What are the num_parallels for?

[ameliacode]

Hi again. I am messing around with the code for months.
I did read the paper over and over again, seems like controller are for synthesizing motion.

So.. I have made my own version of environment to train the characters to compete each other.
and I have notice there is num_parallel variable which visualize numerous characters.

However, as it is on the code, it seems that action space dimension is only for single character unlike other things as root_history which includes dimension of num_parallel.

Is num_parallel variable exists just for training? I am confused..

[belinghy]

I'm glad to hear that you are extending our project, @ameliacode!

By "action space dimension", are you referring to these two lines? If yes, then this is just the standard way of defining the environment; and yes, it is only defined for the one character.

The environment returns a state array, its size is (num_parallel, observation_size). During training and inference, the array is passed through a network which convert state into actions, e.g., (num_parallel, action_size).

[ameliacode]

soo....Is it possible to utilize or interpret num parallel as number of characters ?? (ex using root xz or root facing) or is the num parallel exists just for training?

I just want simplify my environment to two identical agents competiting in my case..

[belinghy]

Yes, they are the same, num parallel is number of characters. With some changes to the environment, it is possible to have two agent competing together.

[ameliacode]

thanks for reply! it helped a lot