss: 0.3029, val_loss: 0.3327\n",
      "Epoch 261, train_loss: 0.3026, val_loss: 0.3326\n",
      "Epoch 262, train_loss: 0.3019, val_loss: 0.3329\n",
      "Epoch 263, train_loss: 0.3013, val_loss: 0.3326\n",
      "Epoch 264, train_loss: 0.3012, val_loss: 0.3318\n",
      "Epoch 265, train_loss: 0.3007, val_loss: 0.3314\n",
      "Epoch 266, train_loss: 0.3006, val_loss: 0.3308\n",
      "Epoch 267, train_loss: 0.2998, val_loss: 0.3310\n",
      "Epoch 268, train_loss: 0.2994, val_loss: 0.3310\n",
      "Epoch 269, train_loss: 0.2992, val_loss: 0.3305\n",
      "Epoch 270, train_loss: 0.2983, val_loss: 0.3300\n",
      "Epoch 271, train_loss: 0.2986, val_loss: 0.3296\n",
      "Epoch 272, train_loss: 0.2978, val_loss: 0.3293\n",
      "Epoch 273, train_loss: 0.2974, val_loss: 0.3293\n",
      "Epoch 274, train_loss: 0.2969, val_loss: 0.3293\n",
      "Epoch 275, train_loss: 0.2970, val_loss: 0.3287\n",
      "Epoch 276, train_loss: 0.2964, val_loss: 0.3279\n",
      "Epoch 277, train_loss: 0.2961, val_loss: 0.3277\n",
      "Epoch 278, train_loss: 0.2959, val_loss: 0.3276\n",
      "Epoch 279, train_loss: 0.2958, val_loss: 0.3278\n",
      "Epoch 280, train_loss: 0.2953, val_loss: 0.3278\n",
      "Epoch 281, train_loss: 0.2943, val_loss: 0.3272\n",
      "Epoch 282, train_loss: 0.2949, val_loss: 0.3261\n",
      "Epoch 283, train_loss: 0.2941, val_loss: 0.3259\n",
      "Epoch 284, train_loss: 0.2937, val_loss: 0.3266\n",
      "Epoch 285, train_loss: 0.2933, val_loss: 0.3265\n",
      "Epoch 286, train_loss: 0.2928, val_loss: 0.3258\n",
      "Epoch 287, train_loss: 0.2931, val_loss: 0.3250\n",
      "Epoch 288, train_loss: 0.2925, val_loss: 0.3248\n",
      "Epoch 289, train_loss: 0.2918, val_loss: 0.3251\n",
      "Epoch 290, train_loss: 0.2912, val_loss: 0.3253\n",
      "Epoch 291, train_loss: 0.2914, val_loss: 0.3250\n",
      "Epoch 292, train_loss: 0.2902, val_loss: 0.3241\n",
      "Epoch 293, train_loss: 0.2903, val_loss: 0.3235\n",
      "Epoch 294, train_loss: 0.2904, val_loss: 0.3233\n",
      "Epoch 295, train_loss: 0.2899, val_loss: 0.3233\n",
      "Epoch 296, train_loss: 0.2894, val_loss: 0.3234\n",
      "Epoch 297, train_loss: 0.2892, val_loss: 0.3226\n",
      "Epoch 298, train_loss: 0.2889, val_loss: 0.3223\n",
      "Epoch 299, train_loss: 0.2886, val_loss: 0.3223\n",
      "Epoch 300, train_loss: 0.2884, val_loss: 0.3225\n",
      "Epoch 301, train_loss: 0.2882, val_loss: 0.3220\n",
      "Epoch 302, train_loss: 0.2880, val_loss: 0.3222\n",
      "Epoch 303, train_loss: 0.2876, val_loss: 0.3225\n",
      "Epoch 304, train_loss: 0.2872, val_loss: 0.3219\n",
      "Epoch 305, train_loss: 0.2869, val_loss: 0.3215\n",
      "Epoch 306, train_loss: 0.2871, val_loss: 0.3219\n",
      "Epoch 307, train_loss: 0.2861, val_loss: 0.3215\n",
      "Epoch 308, train_loss: 0.2861, val_loss: 0.3210\n",
      "Epoch 309, train_loss: 0.2857, val_loss: 0.3204\n",
      "Epoch 310, train_loss: 0.2856, val_loss: 0.3211\n",
      "Epoch 311, train_loss: 0.2849, val_loss: 0.3211\n",
      "Epoch 312, train_loss: 0.2847, val_loss: 0.3202\n",
      "Epoch 313, train_loss: 0.2843, val_loss: 0.3197\n",
      "Epoch 314, train_loss: 0.2845, val_loss: 0.3200\n",
      "Epoch 315, train_loss: 0.2845, val_loss: 0.3196\n",
      "Epoch 316, train_loss: 0.2837, val_loss: 0.3191\n",
