How to save policy model?

I wonder how can I save trained tf2 model - policy to use it.
I used code: “rllib train --run=PPO --env=CartPole-v0”
The training works fine but where the best policy is saved??

Peter Pirog

Hello, you need to setup the config parameter checkpoint_at_end=True , so it will be stored at the end of the training process by default in Ubuntu is stored in: “/home/user/ray_results”.

1 Like

Thank you for the answer. I will try it.