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
Of course, I tried koalas with a model with 500+ features. :)
Unless I'm mistaken (quite possible), this is an easy fix to PySpark. I'm creating the issue here because of the impact on potential koalas users and hoping y'all can encourage the PySpark fix.
The text was updated successfully, but these errors were encountered:
When running a model to generate predictions, koalas+mlflow create PySpark UDFs that take an argument per feature column.
PySpark currently has a limitation of 256 arguments per UDF. This limitation seems shallow and easy to hack around. See https://issues.apache.org/jira/browse/SPARK-28978.
Of course, I tried koalas with a model with 500+ features. :)
Unless I'm mistaken (quite possible), this is an easy fix to PySpark. I'm creating the issue here because of the impact on potential koalas users and hoping y'all can encourage the PySpark fix.
The text was updated successfully, but these errors were encountered: