Supervisor should cancel all jobs that do not have a trial anymore

ErikGartner / hyperdock

A distributed hyperparameter optimizer for machine learning that lives in Docker

https://gartner.io/hyperdock/

Apache License 2.0

8 stars 2 forks source link

Supervisor should cancel all jobs that do not have a trial anymore #70

Closed ErikGartner closed 5 years ago

ErikGartner commented 5 years ago

A crash occured leaving a workqueue item without a trial somehow. The supervisor should monitor the database for any inconsistencies and rectify those workqueue items.