Description
I am working on a test using the sampleOnnxMNIST
from the TensorRT official samples. To optimize the performance, I have enabled CUDA Graph to execute the inference and employed CUPTI to track the kernel activities from the model.
In the test, I have observed that the sequence initTrace
followed by infer
works correctly. However, when I try the sequence infer
followed by initTrace
and then another infer
, I encounter an error with code 700 which means an illegal memory access was encountered
.
Environment
TensorRT Version: v8502
GPU Type: NVIDIA GeForce RTX 3070
Nvidia Driver Version: 510.39.01
CUDA Version: 11.6
CUDNN Version: 8.9.6
Operating System + Version: Ubuntu 20.04.6 LTS
Python Version (if applicable): 3.8.10
TensorFlow Version (if applicable):
PyTorch Version (if applicable):
Baremetal or Container (if container which image + tag):
Relevant Files
trt_samples.tar.gz (15.8 MB)
Steps To Reproduce
- untar the sample code
tar xvzf trt_sample.tar.gz
- cd to the directory
trt_samples/sampleOnnxMNIST
- build the sample
make -j8
- run the sample
../../bin/sample_onnx_mnist
- sample will abort and occur the error
Cuda failure: an illegal memory access was encountered