How to calculate EPIC Distance??

[JongKook-Heo]

Hi, thanks for your nice work!!

I'm wondering if you have any code implementation to calculate epic distance between ground truth reward function and learned reward function??

Concretely, I have no idea how to get true reward from uniformly sampled state, action pair.

How can it possible to calculate reward from certain state-action pair in metaworld environment??

It gonna be possible if the state was sampled by env.reset() and the reward was calculated by env.step(). but as far as I know, it's not available calculate reward from user-defined state.

Best regards,
Jongkook, Heo

[CAI23sbP]

@JongKook-Heo you can see this repo.