.transforms.ToTensor() Issue on Datasets.Cityscapes

Hello Forum,

Im attempting to implement the Cityscapes dataset using torchvision.datasets.Cityscapes, but I keep getting errors when I try to load 2 labels. Loading a single label works fine. However, from my understanding, datasets.Cityscapes would support this. I have tried the following variants:

train_transform = transforms.ToTensor()
    train_dataset = datasets.Cityscapes('./data', split='train', mode='fine',
                     target_type='semantic', transform=train_transform, target_transform=train_transform)
    train_loader = DataLoader(dataset=train_dataset, batch_size = mini_batch_size, shuffle=shuffle, **kwargs)

→ Works as intended, I get my image and the semantic label

    train_dataset = datasets.Cityscapes('./data', split='train', mode='fine',
                     target_type=['semantic','instance], transform=train_transform, target_transform=train_transform)
    train_loader = DataLoader(dataset=train_dataset, batch_size = mini_batch_size, shuffle=shuffle, **kwargs)

TypeError: pic should be PIL Image or ndarray. Got <class ‘tuple’>

How can I apply the transform to both labels?

Thank you in advance.
L