Yolov6 evaluation failing
Hi everyone,
My team and I are working on running yolov6 on the 720 chip but we ran into this issue:
We’re not quite sure what’s causing it. We used the same script to export yolov5 to onnx from torch hub and export the optimised version without issues.
Can you help us out?
Thanks!
The discussion has been closed due to inactivity. To continue with the topic, please feel free to post a new discussion.
Comments
@Aliaa Diab
Hi Aliaa Diab,
After looking at your model, there is a bug in Kneron Toolchain, but the bug has been reported and will be fixed in the next version of Kneron Toolchain.
With this model, theoretically, the KL720 needs to be able to support the "sub" operator, but it has been verified that there is a problem here.
If you need to move this project forward quickly, you can try to remove the operators after the model Transpose and do the removed operators on the host side.
As an additional reminder, I recommend that you check that the operators in the model are compatible with the Kneron chip architecture before compiling the model into a Kneron nef model.
Reference link.
https://doc.kneron.com/docs/#toolchain/appendix/operators/