AloTransform parameters sampling applied twice

[many28]

In AloTransform init function, we initialize the parameters by calling the sample_params() function :

aloception-oss/alodataset/transforms.py

Line 35 in 1e8409e

self.sample_params()

However it's already called in the _call_() function, for each scenario case of the input frames structure:

aloception-oss/alodataset/transforms.py

Line 58 in 1e8409e

def __call__(self, frames: Union[Mapping[str, Frame], List[Frame], Frame], **kwargs):

Which leads to double initialization of the parameters for each instance, if the parameters are chosen randomly, it may rise conflicts of using the two different values of the parameters in the same appliance of the transform, possible inheritance conflicts between other transform classes that come from AloTransform.