What is the Proper Model Save Method?

[CerebralSeed]

The example given in the documentation shows downloading and archiving a pre-existing model from Pytorch. But if serving a custom-built model, what is the correct save method?

For example, on the Save/Loading Documentation, there are several save methods:

https://pytorch.org/tutorials/beginner/saving_loading_models.html

Should the model artifacts be saved as simply torch.save(model, PATH) or should it be saved as torch.save(model.state_dict(), PATH)?

Thank you.

[CerebralSeed]

Upon further inspection, I see more information is given in the examples: https://github.com/pytorch/serve/tree/master/examples/image_classifier/densenet_161