train_link_classification.py

import logging
import time
import sys
import os
from tqdm import tqdm
import numpy as np
import warnings
import shutil
import json
import torch
import torch.nn as nn
from models.TGAT import TGAT
from models.DyGKT import DyGKT
from models.MemoryModel import MemoryModel, compute_src_dst_node_time_shifts
from models.simpleKT import SimpleKT
from models.DyGFormer import DyGFormer
from models.CTNCM import CTNCM
from models.DKT import DKT
from models.DIMKT import DIMKT
from models.QIKT import QIKT
from models.IPKT import IPKT
from models.IEKT import IEKT
from models.AKT import AKT
from models.modules import MergeLayer, MLPClassifier
from utils.utils import set_random_seed, convert_to_gpu, get_parameter_sizes, create_optimizer
from utils.utils import get_neighbor_sampler, NegativeEdgeSampler
from evaluate_models_utils import evaluate_model_link_classification
from utils.metrics import get_link_classification_metrics
from utils.DataLoader import get_idx_data_loader, get_link_classification_data
from utils.EarlyStopping import EarlyStopping
from utils.load_configs import get_link_classification_args

if __name__ == "__main__":

    warnings.filterwarnings('ignore')

    # get arguments
    args = get_link_classification_args(is_evaluation=False)

    # get data for training, validation and testing
    node_raw_features, edge_raw_features, full_data, train_data, val_data, test_data, new_node_val_data, new_node_test_data = \
        get_link_classification_data(dataset_name=args.dataset_name, val_ratio=args.val_ratio, test_ratio=args.test_ratio)
    # print(node_raw_features)
    # initialize training neighbor sampler to retrieve temporal graph
    train_neighbor_sampler = get_neighbor_sampler(data=train_data, sample_neighbor_strategy=args.sample_neighbor_strategy,
                                                  time_scaling_factor=args.time_scaling_factor, seed=0)

    # initialize validation and test neighbor sampler to retrieve temporal graph
    full_neighbor_sampler = get_neighbor_sampler(data=full_data, sample_neighbor_strategy=args.sample_neighbor_strategy,
                                                 time_scaling_factor=args.time_scaling_factor, seed=1)

    # initialize negative samplers, set seeds for validation and testing so negatives are the same across different runs
    # in the inductive setting, negatives are sampled only amongst other new nodes
    # train negative edge sampler does not need to specify the seed, but evaluation samplers need to do so
    train_neg_edge_sampler = NegativeEdgeSampler(src_node_ids=train_data.src_node_ids, dst_node_ids=train_data.dst_node_ids)
    val_neg_edge_sampler = NegativeEdgeSampler(src_node_ids=full_data.src_node_ids, dst_node_ids=full_data.dst_node_ids, seed=0)
    new_node_val_neg_edge_sampler = NegativeEdgeSampler(src_node_ids=new_node_val_data.src_node_ids, dst_node_ids=new_node_val_data.dst_node_ids, seed=1)
    test_neg_edge_sampler = NegativeEdgeSampler(src_node_ids=full_data.src_node_ids, dst_node_ids=full_data.dst_node_ids, seed=2)
    new_node_test_neg_edge_sampler = NegativeEdgeSampler(src_node_ids=new_node_test_data.src_node_ids, dst_node_ids=new_node_test_data.dst_node_ids, seed=3)

    # get data loaders
    train_idx_data_loader = get_idx_data_loader(indices_list=list(range(len(train_data.src_node_ids))), batch_size=args.batch_size, shuffle=False)
    val_idx_data_loader = get_idx_data_loader(indices_list=list(range(len(val_data.src_node_ids))), batch_size=args.batch_size, shuffle=False)
    new_node_val_idx_data_loader = get_idx_data_loader(indices_list=list(range(len(new_node_val_data.src_node_ids))), batch_size=args.batch_size, shuffle=False)
    test_idx_data_loader = get_idx_data_loader(indices_list=list(range(len(test_data.src_node_ids))), batch_size=args.batch_size, shuffle=False)
    new_node_test_idx_data_loader = get_idx_data_loader(indices_list=list(range(len(new_node_test_data.src_node_ids))), batch_size=args.batch_size, shuffle=False)

    val_metric_all_runs, new_node_val_metric_all_runs, test_metric_all_runs, new_node_test_metric_all_runs = [], [], [], []

    for run in range(args.num_runs):

        set_random_seed(seed=run)

        args.seed = run
        args.save_model_name = f'{args.model_name}_seed{args.seed}'

