Currently we authenticate once with model/version supplied in newJob, get worker_id from PyGrid and use it for all jobs.
Instead, we should authenticate for every newJob and store worker_id for given job only.
Why?
Each FL model has own auth settings, hence each job will have individual worker_id that works with corresponding model only.
What?
Currently we authenticate once with model/version supplied in
newJob
, getworker_id
from PyGrid and use it for all jobs. Instead, we should authenticate for everynewJob
and storeworker_id
for given job only.Why?
Each FL model has own auth settings, hence each job will have individual worker_id that works with corresponding model only.
Breakdown
newJob
worker_id
onJob
Additional Context
See https://github.com/OpenMined/PyGrid/issues/628