What would be a correct way of downgrading mlflow say from 2.9 to 1.30 without loosing data? The problem is the sql schema that is not compatible between versions.
I did manage to solve the problem by having 2 backends and a python script that sends experiments from one to the other, but I feel this is a hack/brute force at best
I would appreciate any answear, I really want to get this right.
What was tried:
plain frontend version change (without touching the db) (didn’t work)
python script that takes data from 2.9 to 1.30 (worked)
sql schema manipulation (didn’t work)