I found that the default simulation setting uses all CPU cores. How to properly set number_of_gpus_allocated_per_simulation , for example, for a single node machine with 4 GPUs?
We have not explored running with multiple GPUs on a single machine. I've mentioned in this issue that setting worker=sequential is the easiest way to get GPU access.
I found that the default simulation setting uses all CPU cores. How to properly set
number_of_gpus_allocated_per_simulation
, for example, for a single node machine with 4 GPUs?