Update only partial of the model, reusing the rest of the model during model training

For now, the whole LabelStudioMLBase instance is rebuild each time a new model is trained. However, in some scenarios, only a small part of the whole model is updated, and we want to reuse the rest of the model. Therefore, it is not needed to rebuild a new LabelStudioMLBase instance in these scenarios.

Is there a way to reuse the same LabelStudioMLBase instance after model training?