RescaleAction: how to get the unscaled value after env.step?

[glitchyordis]

My environment's action space has a dim of (4,), each with a different range of values. I'm told PPO are usually train with action ranging between -1 and 1. If I wrap the env with RescaleAction, how do I get the actual value instead of the one scaled between -1 and 1 after performing env.step()?

[pseudo-rnd-thoughts]

You shouldn't need the rescaled actions.
Just imagine that the environment doesn't have (4,) with a different range of values, but a (4,) between -1 and 1.
So you should just use your actions output by the agent not the rescaled ones

[glitchyordis]

I forgot to mention that my physical setup (i.e, my agent but in physical world) operates not only within -1 and 1 but across the entire range specified in my environment's action space. Am I misinterpreting the usage of RescaleAction?

[pseudo-rnd-thoughts]

The purpose of the RescaleAction wrapper is for you to rescale actions from one space to another through linearly projecting the values