GLM quasibinomial regression gives error on real-valued targets in [0,1], seems to expect binary labels

Description

I have a dataset with binary targets in principle, but some uncertain values are coded with scores between 0 and 1. I am using Python H2O. GBM models with quasibinomial loss work as expected. GLM models fail. Here is a part of the stacktrace:
OSError: Job with key $03017f00000132d4ffffffff$_bcd6171c27e81260f58dac108f8e4f5a failed with an exception: java.lang.IllegalArgumentException: Actual column must contain binary class labels, but found cardinality 5!

Assignee

Wendy

Fix versions

Reporter

Thomas Vacek

Support ticket URL

None

Labels

Affected Spark version

None

Customer Request Type

None

Task progress

None

CustomerVisible

Yes

Components

Affects versions

Priority

Major
Configure