        # set up logger
        logging.basicConfig(level=logging.INFO)
        logger = logging.getLogger()
        logger.setLevel(logging.DEBUG)
        os.makedirs(f"./logs/{args.model_name}/{args.dataset_name}/{args.save_model_name}/", exist_ok=True)
        inf = 'paper'
        fh = logging.FileHandler(f"./logs/{args.model_name}/{args.dataset_name}/{args.save_model_name}/{inf+str(time.time())}.log")
        fh.setLevel(logging.DEBUG)
        # create console handler with a higher log level
        ch = logging.StreamHandler()
        ch.setLevel(logging.WARNING)
        # create formatter and add it to the handlers
        formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
        fh.setFormatter(formatter)
        ch.setFormatter(formatter)
        # add the handlers to logger
        logger.addHandler(fh)
        logger.addHandler(ch)

        run_start_time = time.time()
        start_time = time.time()
        initial_memory = torch.cuda.memory_allocated()
        logger.info(f"********** Run {run + 1} starts. **********")

        logger.info(f'configuration is {args}')

        # create model

        if args.model_name == 'DKT':
            dynamic_backbone = DKT(node_raw_features=node_raw_features, 
                                        edge_raw_features=edge_raw_features, 
                                        dropout=args.dropout,
                                        num_neighbors=args.num_neighbors, 
                                        device=args.device)
        elif args.model_name == 'DyGKT':
            dynamic_backbone = DyGKT(node_raw_features=node_raw_features,
                                        edge_raw_features=edge_raw_features,
                                        dropout=args.dropout, 
                                        num_neighbors=args.num_neighbors,
                                        device=args.device,
                                        ablation=args.ablation)

        elif args.model_name == 'CTNCM':
            dynamic_backbone = CTNCM(node_raw_features=node_raw_features,
                                        edge_raw_features=edge_raw_features,
                                        dropout=args.dropout, 
                                        num_neighbors=args.num_neighbors,
                                        device=args.device)

        elif args.model_name == 'AKT':
            dynamic_backbone = AKT(node_raw_features=node_raw_features,
                                        edge_raw_features=edge_raw_features,
                                        dropout=args.dropout, 
                                        num_neighbors=args.num_neighbors,
                                        device=args.device)
        elif args.model_name == 'DIMKT':
            dynamic_backbone = DIMKT(node_raw_features=node_raw_features,
                                    edge_raw_features=edge_raw_features,
                                    dropout=args.dropout,
                                    dataset_name=args.dataset_name,
                                    device=args.device)
        elif args.model_name == 'IPKT':
            dynamic_backbone = IPKT(node_raw_features=node_raw_features,
                                    edge_raw_features=edge_raw_features,
                                    dropout=args.dropout,
                                    device=args.device)
            
        elif args.model_name == 'IEKT':
            dynamic_backbone = IEKT(node_raw_features=node_raw_features,
                                    edge_raw_features=edge_raw_features,
                                    dropout=args.dropout,
                                    device=args.device) 
        elif args.model_name == 'QIKT':
            dynamic_backbone = QIKT(node_raw_features=node_raw_features,
                                    edge_raw_features=edge_raw_features,
                                    dropout=args.dropout,
                                    device=args.device)              
            
        elif args.model_name == 'simpleKT':
            dynamic_backbone = SimpleKT(node_raw_features=node_raw_features,
                                        edge_raw_features=edge_raw_features,
                                        dropout=args.dropout, 
                                        num_neighbors=args.num_neighbors,
                                        device=args.device)

