[SPARK-26410] Support per Pandas UDF configuration

• Type: New Feature
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.1.0
• Component/s: PySpark
• Assignee: Unassigned
• Reporter: Xiangrui Meng
• Created: 19/Dec/18 17:36
• Updated: 17/Mar/20 09:46

Description

We use a "maxRecordsPerBatch" conf to control the batch sizes. However, the "right" batch size usually depends on the task itself. It would be nice if user can configure the batch size when they declare the Pandas UDF.

This is orthogonal to SPARK-23258 (using max buffer size instead of row count).

Besides API, we should also discuss how to merge Pandas UDFs of different configurations. For example,

df.select(predict1(col("features"), predict2(col("features")))

when predict1 requests 100 rows per batch, while predict2 requests 120 rows per batch.

cc: icexelloss bryanc holdenk hyukjin.kwon ueshin smilegator