Output Training Options

RLlib
1

How severe does this issue affect your experience of using Ray?

  • Medium: It contributes to significant difficulty to complete my task, but I can work around it.

Hey,

Is there a way I can output all of the training/configuration options used by a model in Rllib?
I am talking about all the training options even those I haven’t specified myself and would like to know the default value. Looking for either a clean or even a hacky solution, poking around the code base to print the options out.

Any help is very much appreciated.
Thanks

2

You can use the return value of one train() iteration.

config = PPOConfig()
trainer = config.build()
iteration_result = trainer.train()

The iteration_result is a nested dictionary with all (hidden) config settings as well as training results.

If you use tune, you can use the return value of tuner.fit() accordingly.

    tuner = tune.Tuner(..)
    analysis = tuner.fit()
3

Hi Philip, thank you for your reply.

I am actually using tune.run but doing dict on the return of tune.run has similar effect.

Thank you for your help.