        elif args.model_name == 'TGAT':
            dynamic_backbone = TGAT(node_raw_features=node_raw_features, edge_raw_features=edge_raw_features, neighbor_sampler=train_neighbor_sampler,
                                    time_feat_dim=args.time_feat_dim, num_layers=args.num_layers, num_heads=args.num_heads, dropout=args.dropout, device=args.device)

      
        elif args.model_name in ['TGN']:
            # four floats that represent the mean and standard deviation of source and destination node time shifts in the training data, which is used for JODIE
            src_node_mean_time_shift, src_node_std_time_shift, dst_node_mean_time_shift_dst, dst_node_std_time_shift = \
                compute_src_dst_node_time_shifts(train_data.src_node_ids, train_data.dst_node_ids, train_data.node_interact_times)
            dynamic_backbone = MemoryModel(node_raw_features=node_raw_features, edge_raw_features=edge_raw_features, neighbor_sampler=train_neighbor_sampler,
                                           time_feat_dim=args.time_feat_dim, model_name=args.model_name, num_layers=args.num_layers, num_heads=args.num_heads,
                                           dropout=args.dropout, src_node_mean_time_shift=src_node_mean_time_shift, src_node_std_time_shift=src_node_std_time_shift,
                                           dst_node_mean_time_shift_dst=dst_node_mean_time_shift_dst, dst_node_std_time_shift=dst_node_std_time_shift, device=args.device)

        elif args.model_name == 'DyGFormer':
            dynamic_backbone = DyGFormer(node_raw_features=node_raw_features, edge_raw_features=edge_raw_features, neighbor_sampler=train_neighbor_sampler,
                                         time_feat_dim=args.time_feat_dim, channel_embedding_dim=args.channel_embedding_dim, patch_size=args.patch_size,
                                         num_layers=args.num_layers, num_heads=args.num_heads, dropout=args.dropout,
                                         max_input_sequence_length=args.max_input_sequence_length, device=args.device)
        else:
            raise ValueError(f"Wrong value for model_name {args.model_name}!")
        if args.model_name == 'DyGKT':
            link_predictor = MergeLayer(input_dim1=64, input_dim2=64,hidden_dim=64,output_dim=1)
        else:
            link_predictor = MergeLayer(node_raw_features.shape[1],input_dim2=node_raw_features.shape[1], hidden_dim=node_raw_features.shape[1], output_dim=1)#input_dim1=64, input_dim2=64,hidden_dim=64,output_dim=1)#node_raw_features.shape[1],input_dim2=node_raw_features.shape[1], hidden_dim=node_raw_features.shape[1], output_dim=1)
        # MLPClassifier(node_dim, dropout=args.dropout)
        model = nn.Sequential(dynamic_backbone, link_predictor)
        logger.info(f'model -> {model}')
        logger.info(f'model name: {args.model_name}, #parameters: {get_parameter_sizes(model) * 4} B, '
                    f'{get_parameter_sizes(model) * 4 / 1024} KB, {get_parameter_sizes(model) * 4 / 1024 / 1024} MB.')

        optimizer = create_optimizer(model=model, optimizer_name=args.optimizer, learning_rate=args.learning_rate, weight_decay=args.weight_decay)

        model = convert_to_gpu(model, device=args.device)

        save_model_folder = f"./saved_models/{args.model_name}/{args.dataset_name}/{args.save_model_name}/"
        shutil.rmtree(save_model_folder, ignore_errors=True)
        os.makedirs(save_model_folder, exist_ok=True)

        early_stopping = EarlyStopping(patience=args.patience, save_model_folder=save_model_folder,
                                       save_model_name=args.save_model_name, logger=logger, model_name=args.model_name)

