How should I save and load objects in
NeuralPosterior objects are picklable.
import pickle # ... run inference posterior = inference.build_posterior() with open("/path/to/my_posterior.pkl", "wb") as handle: pickle.dump(posterior, handle)
Note: posterior objects that were saved under
sbi v0.17.2 or older can not be loaded under
sbi v0.18.0 or newer.
Note: if you try to load a posterior that was saved under
sbi v0.14.x or earlier under
sbi v0.15.x until
sbi v0.17.x, you have to add:
import sys from sbi.utils import user_input_checks_utils sys.modules["sbi.user_input.user_input_checks_utils"] = user_input_checks_utils
NeuralInference objects are also picklable.
import pickle # ... run inference posterior = inference.build_posterior() with open("/path/to/my_inference.pkl", "wb") as handle: pickle.dump(inference, handle)
inferenceobject will slightly modify the object (in order to make it serializable). These modifications lead to the following two changes in behaviour: 1) Retraining from scratch is not supported, i.e.
.train(..., retrain_from_scratch=True)does not work. 2) When the loaded object calls the
.train()method, it generates a new tensorboard summary writer (instead of appending to the current one).