An error while using TensorRT with NVDLA

I’m using Nsight Compute CLI to profile some TensorRT engines. I have no problems while running the engines without Nsight compute. When I assign all layers to the GPU, I don’t have any problems either. However, TensorRT throws an error from Cuda if I assign the device as DLA rather than GPU. The consecutive errors are the error from each frame/image iterations that I pass as a paramater(3 in this specific example) I tried to pass some parameters to get rid of this error but any attempt did not work.

[E] [TRT] …/rtExt/dla/eglUtils.cpp (275) - Cuda Error in CudaConsumer: -1 (unrecognized error code)

[E] [TRT] FAILED_EXECUTION: std::exception
[E] [TRT] FAILED_EXECUTION: std::exception
[E] [TRT] FAILED_EXECUTION: std::exception

Environmvent:
-Xavier AGX with ubuntu 18.04
-Tensorrt: Version 7.1.3
-Nsight Compute: Version 2019.5.3 (Build 29011660)