"Error Code 10: Internal Error" when running trtexec for int8

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

Description

I’m attempting to create a int8 yolov8 segmentation engine using ‘trtexec’ but I encounter the following error:

Error Code 10: Internal Error (Could not find any implementation for node /model.22/proto/cv3/conv/Conv + PWN(PWN(/model.22/proto/cv3/act/Sigmoid), /model.22/proto/cv3/act/Mul).)
)

I attached the fully log from the trtexec with --verbose for more explicit details
output.txt (2.6 MB)

I also have tried a hard implementation for this not found nodes in the calibration file(attached bellow) but it didnt work.
yolov8m-seg.engine.Orin.int8.4.4.calibration.txt (17.9 KB)

Environment

TensorRT Version: 8.6.2
GPU Type: Jetson Orin Nano development kit
Nvidia Driver Version: JetPack 6.0
CUDA Version: 12.2
CUDNN Version: 12.2
Operating System + Version: Ubuntu 22.04
Python Version (if applicable): 3.10.12
PyTorch Version (if applicable): 2.1.0

Steps To Reproduce

Command that I used:

/usr/src/tensorrt/bin/trtexec --onnx=yolov8m-seg.onnx --saveEngine=yolov8m-int8-seg-b4-onnx.engine --optShapes=‘images’:4x3x480x480 --int8 --calib=yolov8m-seg.engine.Orin.int8.4.4.calibration --verbose

2

Hi @gabrielmstefanello058 ,
This is caused by a CUDA bug and fix in the latest SW

3

Hello, thanks for the reply, so what should I update to solve this problem? the entire JetPack? Cuda? Tensorrt?

4

I have the same problem，could you tell me how to solve it?

5

Need to update the version of CUDA or TensorRT?