[torch_tvm] How do we tune our models?

Questions
#1

Hello,

I’m reading this with great interest, but how does the “autotvm_tuning.log” was generated? The tuning process needs to load the pytorch model in TVM right? or is it possible to do it from Pytorch?

Thanks!

#2

see tutorials here how to tune a relay program, you need to convert pytorch model to relay first