@NathanMerrill Regarding portability of ML models I just stumbled over this project: khronos.org/nnef#nnef_provisional_to_one_changes

@Laikoni oooh, interesting, I'll check it out

ok, so this is interesting: Their proposal is if you standardize the data format, then you are more able to use somebody else's neural net

but that means you have to standardize before you train it.

otherwise, you still have the same problem