      "Epoch 317, train_loss: 0.2836, val_loss: 0.3189\n",
      "Epoch 318, train_loss: 0.2834, val_loss: 0.3196\n",
      "Epoch 319, train_loss: 0.2831, val_loss: 0.3192\n",
      "Epoch 320, train_loss: 0.2828, val_loss: 0.3181\n",
      "Epoch 321, train_loss: 0.2827, val_loss: 0.3189\n",
      "Epoch 322, train_loss: 0.2825, val_loss: 0.3188\n",
      "Epoch 323, train_loss: 0.2822, val_loss: 0.3189\n",
      "Epoch 324, train_loss: 0.2817, val_loss: 0.3178\n",
      "Epoch 325, train_loss: 0.2822, val_loss: 0.3180\n",
      "Epoch 326, train_loss: 0.2811, val_loss: 0.3186\n",
      "Epoch 327, train_loss: 0.2811, val_loss: 0.3177\n",
      "Epoch 328, train_loss: 0.2811, val_loss: 0.3170\n",
      "Epoch 329, train_loss: 0.2809, val_loss: 0.3170\n",
      "Epoch 330, train_loss: 0.2802, val_loss: 0.3179\n",
      "Epoch 331, train_loss: 0.2803, val_loss: 0.3168\n",
      "Epoch 332, train_loss: 0.2795, val_loss: 0.3166\n",
      "Epoch 333, train_loss: 0.2801, val_loss: 0.3171\n",
      "Epoch 334, train_loss: 0.2798, val_loss: 0.3168\n",
      "Epoch 335, train_loss: 0.2791, val_loss: 0.3170\n",
      "Epoch 336, train_loss: 0.2787, val_loss: 0.3166\n",
      "Epoch 337, train_loss: 0.2791, val_loss: 0.3164\n",
      "Epoch 338, train_loss: 0.2793, val_loss: 0.3165\n",
      "Epoch 339, train_loss: 0.2786, val_loss: 0.3160\n",
      "Epoch 340, train_loss: 0.2778, val_loss: 0.3168\n",
      "Epoch 341, train_loss: 0.2780, val_loss: 0.3163\n",
      "Epoch 342, train_loss: 0.2777, val_loss: 0.3150\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 343, train_loss: 0.2775, val_loss: 0.3153\n",
      "Epoch 344, train_loss: 0.2774, val_loss: 0.3165\n",
      "Epoch 345, train_loss: 0.2772, val_loss: 0.3161\n",
      "Epoch 346, train_loss: 0.2768, val_loss: 0.3154\n",
      "Epoch 347, train_loss: 0.2768, val_loss: 0.3141\n",
      "Epoch 348, train_loss: 0.2762, val_loss: 0.3141\n",
      "Epoch 349, train_loss: 0.2763, val_loss: 0.3144\n",
      "Epoch 350, train_loss: 0.2762, val_loss: 0.3149\n",
      "Epoch 351, train_loss: 0.2762, val_loss: 0.3146\n",
      "Epoch 352, train_loss: 0.2756, val_loss: 0.3142\n",
      "Epoch 353, train_loss: 0.2756, val_loss: 0.3145\n",
      "Epoch 354, train_loss: 0.2755, val_loss: 0.3148\n",
      "Epoch 355, train_loss: 0.2755, val_loss: 0.3143\n",
      "Epoch 356, train_loss: 0.2748, val_loss: 0.3142\n",
      "Epoch 357, train_loss: 0.2748, val_loss: 0.3145\n",
      "Epoch 358, train_loss: 0.2744, val_loss: 0.3133\n",
      "Epoch 359, train_loss: 0.2745, val_loss: 0.3136\n",
      "Epoch 360, train_loss: 0.2741, val_loss: 0.3139\n",
      "Epoch 361, train_loss: 0.2745, val_loss: 0.3143\n",
      "Epoch 362, train_loss: 0.2739, val_loss: 0.3147\n",
      "Epoch 363, train_loss: 0.2739, val_loss: 0.3135\n",
      "Epoch 364, train_loss: 0.2739, val_loss: 0.3135\n",
      "Epoch 365, train_loss: 0.2733, val_loss: 0.3134\n",
      "Epoch 366, train_loss: 0.2735, val_loss: 0.3130\n",
      "Epoch 367, train_loss: 0.2734, val_loss: 0.3134\n",
      "Epoch 368, train_loss: 0.2731, val_loss: 0.3124\n",
      "Epoch 369, train_loss: 0.2725, val_loss: 0.3130\n",
      "Epoch 370, train_loss: 0.2726, val_loss: 0.3141\n",
      "Epoch 371, train_loss: 0.2726, val_loss: 0.3129\n",
      "Epoch 372, train_loss: 0.2727, val_loss: 0.3130\n",
