I have some problem to use TensorRT on jetson nano to convert file.pt to file wts, please help me

Autonomous Machines Jetson & Embedded Systems Jetson Nano
,
1

jetson@duongphiv5-desktop:~/JetsonYOLOv7-TensorRT/yolov7/build$ sudo ./yolov7 -s yolov7-tiny.engine t

Using engine, please wait for a while…

weights: yolov7-tiny.wts

…

[02/24/2023-13:22:21] [E] [TRT] (Unnamed Layer* 187) [Convolution]: kernel weights has count 648 but 5760 was expected

[02/24/2023-13:22:21] [E] [TRT] (Unnamed Layer* 187) [Convolution]: count of 32640 weights in kernel, but kernel dimensions (1,1) with 128 input channels, 45 output channels and 1 groups were specified.

weights count is 128 * 45 * 1 * 1 = 5760

…

[02/24/2023-13:22:21] [E] [TRT] Network must have at least one output

[02/24/2023-13:22:21] [E] [TRT] Network validation failed.

/home/duongphiv5/JetsonYOLOv7-TensorRT/yolov7/main.cpp:38: void serialize_engine(): Assertion `engine’ failed.

Aborted

This my all versions in Jetson Nano:

CUDA: 10.2.300

cuDNN: 8.2.1.32

TensorRT: 8.2.1.8

VPI: 1.2.3

Vulkan: 1.2.70

OpenCV: 4.5.1 with CUDA: YES

Module: NVIDIA Jetson Nano (4 GB ram)

SoC: tegra210

CUDA Arch BIN: 5.3

Codename: Porg

L4T: 32.7.4

Jetpack: 4.6.4

3

Hi,

Based on the log, there are some incorrect configure in the yolov7 application.
Have you successfully run the app on other platforms?

It’s also recommended to check with the application owner to see if anything is missing.

Thanks.

4

i used to build it successfully on ras pi 4, but for jetson nano don’t work for me, i think there are some trouble about pip and pytorch version

5

when i use : ```
yolov5_det -s yolov5s.wts yolov5s.engine s ‘’ , it always have problems both yolov7 and yolov5, although i think all version pytorch, torchvision are compatible

6

this is my problems

image
image1920×1440 363 KB

7

Hi,

Raspberry pi 4 doesn’t have GPU so it won’t use TensorRT.
Have you tried it with other platforms that support TensorRT?

Thanks.

8

it is jetson nano

9

Hi,

Based on the eror when building the TensorRT engine, the app isn’t configured correctly.
Please check with the application owner to see if the sample support the TensorRT version you used.

Thanks.

11

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.