[学习过程中的问题记录1]None 自定义Dataset类

1.None 空值，如果对一个对象判断是否存在就用 is None/is not None, 而不是 != None
2. 自定义dataset类时候的数据转换问题，自定义dataset类后，对__getitem__(self, index)函数重载，将数据转换为tensor时候在该函数内进行报错
错误原因：第一次使用__getitem__(self, index)时，所有对象都被转换成tensor类了，后续再次调用时候相当于把已经由numpy类转换成的tensor类再去转换为tensor，指定报错。
修改后的代码：

class myDataset(torch.utils.data.Dataset):
    def __init__(self, feature,clean_labels,noise_labels = None ):
        self.feature = feature
        self.clean_labels = clean_labels
        # 将数据转化为Tensor格式
        self.feature = torch.Tensor(self.feature).type(torch.float32)
        self.clean_labels = torch.from_numpy(self.clean_labels).type(torch.LongTensor)

        if noise_labels is not None:
            self.noise_labels = noise_labels
            self.noise_labels = torch.from_numpy(self.noise_labels).type(torch.LongTensor)
        else:
            self.noise_labels = None

    def __len__(self):
        return len(self.clean_labels)

    def __getitem__(self, index):
        x = self.feature[index]
        clean_y = self.clean_labels[index]
        #如果noise 标签存在就将noise标签返回
        if self.noise_labels is not None:
            noise_y = self.noise_labels[index]
            return x, [clean_y,noise_y]
        else:
            return x, clean_y