Different results from torch.get_num_threads when using docker in x86 and arrch64 cpu

i got this value from x86 cpu is 1.

I’m trying to control the number of threads utilized by OpenMP or Intel Math Kernel Library (MKL) within a Docker container so that it uses only the threads available inside the container instead of using those from the underlying host machine.
now i need to set MKL_NUM_THREADS or OMP_NUM_THREADS;

is there anyone know why this happed?

is to blame arrch64 cpu or pytorch.