多 GPU 测试：类型错误：“int”对象不可迭代

Question

将深度迁移学习单GPU训练代码修改为多GPU训练时，代码测试部分出现如下错误：

Traceback (most recent call last):
  File "../main.py", line 155, in <module>
    main(args)
  File "../main.py", line 57, in main
    t_correct = test(args, model, tar_test_loader, cuda_stat)
  File "../main.py", line 136, in test
    s_output, _, _  = model(data, data, target)
  File "/home/maxin2/project/anaconda3/lib/python3.8/site-packages/torch/nn/modules/module.py", line 889, in _call_impl
    result = self.forward(*input, **kwargs)
  File "/home/maxin2/project/anaconda3/lib/python3.8/site-packages/torch/nn/parallel/data_parallel.py", line 168, in forward
    return self.gather(outputs, self.output_device)
  File "/home/maxin2/project/anaconda3/lib/python3.8/site-packages/torch/nn/parallel/data_parallel.py", line 180, in gather
    return gather(outputs, output_device, dim=self.dim)
  File "/home/maxin2/project/anaconda3/lib/python3.8/site-packages/torch/nn/parallel/scatter_gather.py", line 76, in gather
    res = gather_map(outputs)
  File "/home/maxin2/project/anaconda3/lib/python3.8/site-packages/torch/nn/parallel/scatter_gather.py", line 71, in gather_map
    return type(out)(map(gather_map, zip(*outputs)))
  File "/home/maxin2/project/anaconda3/lib/python3.8/site-packages/torch/nn/parallel/scatter_gather.py", line 71, in gather_map
    return type(out)(map(gather_map, zip(*outputs)))
TypeError: 'int' object is not iterable

我找不到错误的原因。以下是我的测试代码的一部分：

def test(args, model, target_test_loader, cuda_stat):
    model.eval()
    test_loss = 0
    correct = 0
    criterion = torch.nn.CrossEntropyLoss(reduction='sum')
    len_target_dataset = len(target_test_loader.dataset)
    with torch.no_grad():
        for data, target in target_test_loader:
            if cuda_stat:
                data, target = data.cuda(), target.cuda()
            s_output, _, _  = model(data, data, target)
            test_loss += criterion(s_output, target)# sum up batch loss
            pred = torch.max(s_output, 1)[1]  # get the index of the max log-probability
            print(pred)
            correct += torch.sum(pred == target)
        test_loss /= len_target_dataset
        print(args.test_dir, '  Test set:Loss: {} {:.6f}  Accuracy: {}/{} ({:.0f}%)\n'.format(
            test_loss, correct, len_target_dataset,
            100. * correct / len_target_dataset))
    return correct