Matei Zaharia created SPARK-1428: ------------------------------------ Summary: MLlib should convert non-float64 NumPy arrays to float64 instead of complaining Key: SPARK-1428 URL: https://issues.apache.org/jira/browse/SPARK-1428 Project: Spark Issue Type: Improvement Components: MLlib, PySpark Reporter: Matei Zaharia Priority: Minor
Pretty easy to fix, it would avoid spewing some scary task-failed errors. The place to fix this is _serialize_double_vector in _common.py. -- This message was sent by Atlassian JIRA (v6.2#6252)