Running multiples deepstream instances on AGX with tiny yolov3 as pgie , resnet18 as sgie

Please provide complete information as applicable to your setup.

**• AGX ( 32G )
**• DS5.1
**• Jetpack 4.5.1
**• TensorRT 7.1.3
**• Nvidia L4T 32.5.1
**• running multiples deepstream app and consume almost 25G memory with tiny yolov3 as primary detector , resnet18 as classifier , GPU consumption : 25% , Memory : 25G , average CPU : 60% when running 17 deepstream-app .

would like to know why the GPU only consume around 25% but memory is almost hitting the max ( out of 32G )

would like to know any recommended configuration on running multiples instances of deepstream instead of batch mode .

Seperate process take seperate memory allocated, you run multi deepstream-app, and memory hit around 25G, it’s expected. as for GPU loading, it depends on how many inference submit work loaded and other components loading which run on GPU. like tracker, nvvideoconvert, tiler, etc.
You can try run multi pipelines within one process.
deepstream-app -c config -c config …

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.