Process Killed yoloR_w6

[alejoGT1202]

Hello,

I follow the steps described here and was able to generate the .wts and .cfg files. However, when I execute the deepstream-app -c deepstream_app_config.txt command the process gets killed.

The last thing my console prints is:

Total number of YOLO layers: 638

Building YOLO network complete
Building the TensorRT Engine

Killed

I'm running the code in a Jetson Nano Dev Kit, using the yoloR_w6 weights, is it possible to run this weights on Jetson Nano or should I use the CSP weights?

Thanks for the help.

[pabsan-0]

Looks like you're out of RAM. Maybe you can relieve some by killing stuff like text editor, web browser... Use jtop on Jetson for a detailed monitor. If you can't you'll probably need to switch to a lighter model.

I haven't tried TensorRT conversion, but recently we could not infer Yolov5L or Yolov4-P6 (with torch) on the Nano for this very reason, it is quite memory-limited.

[marcoslucianops]

You sould use tiny models on Jetson Nano board. It doesn't have sufficient resources to run heavy models.