多个epoch结果

[Zhangmanman-ict]

为什么训练了多个epoch,结果没有变化，和训练一轮一样

[taishan1994]

为什么训练了多个epoch,结果没有变化，和训练一轮一样

可能之前framework/framework.py里面我限制了：

if epoch == 0:
        eval_start_time = time.time()
        model.eval()
        # call the test function
        precision, recall, f1_score = self.test(test_data_loader, model, current_f1=best_f1_score, output=self.config.result_save_name)
        
        self.logging('epoch {:3d}, eval time: {:5.2f}s, f1: {:4.3f}, precision: {:4.3f}, recall: {:4.3f}'.
                     format(epoch, time.time() - eval_start_time, f1_score, precision, recall))
        
        if f1_score > best_f1_score:
            best_f1_score = f1_score
            best_epoch = epoch
            best_precision = precision
            best_recall = recall
            self.logging("saving the model, epoch: {:3d}, precision: {:4.3f}, recall: {:4.3f}, best f1: {:4.3f}".
                         format(best_epoch, best_precision, best_recall, best_f1_score))
            # save the best model
            path = os.path.join(self.config.checkpoint_dir, self.config.model_save_name)
            if not self.config.debug:
                torch.save(ori_model.state_dict(), path)
        
model.train()

你可以把这里改一下。