Set priority for triton

Please provide complete information as applicable to your setup.

• Hardware Platform (Jetson / GPU) GPU
• DeepStream Version any
• NVIDIA GPU Driver Version (valid for GPU only) 460
• Issue Type( questions, new requirements, bugs) question/new requirement
• Requirement details( This is for new requirement. Including the module name-for which plugin or for which sample application, the function description) nvinferserver / triton

Triton has the option to set priority levels for each request to allow inference in the frames that are more important through the option priority-levels in the configuration (docs). Is it possible to use this feature in deepstream? I have looked the documentation but I don’t find any option that resembles this. If it is not implemented yet, will it be implemented at some point?

Sorry for the late response, we will do the update soon.

Thanks

Hi @1993mwc ,
Sorry for delay!

We don’t support this now. Could you share the use case about this feature, we could evaluate if we need to add it in DS.

Thanks!

Hi @kayccc,

In our case we have several sources and we focus on some of them based on activity, but for security reasons we need to keep non-priority streams also in the pipeline, but due to low activity we do not need to do inference as often as in the high-priority streams. So having the priority-level option would be a perfect solution for all similar cases.

Thanks! @1993mwc

We add it into our roadmap, and will evaluate which release will have it.

1 Like