How can I save model parameters? I want to load the pre-trained model by myself in other experiments

[Phoebe-ovo]

I observed that: the checkpoint.pt only save: report, study, trail_reports, timer, random_state. I want to save the model parameters which can be loaded by anyone else.

[Yura52]

We don't save all model checkpoints during hyperparameter tuning, because that would consume significant amount of memory.

How to implement:

• rename trial_reports to trial_results
• here, do trial_results.append((report, model_state_dict)) where model_state_dict is extracted from checkpoint.pt located in the dir_ directory here
• change the callback accordingly

[Phoebe-ovo]

Thank you!