hi
I trying to make Docker file for perform conversation of yolov8n model to hef based on this totorial
So I prepared yolov8n.onnx model file with v11 command set.
Dockef file use Ubuntu 20.04
Then I prepared packages
And zoo data
(Btw, as you may see to make things workable I had to do a couple tricks - override cfg folder and rename coco data folder. It would be pretty nice if model zoo will not have these bugs)
And then parse, optimize and compile model
It parsed and optimized well.
But on comilation it failed.
How can I fix this?
Thanks
Hi @vitaliy.bondarchuk,
Thanks for the detailed steps.
It seems that your overriding of the cfg files is causing the issue. One of the important steps for yolov8 is setting the start and end node names, which is done automatically if you pass the yolov8 arch name. This is necessary because the post-process is not natively supported, instead, it’s added back to the model during the optimization with the nms_postprocess command.
hi
thanks for reply.
basically I successfully compiled the yolo8n when run steps of this tutorial
directly on WSL, not in a Docker conteiner.
As far as I remember this time i just copied “generic” and "hailo8l@ subfolders of cfg/all to the lib folder.
Also I increased RAM size of WSL VM to 16GB.
So I going to try docker again with these succeedede procedure from WSL