      "Epoch 373, train_loss: 0.2723, val_loss: 0.3134\n",
      "Epoch 374, train_loss: 0.2720, val_loss: 0.3118\n",
      "Epoch 375, train_loss: 0.2722, val_loss: 0.3120\n",
      "Epoch 376, train_loss: 0.2720, val_loss: 0.3127\n",
      "Epoch 377, train_loss: 0.2712, val_loss: 0.3118\n",
      "Epoch 378, train_loss: 0.2720, val_loss: 0.3121\n",
      "Epoch 379, train_loss: 0.2712, val_loss: 0.3125\n",
      "Epoch 380, train_loss: 0.2713, val_loss: 0.3118\n",
      "Epoch 381, train_loss: 0.2707, val_loss: 0.3120\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "IOPub message rate exceeded.\n",
      "The notebook server will temporarily stop sending output\n",
      "to the client in order to avoid crashing it.\n",
      "To change this limit, set the config variable\n",
      "`--NotebookApp.iopub_msg_rate_limit`.\n",
      "\n",
      "Current values:\n",
      "NotebookApp.iopub_msg_rate_limit=1000.0 (msgs/sec)\n",
      "NotebookApp.rate_limit_window=3.0 (secs)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "torch.set_num_threads=16\n",
    "\n",
    "# Reproducibility\n",
    "np.random.seed(42)\n",
    "random.seed(42)\n",
    "torch.manual_seed(42)\n",
    "\n",
    "fc_hidden_dim = 3000\n",
    "gene_net_hidden_dim = 830\n",
    "disease_net_hidden_dim = 500\n",
    "folds = 5\n",
    "max_epochs = 500 \n",
    "early_stopping_window = 16\n",
    "lr = 0.00004\n",
    "weight_decay = 0.15\n",
    "\n",
    "metrics, dis_dict, model = train(\n",
    "    max_epochs=max_epochs, \n",
    "    early_stopping_window=early_stopping_window,\n",
    "    folds=folds, \n",
    "    lr=lr,\n",
    "    weight_decay=weight_decay,\n",
    "    fc_hidden_dim=fc_hidden_dim,\n",
    "    gene_net_hidden_dim=gene_net_hidden_dim,\n",
    "    disease_net_hidden_dim=disease_net_hidden_dim\n",
    ")\n",
    "\n",
    "# Store the results.\n",
    "if not os.path.exists(RESULTS_STORAGE):\n",
    "    os.makedirs(RESULTS_STORAGE)\n",
    "with gzip.open(osp.join(RESULTS_STORAGE, f'{EXPERIMENT_SLUG}_metrics.pickle.gz'), mode='wb') as file:\n",
    "    pickle.dump(metrics, file)\n",
    "with gzip.open(osp.join(RESULTS_STORAGE, f'{EXPERIMENT_SLUG}_dis_dict.pickle.gz'), mode='wb') as file:\n",
    "    pickle.dump(dis_dict, file)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Store the model for later use in disease class prediction.\n",
    "from shutil import copyfile\n",
    "for i in range(1, folds + 1):\n",
    "    copyfile(\n",
    "        osp.join(MODEL_TMP_STORAGE, f'best_model_fold_{i}.ptm'),\n",
    "        osp.join(RESULTS_STORAGE, f'model_fold_{i}.ptm')\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Eval the results.\n",
    "* Load the results.\n",
    "* Compute the ROC for each fold and combined. Distinguish by mono and multigenic diseases."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the results\n",
    "with gzip.open(osp.join(RESULTS_STORAGE, f'{EXPERIMENT_SLUG}_metrics.pickle.gz'), mode='rb') as file:\n",
    "    metrics = pickle.load(file, encoding='bytes')\n",
    "with gzip.open(osp.join(RESULTS_STORAGE, f'{EXPERIMENT_SLUG}_dis_dict.pickle.gz'), mode='rb') as file:\n",
    "    dis_dict = pickle.load(file, encoding='bytes')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from sklearn.metrics import roc_curve, auc\n",
    "\n",
    "\n",
