Hi, I am a mechanical engineer and using rllib for RL control of rocket engines for my phd research. I am using rllib locally on my windows machine just with ray.init() and go. We further have a workstation with 2 CPUs and on this machine ray is only using 1 CPU and the second CPU is idling. Is there an easy way to use both CPUs?
If you use python try at first:
ray.init(num_cpus=2)
Maybe this page will be useful:
https://docs.ray.io/en/master/rllib-training.html
Peter
Hello, thank you for your reply and your time.
I think that by “CPUs” we mean something different.
We have a workstation with 2 separate Intel xxxx CPUs with 18 cores each.
Now if I use ray.init as before, ray.available_resources tells me that I have 36 cores. But if I spawn e.g. a RL training with Tune, I can only use the 18 cores of the first processor. The second one is simply chilling and noot motivated to work:D
This sounds definitely weird. Is it possible to create a script (self-running script so that I can try) that can reproduce the issue?
It will be great if you can create an issue to our Github issue Issues · ray-project/ray · GitHub
Thank you for your time. I finally created the issue on Github