        loss_func = nn.BCELoss()
        torch.autograd.set_detect_anomaly(True)
        final_memory = None
        for epoch in range(args.num_epochs):
            if args.test and epoch > 0:
                end_time = time.time()
                memory_used = final_memory - initial_memory
                logger.info(f'USE_TIME: {end_time-start_time}')
                logger.info(f"USE_GPU: {memory_used / (1024**2):.3f} MB")
                logger.info(f'MODEL_PARA:{get_parameter_sizes(model) * 4 / 1024 / 1024:.3f} MB')
                print(args.model_name)
                sys.exit()
            model.train()
            if args.model_name in ['DyGKT','QIKT','IEKT','IPKT','DIMKT', 'TGAT', 'TGN', 'DyGFormer','DKT','AKT','CTNCM','simpleKT']:
                # training, only use training graph
                model[0].set_neighbor_sampler(train_neighbor_sampler)
            if args.model_name in ['TGN']:
                # reinitialize memory of memory-based models at the start of each epoch
                model[0].memory_bank.__init_memory_bank__()
                model[0].last_node_id = None

            # store train losses and metrics
            train_losses, train_metrics = [], []
            train_predicts, train_labels = [], []
            train_idx_data_loader_tqdm = tqdm(train_idx_data_loader, ncols=120)
            for batch_idx, train_data_indices in enumerate(train_idx_data_loader_tqdm):
                train_data_indices = train_data_indices.numpy()
                batch_src_node_ids, batch_dst_node_ids, batch_node_interact_times, batch_edge_ids, batch_edge_labels= \
                    train_data.src_node_ids[train_data_indices], train_data.dst_node_ids[train_data_indices], \
                    train_data.node_interact_times[train_data_indices], train_data.edge_ids[train_data_indices],\
                    train_data.labels[train_data_indices]

             
                
                if args.model_name in ['DyGKT','DKT','AKT','CTNCM','simpleKT']:
                    batch_src_node_embeddings,batch_dst_node_embeddings = \
                        model[0].compute_src_dst_node_temporal_embeddings(src_node_ids=batch_src_node_ids,
                                                                          edge_ids = batch_edge_ids,
                                                                          node_interact_times=batch_node_interact_times,
                                                                          dst_node_ids=batch_dst_node_ids)
                
                elif args.model_name in ['QIKT','IEKT','IPKT','DIMKT']:
                    batch_src_node_embeddings,batch_dst_node_embeddings = \
                        model[0].compute_src_dst_node_temporal_embeddings(src_node_ids=batch_src_node_ids,
                                                                            dst_node_ids=batch_dst_node_ids,
                                                                            node_interact_times=batch_node_interact_times,
                                                                            edge_ids=batch_edge_ids)
                            

                elif args.model_name in ['TGAT']:
                    # get temporal embedding of source and destination nodes
                    # two Tensors, with shape (batch_size, node_feat_dim)
                    batch_src_node_embeddings, batch_dst_node_embeddings = \
                        model[0].compute_src_dst_node_temporal_embeddings(src_node_ids=batch_src_node_ids,
                                                                          dst_node_ids=batch_dst_node_ids,
                                                                          node_interact_times=batch_node_interact_times,
                                                                          num_neighbors=args.num_neighbors)

                elif args.model_name in ['TGN']:
                    # note that negative nodes do not change the memories while the positive nodes change the memories,
                    # we need to first compute the embeddings of negative nodes for memory-based models
                    # get temporal embedding of negative source and negative destination nodes
                    # two Tensors, with shape (batch_size, node_feat_dim)
                    batch_src_node_embeddings, batch_dst_node_embeddings = \
                        model[0].compute_src_dst_node_temporal_embeddings(src_node_ids=batch_src_node_ids,
                                                                          dst_node_ids=batch_dst_node_ids,
                                                                          node_interact_times=batch_node_interact_times,
                                                                          edge_ids=batch_edge_ids,
                                                                          edges_are_positive=True,
                                                                          num_neighbors=args.num_neighbors)
                
                elif args.model_name in ['DyGFormer']:
                    # get temporal embedding of source and destination nodes
                    # two Tensors, with shape (batch_size, node_feat_dim)
                    batch_src_node_embeddings, batch_dst_node_embeddings = \
                        model[0].compute_src_dst_node_temporal_embeddings(src_node_ids=batch_src_node_ids,
                                                                          dst_node_ids=batch_dst_node_ids,
                                                                          node_interact_times=batch_node_interact_times)
                else:
                    raise ValueError(f"Wrong value for model_name {args.model_name}!")
                
