Save model parameters on each checkpoint

Ray Libraries (Data, Train, Tune, Serve) Ray Tune
21

Hi! I’m facing a similar problem in that I want to be able to save a policy to then loading it in another process.
After exporting the policy with trainer.export_policy_model("/tmp/policy") , how can we load that policy into a new trainer? Ideally, I would want to be able to do something like trainer.compute_single_action(...)

22

Hi @ik12 .

Maybe this example can help you, if you have not already solved the problem. Depending on ray version some adjustments may be needed

BR

Jorgen