composite priors do not work in stores

[elidub]

The composite priors defined as in the tutorial notebook don't work in stores. Changing priors in tutorial notebooks with

from swyft.utils import tensor_to_array, array_to_tensor
from toolz import compose

uniform = torch.distributions.Uniform(array_to_tensor(low), array_to_tensor(high))
prior = swyft.Prior(
        cdf = compose(tensor_to_array, lambda x: uniform.cdf(x), array_to_tensor),
        icdf = compose(tensor_to_array, lambda x: uniform.icdf(x), array_to_tensor),
        log_prob = compose(tensor_to_array, lambda x: uniform.log_prob(x), array_to_tensor),
        n_parameters = n_parameters,
    )

Gives the following error: PicklingError: Can't pickle <function <lambda> at 0x1530b34e7e50>: attribute lookup <lambda> on __main__ failed

[bkmi]

Thank you for reporting the issue. I'll fix it before the 0.3.0 release.

The reason for this is due to how pickle works. (It cannot pickle lambda functions)

[bkmi]

I created a solution to this issue and committed it directly to master c24ce4c5f50e8ebe56eb90e5527c7819fcf4dbc0.

[elidub]

Works great!