                
                predicts = model[1](batch_src_node_embeddings,batch_dst_node_embeddings).squeeze(dim=-1).sigmoid()
                labels = torch.tensor(batch_edge_labels, dtype=torch.float32,device=args.device)
                loss = loss_func(input=predicts, target=labels)
                final_memory = torch.cuda.memory_allocated()
                train_losses.append(loss.item())

                train_predicts.append(predicts)
                train_labels.append(labels)

                optimizer.zero_grad()
                loss.backward()
                optimizer.step()

                train_idx_data_loader_tqdm.set_description(f'Epoch: {epoch + 1}, train for the {batch_idx + 1}-th batch, train loss: {loss.item()}')

                if args.model_name in ['TGN']:
                    # detach the memories and raw messages of nodes in the memory bank after each batch, so we don't back propagate to the start of time
                    model[0].memory_bank.detach_memory_bank()

            if args.model_name in ['TGN']:
                # backup memory bank after training so it can be used for new validation nodes
                train_backup_memory_bank = model[0].memory_bank.backup_memory_bank()
            val_losses, val_metrics = evaluate_model_link_classification(model_name=args.model_name,
                                                                     model=model,
                                                                     neighbor_sampler=full_neighbor_sampler,
                                                                     evaluate_idx_data_loader=val_idx_data_loader,
                                                                     evaluate_neg_edge_sampler=val_neg_edge_sampler,
                                                                     evaluate_data=val_data,
                                                                     loss_func=loss_func,
                                                                     num_neighbors=args.num_neighbors,
                                                                     time_gap=args.time_gap)

            if args.model_name in ['TGN']:
                # backup memory bank after validating so it can be used for testing nodes (since test edges are strictly later in time than validation edges)
                val_backup_memory_bank = model[0].memory_bank.backup_memory_bank()

                # reload training memory bank for new validation nodes
                model[0].memory_bank.reload_memory_bank(train_backup_memory_bank)

            new_node_val_losses, new_node_val_metrics = evaluate_model_link_classification(model_name=args.model_name,
                                                                                           model=model,
                                                                                           neighbor_sampler=full_neighbor_sampler,
                                                                                           evaluate_idx_data_loader=new_node_val_idx_data_loader,
                                                                                           evaluate_neg_edge_sampler=new_node_val_neg_edge_sampler,
                                                                                           evaluate_data=new_node_val_data,
                                                                                           loss_func=loss_func,
                                                                                           num_neighbors=args.num_neighbors,
                                                                                           time_gap=args.time_gap)

            if args.model_name in ['TGN']:
                # reload validation memory bank for testing nodes or saving models
                # note that since model treats memory as parameters, we need to reload the memory to val_backup_memory_bank for saving models
                model[0].memory_bank.reload_memory_bank(val_backup_memory_bank)

            train_predict = torch.cat(train_predicts, dim=0)
            train_label = torch.cat(train_labels, dim=0)
            # standard_label = torch.tensor([0, 1], dtype=torch.float32, device=args.device)
            # predicts,labels = torch.concat((predicts,standard_label)),torch.concat((labels,standard_label))

            train_metrics.append(get_link_classification_metrics(predicts=train_predict, labels=train_label))

