Jetson Nano - display crash on loading NN model

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

I’m fighting with an issue for a week - when I’m trying to load pytorch/tensorflow model the display suddenly turns off. I tried to enable terminal mode, to get more RAM, connected over ssh from my laptop and turned on memory monitoring. I found that a crash happens even when the occupied memory level is 2.5.
When the model was really big, I usually was getting Memory Error.

I have used different versions of:
pytorch - 1.0.0 … 1.3.0,
tensorflow - 1.12 … 2.0 (also tried to convert models to tensorRT)
CUDA - 10.0
I tried different models resnet50, resnet101

Don’t have an idea how to fix it, please help me

2

Hi,

A common issue is the power starvation. You can check this topic for more information:
https://devtalk.nvidia.com/default/topic/1048640/jetson-nano/power-supply-considerations-for-jetson-nano-developer-kit/

To figure out if you are facing the power issue, please set the nvpmodel into 5W and try the pytorch/tensorflow app again.

sudo nvpmodel -m 1

Thanks.

3

Thank you very much!
It helped