search

Stable-Baselines-Team / stable-baselines3-contrib

Contrib package for Stable-Baselines3 - Experimental reinforcement learning (RL) code
https://sb3-contrib.readthedocs.io
MIT License
465 stars 173 forks source link

[Feature Request] STAC algorithm #223

Open EloyAnguiano opened 9 months ago

EloyAnguiano commented 9 months ago

🚀 Feature

Build the STAC algorithm as a callable algorithm: https://arxiv.org/pdf/2002.12928.pdf

Motivation

Hyperparametrization is one of the most time/cost expensive thing when training RL agents. May be this implementation saves some time/cost to some people and it could be the first AC algorithms that deals with meta-gradients to make improvements from here.

Pitch

I would like some to guide me of where to start or to give me some key insights of the posibilities of coding this.

Alternatives

The alternatives are that someone codes it by him/herself.

Additional context

No response

Checklist

araffin commented 8 months ago

Hello, are you willing to implement and benchmark the algorithm?

EloyAnguiano commented 8 months ago

Yes, I would like to try to do so. Is there any oficial benchmark to do so or some coding guides?

EloyAnguiano commented 8 months ago

The algorithm is an Off-policy one. Is there any way or example to begin with this kind of algorithms?

araffin commented 8 months ago

The algorithm is an Off-policy one. Is there any way or example to begin with this kind of algorithms?

https://github.com/Stable-Baselines-Team/stable-baselines3-contrib/pull/4 and please read https://github.com/Stable-Baselines-Team/stable-baselines3-contrib/blob/master/CONTRIBUTING.md