What should i do to compile GPU target code?
Would be great to learn how to optimize for gpu at 6.1. Part 1 — Machine Learing Compilation 0.0.1 documentation.
1 Like
Thank you very much! I it solve my problem
does relax have auto-schedule mechanism now? when many kind of ops are imported from relax front(such as torch FX),should I bind variable to thread manully with cuda target?
Meta-schedule is fully supported by relax for static shapes. Dynamic shape tuning is not ready now