[Question] DDPG Parameters

[mdev11]

Hi I'm new to RL and currently working on my project using DDPG from stable baselines. And I have some questions about the parameters:

1. What is the difference between timesteps and nb_train_steps? If I didn't change the nb_train_steps, would it affect my model?
2. I set the verbose parameter to verbose=1, but the training information only appear every 10k steps, can I change it? like every 1k steps maybe?

Thank you.

[araffin]

What is the difference between timesteps and nb_train_steps?

See #776 (comment)

but the training information only appear every 10k steps, can I change it? like every 1k steps maybe?

you can change log_interval in the learn() method (cf doc)

it also depends on nb_training_steps (see https://github.com/hill-a/stable-baselines/blob/master/stable_baselines/ddpg/ddpg.py#L866)

[mdev11]

Okay thank you so much :)
More question please.. its kind out of topic but, my actor keep getting negative loss, is it normal?

[araffin]

my actor keep getting negative loss

Please take a look at the resources in our documentation to understand more about the meaning of the losses (notably take a look at spinning up).
Usually actor loss value does not have much sense.

Closing this issue then.