Could we run deepstream with Cuda 11?

may main question is, Could i install and run deepstream 5 with cuda 11??
because I’ve a model that failed to convert into engine with tensorRT7, but success with tensorRT7.2
and while deploying that model with deepstream it gives me an error because the model made with tensorRT7.2 and deepstream docker use tensorRT7.0.0
if you have any work around this situation I’d be grateful

• Hardware Platform GPU
• DeepStream Version
• TensorRT Version 7.2.0

Deepstream 5.0 is based on cuda 10.2, you can’t use cuda 11 for 5.0, TensorRT 7.2 compatible cuda version is 11.0, so you may wait for a newer deepstream release which support cuda 11.0


So can I convert my model to engine with TensorRT 7.2 then get it working on deepstream5 with tensorRT7.0 ?

Your engine built by TensorrRT version must be same as the TensorRT version while you run it. it will do some check. if not equal, it will report error.

I see that DeepStream has been released for Cuda 11 and A100. Does it support other GPUs too?

What’s A100 ?

I think that’s a GPU for data center