`get_vars_in_point_list` should ignore variables that are not in the model

[lucianopaz]

Description of your problem

This thread pointed out a bug in get_vars_in_point_list.

with pm.Model() as model:
    a = pm.Normal(“a”,0,1)
    b = pm.Normal(“b”,0,1)

vars_in_trace = pm.sampling.get_vars_in_point_list(
    [{“a”:0, “c”:0}],
    model,
)

This should return a list with a, but raises a KeyError instead because ”c” is not in model.

The solution is simple, just include the tensors that are in the model and in the trace, ignoring the variables in the trace that yield KeyError