How to load data using torchmeta without any class-per-task break up?

vision
1

I want to use the tiered-imagenet data for an image classification problem. I want to use the whole data together and don’t want to perform few-shot learning. The tiered imagenet data is available directly on torchmeta and I would like to use it. How do I load the data without any break up of the dataset?

    dataset = torchmeta.datasets.TieredImagenet(root, num_classes_per_task=1,
    meta_train=False, meta_val=False, meta_test=False, meta_split=mode,
    transform=None, target_transform=None, dataset_transform=None,
    class_augmentations=None, download=True)

However, even this is creating errors. I checked the documentation and seems like the num_classes_per_task is mandatory. How do I resolve this?

Thanking you in advance!