HeatWave ML has the following limitations:
ML_TRAINroutine does not support MySQL user names that contain a period; for example, a user named
'joe.smith'@'cannot run the
ML_TRAINroutine. The model catalog schema created by the
ML_TRAINprocedure incorporates the user name in the schema name (e.g.,
ML_SCHEMA_joesmith), and a period is not a permitted schema name character.
The table used to train a model (the training dataset) cannot exceed 10 GB, 100 million rows, or 900 columns.
To avoid taking up too much space in memory, the number of loaded models should be limited to three.
“Bring your own model” is not supported. Use of non-HeatWave ML models or manually modified HeatWave ML models can cause undefined behavior.
There is currently no way to monitor HeatWave ML query progress.
ML_TRAINis typically the most time consuming routine. The time required to train a model depends on the number of rows and columns in the dataset and the specified
ML_TRAINparameters and options.
ML_PREDICT_TABLEare compute intensive processes, with
ML_EXPLAIN_TABLEbeing the most compute intensive. Limiting operations to batches of 10 to 100 rows by splitting large tables into smaller tables is recommended.
ML_EXPLAIN_*routines limit explanations to the 100 most relevant features.
Concurrent HeatWave analytics and HeatWave ML queries are not supported. A HeatWave ML query must wait for HeatWave analytics queries to finish, and vice versa. HeatWave analytics queries are given priority over HeatWave ML queries.
On AWS, HeatWave Machine Learning is only supported with the
HeatWave.256GBnode shape. If you intend to use HeatWave Machine Learning functionality, select that shape when creating a HeatWave Cluster.