Does all_reduce_multigpu work with shared list (created by multiprocess.Module.list())?


I’m trying to use “all_reduce_multigpu” function with shared memory (list created by multiprocess module). However, it does not work with the shared list object (only working with a normal list).

I want to use the shared list to get return values from the child processes. Is there any alternative way?
Thank you in advance :slight_smile: