Loading a custom trained checkpoint/state_dict for inference/testing

[christianjosef27]

Hello,

can someone explain me how to do:

I have custom trained a model and used the save_state_dict argument in the Trainer Function.
So I have folders of different checkpoints generated by the Trainer. But how to use a saved checkpoint for model inference?

Do I need to pass the name of my checkpoint folder to the checkpoint argument of the get_aspect_extractor() method?
Or how does it work?

Thank you in advance