how to re-use same object on different parts of a LazyCall object

[aldopareja]

When using LazyCall and instantiate an instantiation creates different objects for each new target. Sometimes it is helpful to reuse the object on different parts of a higher scope. A good example would be siamese networks or self-play.

This code could be a good example. Here, the model for the visualizer would be a different object than for the trainer at instantiation time.

import LazyCall as L
from torch.nn import Linear

my_model = L(Linear)(
     in_features = 10,
     out_features = 10
)

visualizer = L(my_visualizer)(
     model = my_model,
)

trainer = L(my_trainer)(
    model = my_model,
)

I could solve this by having an instantiate at a wrapper class and then reusing the instantiated object. Still, I think instantiating an object and making it transparent for the user would be better.