Comparison with hidet

[austinmw]

Hi, do you happen to know how the optimization performance for this TVM based solution might compare to https://github.com/hidet-org/hidet for NVIDIA edge devices (like NVIDIA Jetson Xavier)?

I'm curious to understand which has the capability to make LLMs smaller/faster for these devices?

I assume they are not compatible, but please let me know if I can use both.