Using MLIR on Conv2D (and other TF filters)

Hello, where should I look for tutorials on how to take classical TensorFlow transforms (e.g. Conv2D) and transform them into efficient single/multi-core code using MLIR? I know of the very good tutorial of Uday on matmul (https://github.com/bondhugula/llvm-project/blob/hop/mlir/docs/HighPerfCodeGen.md). Is there something equivalent for other transforms, most notably Conv2D? If it’s just code (in some MLIR-based TF back-end), its also good, but I don’t know where to look for it.