num_ops argument of RandAugment() shouldn't accept negative values

[hyperkai]

🐛 Describe the bug

Setting negative values to num_ops argument of RandAugment() doesn't do augmentation transformations at all as shown below:

from torchvision.datasets import OxfordIIITPet
from torchvision.transforms.v2 import RandAugment

my_data = OxfordIIITPet(
    root="data",
    transform=RandAugment(num_ops=-1)
    # transform=RandAugment(num_ops=-10)
    # transform=RandAugment(num_ops=-100)
)

my_data[0][0]

And, num_ops argument is the number of augmentation transformations according to the doc as shown below:

Parameters:

• num_ops (int, optional) – Number of augmentation transformations to apply sequentially

So num_ops argument of RandAugment() shouldn't accept negative values.

Versions

import torchvision

torchvision.__version__ # '0.20.1'

[NicolasHug]

Thanks for the report, feel free to submit a PR for this