WARNING:root:Use native layout for depthwise convolution on NNVM

I am trying to optimize my mobilenet based model. Using the default auto tune i am able to export but i am getting this warning. Is there further optimization possible ? Can you please suggest what changes i need to make in the script to this possible.