    "def plot_folds_roc(disease_gene_type):\n",
    "    average_roc = sum([m[disease_gene_type]['AUC'] for m in metrics])/len(metrics)\n",
    "    colors = iter(plt.cm.copper(np.linspace(0, 1, len(metrics))))\n",
    "    plt.figure(figsize=(8, 8))\n",
    "    labels, lines = [], []\n",
    "\n",
    "    for i, color in zip(range(len(metrics)), colors):\n",
    "        l, = plt.plot(metrics[i][disease_gene_type]['FPR'], metrics[i][disease_gene_type]['TPR'], color=color, lw=2)\n",
    "        lines.append(l)\n",
    "        auc_val = metrics[i][disease_gene_type]['AUC']\n",
    "        labels.append(f'Fold {i + 1} (AUC = {auc_val:0.2f})')\n",
    "\n",
    "    l, = plt.plot([0, 1], [0, 1], color=[0, 0, 0, 0.3], lw=2, linestyle='--')\n",
    "    lines.append(l)\n",
    "\n",
    "    fig = plt.gcf()\n",
    "    fig.subplots_adjust(bottom=0.25)\n",
    "    plt.xlim([0.0, 1.0])\n",
    "    plt.ylim([0.0, 1.0])\n",
    "    plt.xlabel('False Positive Rate')\n",
    "    plt.ylabel('True Positive Rate')\n",
    "    plt.title(f'Disease gene prediction ROC. (Average AUC: {average_roc:0.2f})')\n",
    "    plt.legend(lines, labels, loc=(0.56, 0.02), prop=dict(size=14))\n",
    "\n",
    "    fig.savefig(\n",
    "        osp.join(RESULTS_STORAGE, f\"Disease_gene_prediction_ROC_by_fold_{disease_gene_type}genic_diseases.pdf\"),\n",
    "        bbox_inches='tight'\n",
    "    )\n",
    "    plt.show()\n",
    "\n",
    "plot_folds_roc('mono')\n",
    "plot_folds_roc('multi')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "from sklearn.metrics import roc_curve, auc\n",
    "\n",
    "colors = iter(plt.cm.copper(np.linspace(0, 1, 2)))\n",
    "plt.figure(figsize=(8, 8))\n",
    "labels, lines = [], []\n",
    "\n",
    "# Average FPR and TPR over the folds.\n",
    "avg_metrics = {\n",
    "    'mono': [],\n",
    "    'multi': []\n",
    "}\n",
    "\n",
    "for mode, color in zip(avg_metrics.keys(), colors):\n",
    "    t_pred_tmp = metrics[0][mode]['TEST_PREDICT'].copy()\n",
    "    t_pos_tmp = metrics[0][mode]['TEST_Y'].copy()\n",
    "    for i in range(len(metrics)):\n",
    "        t_pred_tmp = np.concatenate((t_pred_tmp, metrics[0][mode]['TEST_PREDICT'].copy()), axis=0)\n",
    "        t_pos_tmp = np.concatenate((t_pos_tmp, metrics[0][mode]['TEST_Y'].copy()), axis=0)\n",
    "    \n",
    "    fpr, tpr, _ = roc_curve(t_pos_tmp, t_pred_tmp[:,1], pos_label=1)\n",
    "    average_roc = sum([m[mode]['AUC'] for m in metrics])/len(metrics)\n",
    "    \n",
    "    l, = plt.plot(fpr, tpr, color=color, lw=2)\n",
    "    lines.append(l)\n",
    "    labels.append(f'M{mode[1:]}genetic (AUC = {average_roc:0.2f})')\n",
    "\n",
    "l, = plt.plot([0, 1], [0, 1], color=[0, 0, 0, 0.3], lw=2, linestyle='--')\n",
    "lines.append(l)\n",
    "\n",
    "fig = plt.gcf()\n",
    "fig.subplots_adjust(bottom=0.25)\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.0])\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.title(f'Disease gene prediction ROC.')\n",
    "plt.legend(lines, labels, loc=(0.45, 0.02), prop=dict(size=14))\n",
    "\n",
    "fig.savefig(\n",
    "    osp.join(RESULTS_STORAGE, \"Disease_gene_prediction_ROC_combined.pdf\"),\n",
    "    bbox_inches='tight'\n",
    ")\n",
    "plt.show()"
   ]
  }
 ],
 "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}