Hi everyone,
I have started experimenting with TVM for some of my projects & I must say it is exciting but also a bit overwhelming at times. I have been trying to understand the best ways to optimize performance when deploying models, on edge devices where resources are limited. I am testing smaller models but I want to scale it up gradually.
I keep facing is figuring out the right configurations & schedules that improve performance without breaking things. The docs are helpful but sometimes it feels such as I need a bit of real-world insight from folks who have already dealt with these issues.
Also, I want to know if there are suggestion learning paths or resources to get a deeper understanding of performance tuning. I came across a Golang Course Online while searching for related programming resources & it got me thinking if branching into different languages helps when working with TVM.
I want to hear your thoughts, tips or even mistakes you have learned from while tuning models with TVM. Also i have check this Optimizing a loop body expressed with tvm.extern still need advice.
Thank you.