Correct way to convert a model

Are there any fixed set of steps or probable modifications to the script that one needs to make to correctly convert an existing model into it’s .nb and .so components?

I’ve tried converting resnet models (which is a seemingly popular architecture) from tensorflow slim repository but with any luck (wrong accuracy values). Whereas the same inception network from tf_slim gets converted without any hassle. Please do suggest a solution

Any hints on the steps followed to get the models in the zoo would also be great

https://docs.khadas.com/linux/vim3/KSNNConvert.html

And you can follow the README in ksnn repo. Every model I have push the convert parameter in README

All the models in my warehouse have been converted into demos in the ksnn repo. The converted parameters are all in ksnn/examples/${platform}/${model}/README.md

Upon following the exact steps as mentioned in the docs, I get highly inconsistent accuracy results

Linked issue