Setting the value of `x <= 0` to `alpha` argument of `AugMix()` doesn't get error even if `Dirichlet()` is used for it

[hyperkai]

🐛 Describe the bug

Dirichlet() is used for alpha argument of AugMix(), then setting the value of x <= 0 to concentration argument of Dirichlet() gets error as shown below:

from torch.distributions.dirichlet import Dirichlet
import torch

Dirichlet(concentration=torch.tensor([0.0])) # Error
Dirichlet(concentration=torch.tensor([-1.0])) # Error
Dirichlet(concentration=torch.tensor([-2.0])) # Error

But setting the value of x <= 0 to alpha argument of AugMix() doesn't get error as shown below:

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

my_data = OxfordIIITPet(
    root="data",
    transform=AugMix(alpha=0.0)
    # transform=AugMix(alpha=-1.0)
    # transform=AugMix(alpha=-2.0)
)

my_data[0]
# (<PIL.Image.Image image mode=RGB size=394x500>, 0)

So, setting the value of x <= 0 to alpha argument of AugMix() should get error same as concentration argument of Dirichlet().

Versions

import torchvision

torchvision.__version__ # '0.20.1'