Platform Dependency of Calibration Cache

I was wondering if the calibration caches created during INT8 calibration are platform specific? Can I build one cache and use it for the same network at another platform?

Hi @orcdnz,

You can build cache on host and can be re-used by the builder on the other target as well. Please refer Notes in the 12.2 section of developer guide.
https://docs.nvidia.com/deeplearning/tensorrt/developer-guide/index.html#deploy_embed

Thank you.