How to change available resources when restoring a checkpoint?


I am training a PPO agent on a remote machine where GPUs are available. I transferred the resulting checkpoints to my local machine where there are no GPUs available. Restoring the policy using ppo.PPO.from_checkpoint(checkpoint=checkpoint_path) fails because it cannot find GPUs. How can I change the allocated resources in the call so I can e.g. evaluate on my local device?

Thank you,