tensorrtx icon indicating copy to clipboard operation
tensorrtx copied to clipboard

Segmentation fault (core dumped)

Open VanCholen opened this issue 1 year ago • 2 comments

Env

  • GPU, RTX3090

  • OS, Ubuntu 20.04

  • Cuda, 11.3

  • TensorRT version, TensorRT-8.0.1.6

  • which branch/tag/commit are you using? yolov5-5.0

  • which model? yolov5, retinaface? yolov5s download from https://github.com/ultralytics/yolov5/releases/download/v5.0/yolov5s.pt

Your problem

I can successfully generate yolov5 and yolov5s.wts from source file, but when I try "sudo ./yolov5 -s yolov5s.wts yolov5s.engine s" , I get an error such as follows: Loading weights: yolov5s.wts Building engine, please wait for a while... [09/15/2022-23:33:27] [W] [TRT] Detected invalid timing cache, setup a local cache instead Segmentation fault (core dumped)

VanCholen avatar Sep 15 '22 15:09 VanCholen

Seems like a bug in TensorRT. Can you try tensorrt7.x or tensorrt8.2 or 8.4?

wang-xinyu avatar Sep 16 '22 05:09 wang-xinyu

Thanks for your apply, I will try it again.

VanCholen avatar Sep 16 '22 15:09 VanCholen

Thanks for your apply, I will try it again.

hi, have you solve that, I have the same error.

ClydeU420 avatar Oct 24 '22 08:10 ClydeU420

Thanks for your apply, I will try it again.

hi, have you solve that, I have the same error.

The TensorRT API have updated in TensorRT 8 , you'd better to keep the same version of TensorRT with author.

VanCholen avatar Oct 25 '22 02:10 VanCholen