Packages

  • package root
    Definition Classes
    root
  • package ai
    Definition Classes
    root
  • package eto
    Definition Classes
    ai
  • package rikai
    Definition Classes
    eto
  • package sql

    Rikai SQL-ML extension.

    Rikai SQL-ML extension.

    Rikai offers DDL to manipulate ML Models:

    CREATE MODEL model_name
[ OPTIONS (key=value, key=value, ...) ]
[ AS "model_registry_uri" ]

# List all registered models.
SHOW MODELS

# Describe the details of a model.
(DESC | DESCRIBE) MODEL model_name

# Drop a Model
DROP MODEL model_name

    A ML_PREDICT function is implemented to run model inference.

    SELECT id, ML_PREDICT(model_name, col1, col2, col3) as predicted FROM table
    Definition Classes
    rikai
  • package model
    Definition Classes
    sql
  • package spark
    Definition Classes
    sql
  • package execution
  • package expressions
  • package functions
  • Python
  • RikaiSparkSessionExtensions
  • SparkRunnable
p

ai.eto.rikai.sql

spark

package spark

Ordering
  1. Alphabetic
Visibility
  1. Public
  2. All

Type Members

  1. trait Python extends AnyRef

    Python is the callback service to call arbitrary Python code in the SparkSessions' main python interpreter.

  2. class RikaiSparkSessionExtensions extends (SparkSessionExtensions) ⇒ Unit

    Rikai SparkSession extensions to enable Spark SQL ML.

  3. trait SparkRunnable extends AnyRef

    Make ai.eto.rikai.sql.model.Model runnable on Spark.

    Make ai.eto.rikai.sql.model.Model runnable on Spark.

    For a ML_PREDICT expression in Spark SQL,

    SELECT ML_PREDICT(model_zoo, col1, col2, col3) FROM t1

    It generates a LogicalPlan equivalent to

    SELECT <Model{model_zoo}.asSpark(col1, col2, col3)> FROM t1
    Example:

    1. To implement a ai.eto.rikai.sql.model.Model for RegistryFoo:

      class FooModel(name, uri) extends Model with SparkRunnable {

    /** Use a Spark UDF with the same name to run RegistryFoo's model */
   def asSpark(args: Seq[Expression]) : Expression = {
       UnresolvedFunction(
          new FunctionIdentifier(s"${name}"),
          arguments,
          isDistinct = false,
          Option.empty
       )
   }
    Note

    It is the ai.eto.rikai.sql.model.Registry's responsibility to implement a Model that runs with Spark SQL.

Value Members

  1. object Python

Ungrouped