Using a Script Language Container

Loading MLflow models in a UDF or using the REST API UDFs requires the installation of the Exasol MLflow Plugin and its dependencies into the Exasol Database. This is done by creating a dedicated Script Language Container (SLC) for running the UDF.

Depending on the models you want to use in your UDF implementation, you might need to add additional libraries as dependencies to be included in the SLC. See also

• The MLflow documentation on dependencies in general

• An overview of MLflow optional dependencies (aka. “Extras”)

• The entire list of MLflow extras

ALTER SESSION or ALTER SYSTEM Statement

For general information on SLCs, incl. deployment (upload to Exasol’s Bucket File System) and activation, see the documentation at docs.exasol.com.

When activating the Exasol’s MLflow SLC, the language is python and the language alias must be EXA_MLFLOW.

EXA_MLFLOW=localzmq+protobuf:///<bucketfs_name>/<bucket_name>/<path_in_bucket>/<container_name>/?lang=python#buckets/<bucketfs_name>/<bucket_name>/<path_in_bucket>/<container_name>/exaudf/exaudfclient