            logger.info(f'Epoch: {epoch + 1}, learning rate: {optimizer.param_groups[0]["lr"]}, train loss: {np.mean(train_losses):.4f}')
            for metric_name in train_metrics[0].keys():
                logger.info(f'train {metric_name}, {np.mean([train_metric[metric_name] for train_metric in train_metrics]):.4f}')
            logger.info(f'validate loss: {np.mean(val_losses):.4f}')
            for metric_name in val_metrics[0].keys():
                logger.info(f'validate {metric_name}, {np.mean([val_metric[metric_name] for val_metric in val_metrics]):.4f}')
            logger.info(f'new node validate loss: {np.mean(new_node_val_losses):.4f}')
            for metric_name in new_node_val_metrics[0].keys():
                logger.info(f'new node validate {metric_name}, {np.mean([new_node_val_metric[metric_name] for new_node_val_metric in new_node_val_metrics]):.4f}')

            if (epoch + 1) % args.test_interval_epochs == 0:
                test_losses, test_metrics = evaluate_model_link_classification(model_name=args.model_name,
                                                                           model=model,
                                                                           neighbor_sampler=full_neighbor_sampler,
                                                                           evaluate_idx_data_loader=test_idx_data_loader,
                                                                           evaluate_neg_edge_sampler=test_neg_edge_sampler,
                                                                           evaluate_data=test_data,
                                                                           loss_func=loss_func,
                                                                           num_neighbors=args.num_neighbors,
                                                                           time_gap=args.time_gap)

                if args.model_name in ['TGN']:
                    # reload validation memory bank for new testing nodes
                    model[0].memory_bank.reload_memory_bank(val_backup_memory_bank)

                new_node_test_losses, new_node_test_metrics = evaluate_model_link_classification(model_name=args.model_name,
                                                                                             model=model,
                                                                                             neighbor_sampler=full_neighbor_sampler,
                                                                                             evaluate_idx_data_loader=new_node_test_idx_data_loader,
                                                                                             evaluate_neg_edge_sampler=new_node_test_neg_edge_sampler,
                                                                                             evaluate_data=new_node_test_data,
                                                                                             loss_func=loss_func,
                                                                                             num_neighbors=args.num_neighbors,
                                                                                             time_gap=args.time_gap)

                if args.model_name in ['TGN']:
                    # reload validation memory bank for testing nodes or saving models
                    # note that since model treats memory as parameters, we need to reload the memory to val_backup_memory_bank for saving models
                    model[0].memory_bank.reload_memory_bank(val_backup_memory_bank)

                logger.info(f'test loss: {np.mean(test_losses):.4f}')
                for metric_name in test_metrics[0].keys():
                    logger.info(f'test {metric_name}, {np.mean([test_metric[metric_name] for test_metric in test_metrics]):.4f}')
                logger.info(f'new node test loss: {np.mean(new_node_test_losses):.4f}')
                for metric_name in new_node_test_metrics[0].keys():
                    logger.info(f'new node test {metric_name}, {np.mean([new_node_test_metric[metric_name] for new_node_test_metric in new_node_test_metrics]):.4f}')

            # select the best model based on all the validate metrics
            val_metric_indicator = []
            for metric_name in val_metrics[0].keys():
                val_metric_indicator.append((metric_name, np.mean([val_metric[metric_name] for val_metric in val_metrics]), True))
            early_stop = early_stopping.step(val_metric_indicator, model)

            if early_stop:
                break

        # load the best model
        early_stopping.load_checkpoint(model)

        # evaluate the best model
        logger.info(f'get final performance on dataset {args.dataset_name}...')

        # the saved best model of memory-based models cannot perform validation since the stored memory has been updated by validation data
        if args.model_name not in ['TGN']:
            val_losses, val_metrics = evaluate_model_link_classification(model_name=args.model_name,
                                                                     model=model,
                                                                     neighbor_sampler=full_neighbor_sampler,
                                                                     evaluate_idx_data_loader=val_idx_data_loader,
                                                                     evaluate_neg_edge_sampler=val_neg_edge_sampler,
                                                                     evaluate_data=val_data,
                                                                     loss_func=loss_func,
                                                                     num_neighbors=args.num_neighbors,
                                                                     time_gap=args.time_gap)

