How to test a policy on unseen test samples

[travek]

Hello!

I'm new in Reinforcement Learning and I studied RLLib User Guide and well as examples included into RLLib.
All learning examples are ended like that:

Simulator* sim = new Simulator(agent, problem, 5000, 100, 10);
sim->setTestEpisodesAfterEachRun(true);
sim->run();
sim->computeValueFunction();

I had some experience of using Supervised learning functionality. In most cases we construct a model and train it on Train samples and then test on Testing samples. Test and Train samples are different. And this is done to estimate how well the model behave on unseen data.

Could someone advise how RLLib based code should be structured to train a policy on train data and then to feed one by one test samples and estimate the policy on unseen samples.

Thanks

[travek]

No feedback... closing