Closed jaried closed 2 years ago
It's more or less because of python subprocess and GIL. If your environment is written in Python, unfortunately, for a single-machine case there's no solution IMHO.
Is it possible to avoid this problem by writing in C++ and using envpool?
I think so.
Thanks for your answer.
When I use tianshou multi-environment concurrently, due to the high cpu occupancy rate of the main program, the multi-core of the CPU is not fully utilized. Is there any solution?