Method main

code/run.py:459–503

(self)

    def main(self):
        if self.mode_eval == "nocv":
            self.model = self.get_model()
            dataloaders = self.get_dataloader(data_type=self.data_type, data_fold=self.fold)
            trainer = Trainer(model=self.model, device = self.device, lr = self.lr, dataloaders = dataloaders, epoches = self.epoches, dropout = self.dropout, weight_decay = self.weight_decay, mode = self.mode, model_name = self.model_name, event_num = self.event_num, 
                    epoch_stop = self.epoch_stop, save_param_path = self.save_param_dir+self.data_type+"/"+self.model_name+"/", writer = SummaryWriter(self.path_tensorboard))
            result=trainer.train()

        elif self.mode_eval == "temporal":
            self.model = self.get_model()
            dataloaders = self.get_dataloader_temporal(data_type=self.data_type)
            trainer = Trainer3(model=self.model, device = self.device, lr = self.lr, dataloaders = dataloaders, epoches = self.epoches, dropout = self.dropout, weight_decay = self.weight_decay, mode = self.mode, model_name = self.model_name, event_num = self.event_num, 
                    epoch_stop = self.epoch_stop, save_param_path = self.save_param_dir+self.data_type+"/"+self.model_name+"/", writer = SummaryWriter(self.path_tensorboard))
            result=trainer.train()
            return result
        
        elif self.mode_eval == "cv":
            collate_fn=None
            if self.model_name == 'TextCNN':
                wv_from_text = KeyedVectors.load_word2vec_format("./stores/tencent-ailab-embedding-zh-d100-v0.2.0-s/tencent-ailab-embedding-zh-d100-v0.2.0-s.txt", binary=False)

            history = collections.defaultdict(list) 
            for fold in range(1, 6): 
                print('-' * 50)
                print ('fold %d:' % fold)
                print('-' * 50)
                self.model = self.get_model()
                dataloaders = self.get_dataloader(data_type=self.data_type, data_fold=fold)
                trainer = Trainer(model = self.model, device = self.device, lr = self.lr, dataloaders = dataloaders, epoches = self.epoches, dropout = self.dropout, weight_decay = self.weight_decay, mode = self.mode, model_name = self.model_name, event_num = self.event_num, 
                    epoch_stop = self.epoch_stop, save_param_path = self.save_param_dir+self.data_type+"/"+self.model_name+"/", writer = SummaryWriter(self.path_tensorboard+"fold_"+str(fold)+"/"))
       
                result = trainer.train()

                history['auc'].append(result['auc'])
                history['f1'].append(result['f1'])
                history['recall'].append(result['recall'])
                history['precision'].append(result['precision'])
                history['acc'].append(result['acc'])
                
            print ('results on 5-fold cross-validation: ')
            for metric in ['acc', 'f1', 'precision', 'recall', 'auc']:
                print ('%s : %.4f +/- %.4f' % (metric, np.mean(history[metric]), np.std(history[metric])))
        
        else:
            print ("Not Available")