PyTorch implementation of Advantage Actor Critic (A2C), Proximal Policy Optimization (PPO), Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation (ACKTR) and Generative Adversarial Imitation Learning (GAIL).
MIT License
3.53k
stars
832
forks
source link
New parallel PyTorchRL library based on this one #267
I think that this is relevant for people who would like to try multiple algorithms or scale it up. We would be happy to work with you somehow if you still keen in this. In any case, I hope that this message is useful for you and everyone else, but feel free to close it, if it's not.
@ikostrikov We have developped a new library starting from this one. We have submitted now what we think that is a practically final version.
It is still based on pytorch but has two differences:
Few years ago we used your PPO implementation to get second position in this challenge from Unity: https://blogs.unity3d.com/2019/08/07/announcing-the-obstacle-tower-challenge-winners-and-open-source-release/
But we felt we would have won if the library was parallel, so here it is PyTorchRL: https://github.com/PyTorchRL/pytorchrl
and the paper: https://arxiv.org/abs/2007.02622
The docs: https://pytorchrl.readthedocs.io/en/latest/
I think that this is relevant for people who would like to try multiple algorithms or scale it up. We would be happy to work with you somehow if you still keen in this. In any case, I hope that this message is useful for you and everyone else, but feel free to close it, if it's not.
Best, Gianni