search

NM512 / dreamerv3-torch

Implementation of Dreamer v3 in pytorch.
MIT License
422 stars 96 forks source link

Multigpus #51

Closed realjoenguyen closed 1 month ago

realjoenguyen commented 7 months ago

Hi, How to extend the codebase to multi-gpus ? Seems like non-trivial to do that? Thank you,

NM512 commented 7 months ago

Hello,

One option worth exploring is leveraging the implementation of RLlib, as it inherently supports multi-GPU training. RLlib provides a robust framework for reinforcement learning, and its multi-GPU capabilities can serve as a valuable reference point for your codebase extension.

Moreover, if you can accept TF implementation, you might find DreamerV3 within the RLlib library.