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
We utilize create_custom_training_job_from_component to take advantage of other Vertex-specific functionality like distributed training clusters and NFS mounts, within our pipeline construction and when newly added features are added we are struggling to take advantage of them because they are missing from KFP.
Is there a workaround currently?
We can add it into the machine spec manually by editing the spec returned from create_custom_training_job_from_component, but a much better user experience would be for create_custom_training_job_from_component to accept maxWaitDuration and strategy
Love this idea? Give it a 👍.
The text was updated successfully, but these errors were encountered:
Feature Area
/area components
What feature would you like to see?
Add support for defining the
maxWaitDuration
so Vertex can be used with DWShttps://cloud.google.com/vertex-ai/docs/training/schedule-jobs-dws#examples
What is the use case or pain point?
We utilize
create_custom_training_job_from_component
to take advantage of other Vertex-specific functionality like distributed training clusters and NFS mounts, within our pipeline construction and when newly added features are added we are struggling to take advantage of them because they are missing from KFP.Is there a workaround currently?
We can add it into the machine spec manually by editing the spec returned from
create_custom_training_job_from_component
, but a much better user experience would be forcreate_custom_training_job_from_component
to acceptmaxWaitDuration
andstrategy
Love this idea? Give it a 👍.
The text was updated successfully, but these errors were encountered: