I am trying to train an AlienGo control strategy based on depth images using raisimGymtorch. However, my program crashed when I set num_envs to 512. Through my attempts, I found that I can run up to around 240 environments simultaneously. However, I noticed that my GPU memory usage is not high but my CPU usage is very high.I think the problem might be related to the computational load of the simulator. My computer configuration is Ubuntu 20.04, CPU is i5, RAM is 32GB, and GPU memory is 24GB. Could you please explain the role of num_thread in the cfg file and suggest a reasonable value for num_envs?
Thank you for your work!
I am trying to train an AlienGo control strategy based on depth images using raisimGymtorch. However, my program crashed when I set num_envs to 512. Through my attempts, I found that I can run up to around 240 environments simultaneously. However, I noticed that my GPU memory usage is not high but my CPU usage is very high.I think the problem might be related to the computational load of the simulator. My computer configuration is Ubuntu 20.04, CPU is i5, RAM is 32GB, and GPU memory is 24GB. Could you please explain the role of num_thread in the cfg file and suggest a reasonable value for num_envs?