Hi, I am preparing to deploy a multi-agent environment with rllib on my ubuntu20.04 for training my own defined env. My graphics card is 3090, cuda version is 11.3, I have tried to install it under python3.9, but according to the instructions of the document, there is incompatibility between torch, numpy, and cuda versions. Can you recommend a usable version combination for me? I don’t want to change my cuda version, thank you!
Hi Fhy0720 welcome to the Ray community! Can you send over the document you are using to set this up and what version of torch, numpy, ray, you tried to install that showed the errors? Thanks!!