            new_node_val_losses, new_node_val_metrics = evaluate_model_link_classification(model_name=args.model_name,
                                                                                       model=model,
                                                                                       neighbor_sampler=full_neighbor_sampler,
                                                                                       evaluate_idx_data_loader=new_node_val_idx_data_loader,
                                                                                       evaluate_neg_edge_sampler=new_node_val_neg_edge_sampler,
                                                                                       evaluate_data=new_node_val_data,
                                                                                       loss_func=loss_func,
                                                                                       num_neighbors=args.num_neighbors,
                                                                                       time_gap=args.time_gap)

        if args.model_name in ['TGN']:
            # the memory in the best model has seen the validation edges, we need to backup the memory for new testing nodes
            val_backup_memory_bank = model[0].memory_bank.backup_memory_bank()

        test_losses, test_metrics = evaluate_model_link_classification(model_name=args.model_name,
                                                                   model=model,
                                                                   neighbor_sampler=full_neighbor_sampler,
                                                                   evaluate_idx_data_loader=test_idx_data_loader,
                                                                   evaluate_neg_edge_sampler=test_neg_edge_sampler,
                                                                   evaluate_data=test_data,
                                                                   loss_func=loss_func,
                                                                   num_neighbors=args.num_neighbors,
                                                                   time_gap=args.time_gap)

        if args.model_name in ['TGN']:
            # reload validation memory bank for new testing nodes
            model[0].memory_bank.reload_memory_bank(val_backup_memory_bank)

        new_node_test_losses, new_node_test_metrics = evaluate_model_link_classification(model_name=args.model_name,
                                                                                     model=model,
                                                                                     neighbor_sampler=full_neighbor_sampler,
                                                                                     evaluate_idx_data_loader=new_node_test_idx_data_loader,
                                                                                     evaluate_neg_edge_sampler=new_node_test_neg_edge_sampler,
                                                                                     evaluate_data=new_node_test_data,
                                                                                     loss_func=loss_func,
                                                                                     num_neighbors=args.num_neighbors,
                                                                                     time_gap=args.time_gap)
        # store the evaluation metrics at the current run
        val_metric_dict, new_node_val_metric_dict, test_metric_dict, new_node_test_metric_dict = {}, {}, {}, {}

        if args.model_name not in ['TGN']:
            logger.info(f'validate loss: {np.mean(val_losses):.4f}')
            for metric_name in val_metrics[0].keys():
                average_val_metric = np.mean([val_metric[metric_name] for val_metric in val_metrics])
                logger.info(f'validate {metric_name}, {average_val_metric:.4f}')
                val_metric_dict[metric_name] = average_val_metric

            logger.info(f'new node validate loss: {np.mean(new_node_val_losses):.4f}')
            for metric_name in new_node_val_metrics[0].keys():
                average_new_node_val_metric = np.mean([new_node_val_metric[metric_name] for new_node_val_metric in new_node_val_metrics])
                logger.info(f'new node validate {metric_name}, {average_new_node_val_metric:.4f}')
                new_node_val_metric_dict[metric_name] = average_new_node_val_metric

        logger.info(f'test loss: {np.mean(test_losses):.4f}')
        for metric_name in test_metrics[0].keys():
            average_test_metric = np.mean([test_metric[metric_name] for test_metric in test_metrics])
            logger.info(f'test {metric_name}, {average_test_metric:.4f}')
            test_metric_dict[metric_name] = average_test_metric

        logger.info(f'new node test loss: {np.mean(new_node_test_losses):.4f}')
        for metric_name in new_node_test_metrics[0].keys():
            average_new_node_test_metric = np.mean([new_node_test_metric[metric_name] for new_node_test_metric in new_node_test_metrics])
            logger.info(f'new node test {metric_name}, {average_new_node_test_metric:.4f}')
            new_node_test_metric_dict[metric_name] = average_new_node_test_metric

        single_run_time = time.time() - run_start_time
        logger.info(f'Run {run + 1} cost {single_run_time:.2f} seconds.')

        if args.model_name not in ['TGN']:
            val_metric_all_runs.append(val_metric_dict)
            new_node_val_metric_all_runs.append(new_node_val_metric_dict)
        test_metric_all_runs.append(test_metric_dict)
        new_node_test_metric_all_runs.append(new_node_test_metric_dict)

