How can I save a predictor for using it again

[MarleneBs]

On the page it is said with Path how to save a predictor, but then also comes the warning that you can no longer use it somehow. Can you save the model so that you can open it in another notebook and use it with other data?

[jaheba]

Hi, can you share the warning?

Generally, you can use predictor.serialize(path) to store the predictor, and Predictor.deserialize(path) to load it again.

Here is an example:

gluonts/test/mx/model/simple_feedforward/test_serde.py

Lines 50 to 52 in 4f423ad

	predictor.serialize(model_path)
	new_symbol_predictor = Predictor.deserialize(model_path)

[MarleneBs]

WARNING: root: Serializing RepresentableBlockPredictor instances does not save the prediction network stucture in a backwards-compatible manner. Be careful not to use this method in production