trtexec, segmentation fault #4331
Description
Description
I have an ONNX model that can be correctly inferred using Python and ONNX Runtime.
However, when converting the ONNX model to a TensorRT model using trtexec, I encountered a segmentation fault.
The command I used to convert the model is as follows:
trtexec --onnx=onnx_model_path --device=$gpuid --minShapes=input_1:1,input_2:1x256,input_3:1x256,input_4:1x684x8x1,input_5:1x512x8x1,input_6:1x1x8x1 --optShapes=input_1:1,input_2:1x256,input_3:1x256,input_4:1x684x8x30,input_5:1x512x8x30,input_6:1x1x8x30 --maxShapes=input_1:1,input_2:1x256,input_3:1x256,input_4:1x684x8x75,input_5:1x512x8x75,input_6:1x1x8x75 --saveEngine=tensorrt_model_path
The ONNX model is as follows:
https://drive.google.com/file/d/13Xrm-13Hyjo2dyzmtrpgE-Miuc7JaDCO/view?usp=sharing
Environment
TensorRT Version: 10.7
NVIDIA GPU: NVIDIA RTX A6000
NVIDIA Driver Version: 550.76
CUDA Version: 12.4
CUDNN Version:
Operating System: CentOS Linux 7
Python Version (if applicable): 1.12