Must specify a non-zero or non-empty correlation ID - Triton with sequence batching, Tensorrt

We are investigating the issue. Will be back when there is progress.

We’ve got some clue, can you share your models and configurations for us to reproduce the problem?

I prepared a zip file with all you need to reproduce this error with a simple identity Python model.

I think that the issue is that you can’t configure Deepstream to send correlation id’s

See the attached zip:
demo_correlation_id.zip (6.7 KB)

it includes:

  • deepstream app config file
  • model config
  • triton model repo

Tested inside container version: deepstream:6.1.1:triton

gst-nvinferserver does not support such feature now. We will add this feature to our roadmap.