Search code examples
pythongoogle-cloud-platformxgboostgoogle-ai-platform

How can I do batch predictions on Google AI Platform Unified using a XGBoost model?

How can I use XGBoost on Google AI Platform Unified for batch predictions?

I can perform Online predictions with no problem, but when I submit a batch job request, it returns an error ("Could not predict, DMatrix initialization error).

I'm submiting a csv file with the same column names as the ones used for training. I try doing it with the dataframe formatted as jsonl file and have no result. Is that a bug or am I doing something wrong?

Solution

  • I found out: you can't. Perform batch predictions using Apache Beam running on Dataflow instead. Runs like a charm.