use DLA fail in half mode on tensorrt_version_5_0_6_3

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

Platform: Jetson xavier on tensorrt_version_5_0_6_3
Hello,
when i enable DLA to create my engine, engine created successfully,but when inferring it,I got below issue,is anybody got same issue?

NVMEDIA_DLA : 1097, ERROR: setInputTensorDesc failed 
NVMEDIA_DLA : 1193, ERROR: SetInputTensorDesc failed for tensor: 7. status: 0x0.
NVMEDIA_DLA : 1338, ERROR: BindArgs failed (Input). status: 0x7.
ERROR: dla/dlaUtils.cpp (536) - DLA Error in submit: 7 (Failure to submit program to DLA engine.)
ERROR: dla/dlaUtils.cpp (536) - DLA Error in submit: 7 (Failure to submit program to DLA engine.)
[INFO]::photoclass 
imageClassification: dla/nvmRegion.cpp:474: nvinfer1::rt::dla::NvmItem::~NvmItem(): Assertion `!mCudaMemory || !mNvmTensor' failed.
Aborted (core dumped)

my code related dla is as follows:

void enableDLA(IBuilder *b,int useDLACore)
{
	if (useDLACore >= 0)
	{
		const int nbDLACores = b->getNbDLACores();
		if (nbDLACores<=0)
		{
			LOG_ERROR("DLA is not support on this platform");
			return;
		}
		else if(nbDLACores<=useDLACore)
		{
			LOG_ERROR("requested DLA core num[%d] is greater than DLA core num provided[%d]",
				useDLACore,nbDLACores);
			return;
		}
		b->allowGPUFallback(true);
                b->setFp16Mode(true);
                b->setDefaultDeviceType(DeviceType::kDLA);
                b->setDLACore(useDLACore);
	}
	return;
}
//enable dla
if (modeData != kFP32 )
{
	if (gUseDLACore >= 0)
	
		if ((int)maxBatchSize > builder->getMaxDLABatchSize())
		{
			LOG_ERROR("Requested maxBatchSize %d is greater than the max DLA batchSize %d,reducing it accordingly!",maxBatchSize,builder->getMaxDLABatchSize());
			assert(0); //not support required batchsize when use dla

		}
		enableDLA(builder,gUseDLACore);
	}
}
2

Hello，
According to the way of calling DLA in the example sampleMNIST,I have modified my code,but the following problems still occur. May I ask what do these logs mean and what causes these problems?

NVMEDIA_DLA : 1097, ERROR: setInputTensorDesc failed 
NVMEDIA_DLA : 1193, ERROR: SetInputTensorDesc failed for tensor: 7. status: 0x0.
NVMEDIA_DLA : 1338, ERROR: BindArgs failed (Input). status: 0x7.
ERROR: dla/dlaUtils.cpp (536) - DLA Error in submit: 7 (Failure to submit program to DLA engine.)
ERROR: dla/dlaUtils.cpp (536) - DLA Error in submit: 7 (Failure to submit program to DLA engine.)
3

Hi,
I’m seeing something similar. Were you able to find the reason for this?

thanks
Eyal