I wish to to separate the training logic (which is encapsulated in the Lightning Module) and the actual model. Thus I want to pass a model to the initializer. Is it possible to do this with lightning cli and a yaml file?
I am imagining something like this:
model:
class_path: path_to_lightning_module.MyLightningModule
init_args:
model :
class_path: path_to_model.MyModel
However this does not work. Is there another way to achieve this?