Open
Description
Hi, Is it possible for using mlir to make LLVM IR in machine training like GPU support?
I can not find any code in tensroflow to use mlir turn back to tensorflow executor. Therefore, mlir is only useful for inference? I wonder if tensorflow could change make some ops into IR process and then merge IR result back to tensorflow process?
Metadata
Metadata
Assignees
Labels
No labels
Activity