You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Trainingjob is a stateful application: each parameter server and worker needs to be uniquely addressable to support all the different patterns of distributed training
Trainingjob
is a stateful application: each parameter server and worker needs to be uniquely addressable to support all the different patterns of distributed trainingTensorflow/k8s
create many Kubernetes Jobs with unique stateful name, so they avoid to use etcd to get the unique-persist name: https://github.com/tensorflow/k8s/blob/master/tf_job_design_doc.md#controllerThe text was updated successfully, but these errors were encountered: