1

We are playing mxnet for a while. still couldnt get full cpu utilization as we have in tensorflow in default.

we have: python 3.6 mxnet 1.2.1 pos1 mxnet-mkl 1.2.1 pos1

installed via pip (in Pycharm). mac os x Sierra 10.12.6

we have only 2 core of the cpu used out of the 8 cores.

Thanks

2adnielsenx xx
  • 488
  • 1
  • 5
  • 22

1 Answers1

1

MXNET has a number of environment variables that determine the thread count. Try increasing MXNET_CPU_WORKER_NTHREADS. (The default value is 1.)

For more information:

Stephen C
  • 698,415
  • 94
  • 811
  • 1,216