Im using Vim3 Pro 1.5
Had to move from tengine to KSNN, and here I have few questions:
1)Can I convert on different device?
2)I followed your instructions in KSNN github, but here is one thing - how to use convert?
where is ./convert ?
what inputs should contain?
how to count mean-value?
thx
Please check the aml_npu_sdk repo for documentation and instructions on file conversion:
Okay
And the last question, while using tengine I could work with NPU, there were neural networks on timvx, and how to do it on ksnn?
can you elaborate, what models are you referring to ?
I was using Yolov5s, but now I planned move to RetinaFace
On tengine the process took 44 ms. Will it be the same on KSNN?
Most likely it should have same inference time, if there is no difference in the model.
