VertexAI autoML pipeline template error (tabular-regression)

[Zoltara]

I have been using the the autoML regression pipeline templates for a while successfully.  Three days ago overnight they broke.  The same job that worked the night before hangs with the error message: "Unable to create pipeline run due to the following error: Input parameter type mismatch. PipelineSpec.root.input_definitions.parameters['dataflow_use_public_ips'] is defined as BOOLEAN that parses BOOL_VALUE type, but the default value is provided as NUMBER_VALUE type."

If I want to re-run a clone of a previously successful training the same happens.  I have tried to set public IP setting to different values -- no success.  I have downloaded and edited the yaml according to the error-- no luck either!

Anybody encountered the same?  Is there a workaround?

[Lloyd225]

Hello,

It seems like a recent update may have caused a compatibility issue with your pipeline. Try rolling back to a previous version or reaching out to the autoML support team for assistance with resolving the issue.

MyMorri App

Best regard,
Lloyd225

[Zoltara]

Hi Lloyd225

Thanks for reaching out!  The issue is that the original pipeline template on GCP is broken!
It does not work "out of the box"!

I don't know how to reach the autoML support team, I searched for it but came up empty.
Could you point me to a channel to reach out to?

Thanks

[Zoltara · Accepted solution]

OK!  The original Google provided template is now fixed and running!

Thanks for the help!