当前位置: 代码迷 >> 综合 >> 文本分类(二) | (5) 训练、验证以及测试
  详细解决方案

文本分类(二) | (5) 训练、验证以及测试

热度:16   发布时间:2024-01-15 00:32:45.0

完整项目???????

最后,介绍一下模型的训练、验证以及测试流程。

目录

1. 训练

2. 验证

3. 测试


1. 训练

def train(config, model, train_iter, dev_iter, test_iter):start_time = time.time()model.train() #训练模式optimizer = torch.optim.Adam(model.parameters(), lr=config.learning_rate)#Adam优化器# 学习率指数衰减,每过一个epoch:学习率 = gamma * 学习率 也可以采用达观杯中的方式,当模型在验证集上指标不再提高时,学习率减半# scheduler = torch.optim.lr_scheduler.ExponentialLR(optimizer, gamma=0.9)total_batch = 0  # 记录进行到多少batchdev_best_loss = float('inf') #跟踪验证集最小的loss 或 最大f1-score或准确率last_improve = 0  # 记录上次验证集loss下降的batch数flag = False  # 记录是否很久没有效果提升#可视化 tensorBoardwriter = SummaryWriter(log_dir=config.log_path + '/' + time.strftime('%m-%d_%H.%M', time.localtime()))for epoch in range(config.num_epochs):print('Epoch [{}/{}]'.format(epoch + 1, config.num_epochs))# scheduler.step() # 学习率衰减for i, (trains, labels) in enumerate(train_iter):outputs = model(trains) #前向传播 获取输出model.zero_grad()  #清空梯度loss = F.cross_entropy(outputs, labels) #计算交叉熵损失(内部包含softmax log等操作) 可以用nn里面的函数 也可以用F中的函数 labels为整数索引 内部会自动转换为one-hotloss.backward() #计算梯度optimizer.step() #更新参数if total_batch % 100 == 0: #每100个batch 计算一下在验证集上的指标 或者像之前项目中那样 每一个epoch在验证集上计算相关指标true = labels.data.cpu()predic = torch.max(outputs.data, 1)[1].cpu()train_acc = metrics.accuracy_score(true, predic) #当前batch上训练集的准确率 因为是类别均衡数据集,所以可以直接用准确率作评估指标 不然使用 macro f1-scoredev_acc, dev_loss = evaluate(config, model, dev_iter) #计算此时模型在验证集上的损失和准确率if dev_loss < dev_best_loss:dev_best_loss = dev_loss#更新验证集最小损失torch.save(model.state_dict(), config.save_path) #保存在验证集上损失最小的参数improve = '*' #效果提升标志last_improve = total_batch #计算上次提升 位于哪个batchelse:improve = ''time_dif = get_time_dif(start_time)msg = 'Iter: {0:>6},  Train Loss: {1:>5.2},  Train Acc: {2:>6.2%},  Val Loss: {3:>5.2},  Val Acc: {4:>6.2%},  Time: {5} {6}'print(msg.format(total_batch, loss.item(), train_acc, dev_loss, dev_acc, time_dif, improve))#保存 训练集(当前batch)、验证集的损失和准确率信息 方便可视化 以batch为单位writer.add_scalar("loss/train", loss.item(), total_batch) writer.add_scalar("loss/dev", dev_loss, total_batch)writer.add_scalar("acc/train", train_acc, total_batch)writer.add_scalar("acc/dev", dev_acc, total_batch)model.train()#回到训练模式total_batch += 1if total_batch - last_improve > config.require_improvement: #如果长期没有提高 就提前终止# 验证集loss超过1000batch没下降,结束训练print("No optimization for a long time, auto-stopping...")flag = Truebreakif flag:breakwriter.close()test(config, model, test_iter) #模型训练结束后 进行测试

2. 验证

def evaluate(config, model, data_iter, test=False):model.eval() #测试模式loss_total = 0predict_all = np.array([], dtype=int) #存储验证集所有batch的预测结果labels_all = np.array([], dtype=int) #存储验证集所有batch的真实标签with torch.no_grad():for texts, labels in data_iter:outputs = model(texts)loss = F.cross_entropy(outputs, labels)loss_total += losslabels = labels.data.cpu().numpy()predic = torch.max(outputs.data, 1)[1].cpu().numpy()labels_all = np.append(labels_all, labels)predict_all = np.append(predict_all, predic)acc = metrics.accuracy_score(labels_all, predict_all) #计算验证集准确率if test: #如果是测试集的话 计算一下分类报告和混淆矩阵report = metrics.classification_report(labels_all, predict_all, target_names=config.class_list, digits=4)confusion = metrics.confusion_matrix(labels_all, predict_all) #计算混淆矩阵return acc, loss_total / len(data_iter), report, confusionreturn acc, loss_total / len(data_iter) #返回准确率和每个batch的平均损失

3. 测试

def test(config, model, test_iter):# testmodel.load_state_dict(torch.load(config.save_path)) #加载使验证集损失最小的参数model.eval() #测试模式start_time = time.time()test_acc, test_loss, test_report, test_confusion = evaluate(config, model, test_iter, test=True) #计算测试集准确率,每个batch的平均损失 分类报告、混淆矩阵msg = 'Test Loss: {0:>5.2},  Test Acc: {1:>6.2%}'print(msg.format(test_loss, test_acc))print("Precision, Recall and F1-Score...")print(test_report)print("Confusion Matrix...")print(test_confusion)time_dif = get_time_dif(start_time)print("Time usage:", time_dif)

 

  相关解决方案