This message was deleted.

Answer:

--production

uses a different runner- which does not model the same behavior as the local runner (no production flag) when dealing with

FileLike(_wrapped=<tempfile.SpooledTemporaryFile object at 0x7fb025457a30>, _name='file.h5ad'

. Make sure to convert the file into a pickleable python object before passing it into the runnable method, as it will try to pickle the FileLike object throw the error

TypeError: cannot pickle '_io.BufferedRandom' object