Setting `5.3` and `9` to `bits` argument of `RandomPosterize()` gets an indirect error message

[hyperkai]

🐛 Describe the bug

Setting 5.3 and 9 to bits argument of RandomPosterize() gets the indirect error message as shown below:

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

my_data1 = OxfordIIITPet(
    root="data",                 # ↓↓↓
    transform=RandomPosterize(bits=5.3)
)

my_data2 = OxfordIIITPet(
    root="data",                 # ↓
    transform=RandomPosterize(bits=9)
)

my_data1[0] # Error
my_data2[0] # Error

TypeError: bad operand type for unary ~: 'float'

TypeError: bad operand type for unary ~: 'float'

So, they should be something direct like below respectively:

TypeError: bits must be int

ValueError: bits must be [0, 8]

Versions

import torchvision

torchvision.__version__ # '0.20.1'

[NicolasHug]

happy to consider a PR @hyperkai , thanks