HumanCompatibleAI / imitation

Clean PyTorch implementations of imitation and reward learning algorithms
https://imitation.readthedocs.io/
MIT License
1.27k stars 241 forks source link

Save reward_net's checkpoint? #698

Open Liuzy0908 opened 1 year ago

Liuzy0908 commented 1 year ago

Problem

Hello! Thanks so much for sharing the code! There is a question.

I want to load the trained learner net and the trained reward_net to train learner and reward_net again. So how do I save reward_net's checkpoint after the first training GAIL?

Thanks ;).

ernestum commented 1 year ago

Hi @Liuzy0908, thanks for making us aware of the lack of documentation here. We have an an overhaul of the documentation on reward networks planned in #714 Until then maybe have a look at this function where the reward net of an adversarial trainer is stored.