Parameter check failed while using dla

Hey all,

I took the code located at this repository:

https://github.com/NVIDIA-AI-IOT/tf_to_trt_image_classification

and modified the uff_to_plan file by adding this code

builder->setDefaultDeviceType(DeviceType::kDLA);
builder-> setDLACore(1);
builder->allowGPUFallback(true);

right below the builder to enable the dla. When I run either the classify_image script or the test_trt.py script I get them to successfully run but I get this series of errors at the end:

Parameter check failed at: dla/egUtils.cpp::~CudaProducer::162, condition cudaEGLStreamProducerDisconnect(&mConnection) failure.
Parameter check failed at: dla/egUtils.cpp::operator()::137, condition: eglDestroyStreamKHR(display, stream) failure.
Parameter check failed at: dla/egUtils.cpp::~CudaConsumer::245, condition cudaEGLStreamConsumerDisconnect(&mConnection) failure.
Parameter check failed at: dla/egUtils.cpp::operator()::137, condition: eglDestroyStreamKHR(display, stream) failure.
Parameter check failed at: dla/egUtils.cpp::~CudaProducer::162, condition cudaEGLStreamProducerDisconnect(&mConnection) failure.
Parameter check failed at: dla/egUtils.cpp::operator()::137, condition: eglDestroyStreamKHR(display, stream) failure.
Parameter check failed at: dla/egUtils.cpp::~CudaConsumer::245, condition cudaEGLStreamConsumerDisconnect(&mConnection) failure.
Parameter check failed at: dla/egUtils.cpp::operator()::137, condition: eglDestroyStreamKHR(display, stream) failure.

any thought to what might be causing it.

Hi,

May I know which JetPack version do you use?
This is a known issue and already fixed in the last summer.

If you are not using JetPack4.2, would you mind to give it a try?
Thanks.

Hey Aasta,

It looks like I’m using JetPack4.1, is there a way to use install Jetpack4.2 without deleting existing data on the AGX Xavier?

Thanks.

Hi,

JetPack4.2 can be installed with sdkmanager here:
https://developer.nvidia.com/embedded/jetpack

Thanks.

Hey Aasta,

This error resurfaced in conjuction with the one on the other thread:
https://devtalk.nvidia.com/default/topic/1052449/jetson-agx-xavier/cuda-memory-error-when-enabling-the-dla/post/5342754/#5342754, I am running JetPAck4.2 and TensorRT 5.0.6.3