Question about agent.eval_mode()

[xylee95]

How do I switch back the agent to training mode after using with agent.eval_mode()?

I'm following the quick-start example and I want to evaluate my agent intermittently during training, rather than at the end of the training. What is the proper way of returning the agent to training setting after running a certain number of episodes of evaluation?

I'm writing my own custom training loop rather than using the pre-defined experiments function.

[muupan]

The agent will switch back to the original mode automatically once it leaves the with agent.eval_mode() context.

print("training", agent.training)
with agent.eval_mode():
    print("training", agent.training)
print("training", agent.training)

will print

training True
training False
training True

[xylee95]

Great, thank you for clarifying!