Generalising Map: am I able to contribute?

[Joel]

Hi all,

I raised this ticket https://github.com/tensorflow/tensorflow/issues/55246 for generalising Map to work with functions with arbitrary shapes.

I'm now wondering if I'd be able to attempt the change myself. I have no experience of the XLA internals, nor compilers, but I have used the XLA C++ API quite a bit over the last year. Is someone able to give me a very rough idea of how much work it would be, and perhaps how complex the task would be, so that I can figure out if it's worth me learning the necessary details and giving it a try.

Thanks,

Joel