Exploring ML Compiler Optimizations with microTVM