Serialization

[bhardwaj-gopika]

This pull request adds full serialization and deserialization capabilities to the LUMETorchModel wrapper class, enabling models to be saved to and loaded from YAML configuration files.

from lume_torch.base import LUMETorch, LUMETorchModel
from lume_torch.models import TorchModel
model = TorchModel("model_config.yaml")
ltmodel = LUMETorchModel(model)
ltmodel.dump("my_lume_torch_model.yaml")
loaded_model = LUMETorchModel.from_file("my_lume_torch_model.yaml")
loaded_model.set({"ACCL:LI25:1:ADES": 6260.0})
print(loaded_model.get(loaded_model.torch_model.output_names))
{'hxr_pulse_intensity': tensor(1.5391, dtype=torch.float64)}

[roussel-ryan]

This looks good to me, @pluflou can also review. One thing to consider is that soon^(TM) we will add serialization / deserialization logic to the LUMEModel base class, which should handle a lot of this logic, so be prepared to move a lot of this to lume-base when the time is right

[pluflou]

Looks good to me as well.