CUDA error: initialization errorCUDA kernel errors

第一次玩神经网络，之前环境没装好都是用CPU跑的，这次试了下用GPU，发现一直报错提示RuntimeError: CUDA error: initialization error CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1.

解决办法：调用GPU的代码，一定要放到创建好的Dataloader后！

 for fold, (train_indices, val_indices) in enumerate(kfold.split(train_dataset)):
    # 根据当前折划分训练集和验证集
        train_dataset_fold = torch.utils.data.Subset(train_dataset, train_indices)
        val_dataset_fold = torch.utils.data.Subset(train_dataset, val_indices)
    
    # 创建 DataLoader
        train_loader_fold = DataLoader(train_dataset_fold, batch_size=50,shuffle=True)
        val_loader_fold = DataLoader(val_dataset_fold, batch_size=50, shuffle=False)
    
        device = torch.device("cuda:1" if torch.cuda.is_available() else "cpu")
    
 同个cell里如果还有外部验证集的Dataloader也不行，会报错，一定要把它放到另一个cell里运行（楼主用的是jupyterlab），同样也是放到Dataloader后！！！