Please provide the following info (check/uncheck the boxes after creating this topic):
Software Version
Linux nvidia 4.9.253-tegra aarch64 GNU/Linux
Target Operating System
Linux
Hardware Platform
NVIDIA DRIVE™ AGX Xavier DevKit (E3550)
My environment is tensorRT8.0.1.
I have four layer to run concat operation. Dims{32, 20, 20},Dims{32, 20, 20}, Dims{64, 20, 20}, Dims{128, 20, 20}. I want to run it on dla. so I write that example
inputTensors[0] = Dims{32, 20, 20};
inputTensors[1] = Dims{32, 20, 20};
inputTensors[2] = Dims{64, 20, 20};
inputTensors[3] = Dims{128, 20, 20};
nvinfer1::ITensor* inputTensorsSub[] = {inputTensors[0], inputTensors[1]};
nvinfer1::IConcatenationLayer* input_32 = network->addConcatenation(inputTensorsSub, 2);
nvinfer1::ITensor* inputTensorsSub1[] = {input_32->getOutput(0),inputTensors[2]};
nvinfer1::IConcatenationLayer* input_64 = network->addConcatenation(inputTensorsSub1, 2);
nvinfer1::ITensor* inputTensorsSub2[] = {input_64->getOutput(0),inputTensors[3]};
nvinfer1::IConcatenationLayer* input_128 = network->addConcatenation(inputTensorsSub2, 2);
return input_128->getOutput(0);
It get some wrong problem.Module_id 33 Severity 2 : NVMEDIA_DLA 684 Module_id 33 Severity 2 : Failed to bind input tensor. err : 0x00000b Module_id 33 Severity 2 : NVMEDIA_DLA 2866 Module_id 33 Severity 2 : Failed to bind input tensor args. status: 0x000007
And when I write plugin to implementation it and run on gpu, it’ OK.
Can you help me, Thank you.