How can I speed up the loading of Industrial Heat Sink?


I am running Industrial Heat Sink’s using mpirun --allow-run-as-root -np 4 with 4 GPUs, but it takes a very long time from
GraphManager: {'_func_arch': False, '_debug': False, '_func_arch_allow_partial_hessian': True}
to the start of the training. I believe the conversion process of the STL files is taking a significant amount of time. During this period, I checked the CPU and GPU usage and noticed that the 4 CPU cores were at 100% while the GPUs were at 0%. It seems like one CPU core is handling the processing for one GPU.

Therefore, I would like to increase the number of active CPU cores to speed up the loading process. How can I achieve this?

Thank you very much for your help