        # avoid the overlap of logs
        if run < args.num_runs - 1:
            logger.removeHandler(fh)
            logger.removeHandler(ch)

        # save model result
        if args.model_name not in ['TGN']:
            result_json = {
                "validate metrics": {metric_name: f'{val_metric_dict[metric_name]:.4f}' for metric_name in val_metric_dict},
                "new node validate metrics": {metric_name: f'{new_node_val_metric_dict[metric_name]:.4f}' for metric_name in new_node_val_metric_dict},
                "test metrics": {metric_name: f'{test_metric_dict[metric_name]:.4f}' for metric_name in test_metric_dict},
                "new node test metrics": {metric_name: f'{new_node_test_metric_dict[metric_name]:.4f}' for metric_name in new_node_test_metric_dict}
            }
        else:
            result_json = {
                "test metrics": {metric_name: f'{test_metric_dict[metric_name]:.4f}' for metric_name in test_metric_dict},
                "new node test metrics": {metric_name: f'{new_node_test_metric_dict[metric_name]:.4f}' for metric_name in new_node_test_metric_dict}
            }
        result_json = json.dumps(result_json, indent=4)

        save_result_folder = f"./saved_results/{args.model_name}/{args.dataset_name}"
        os.makedirs(save_result_folder, exist_ok=True)
        save_result_path = os.path.join(save_result_folder, f"{args.save_model_name}.json")

        with open(save_result_path, 'w') as file:
            file.write(result_json)

    # store the average metrics at the log of the last run
    logger.info(f'metrics over {args.num_runs} runs:')

    if args.model_name not in ['TGN']:
        for metric_name in val_metric_all_runs[0].keys():
            logger.info(f'validate {metric_name}, {[val_metric_single_run[metric_name] for val_metric_single_run in val_metric_all_runs]}')
            logger.info(f'average validate {metric_name}, {np.mean([val_metric_single_run[metric_name] for val_metric_single_run in val_metric_all_runs]):.4f} '
                        f'± {np.std([val_metric_single_run[metric_name] for val_metric_single_run in val_metric_all_runs], ddof=1):.4f}')

        for metric_name in new_node_val_metric_all_runs[0].keys():
            logger.info(f'new node validate {metric_name}, {[new_node_val_metric_single_run[metric_name] for new_node_val_metric_single_run in new_node_val_metric_all_runs]}')
            logger.info(f'average new node validate {metric_name}, {np.mean([new_node_val_metric_single_run[metric_name] for new_node_val_metric_single_run in new_node_val_metric_all_runs]):.4f} '
                        f'± {np.std([new_node_val_metric_single_run[metric_name] for new_node_val_metric_single_run in new_node_val_metric_all_runs], ddof=1):.4f}')

    for metric_name in test_metric_all_runs[0].keys():
        logger.info(f'test {metric_name}, {[test_metric_single_run[metric_name] for test_metric_single_run in test_metric_all_runs]}')
        logger.info(f'average test {metric_name}, {np.mean([test_metric_single_run[metric_name] for test_metric_single_run in test_metric_all_runs]):.4f} '
                    f'± {np.std([test_metric_single_run[metric_name] for test_metric_single_run in test_metric_all_runs], ddof=1):.4f}')

    for metric_name in new_node_test_metric_all_runs[0].keys():
        logger.info(f'new node test {metric_name}, {[new_node_test_metric_single_run[metric_name] for new_node_test_metric_single_run in new_node_test_metric_all_runs]}')
        logger.info(f'average new node test {metric_name}, {np.mean([new_node_test_metric_single_run[metric_name] for new_node_test_metric_single_run in new_node_test_metric_all_runs]):.4f} '
                    f'± {np.std([new_node_test_metric_single_run[metric_name] for new_node_test_metric_single_run in new_node_test_metric_all_runs], ddof=1):.4f}')

    sys.exit()