Skip to main content
Version: 6.1

ML Studio

Overview​

ML Studio is a Search Anywhere Framework module for managing the machine learning model lifecycle. The module allows you to register algorithms, deploy them to compute environments, train models, and apply them to data. Management operations are performed in the Search Anywhere Framework interface, while computations are handled by sm-ml-service. Service configuration is described in SM-ML Settings.

ML Studio is not tied to specific algorithms. Each algorithm is connected as a separate object consisting of a container image and notebook code. Algorithms run in isolated environments whose lifecycle is managed by the module.

The distribution includes three ready-to-use example algorithms (classifiers):

  • Binary Neural Network Classifier - binary classification
  • Logistic Regression - logistic regression
  • Multi Class Neural Network Classifier - multiclass classification

You can use these examples without modification or as a basis for your own algorithms.

Entity Relationships​

  • Algorithm - an implementation of a machine learning method. An algorithm is associated with the container image in which its code runs
  • Runtime environment - a compute environment (Docker or Kubernetes) in which images are deployed and jobs run
  • Deployment - an algorithm instance running in a specific environment with assigned resources (CPU, memory, and GPU). Jobs send requests to a deployment
  • Job - a single execution of a fit (training) or apply (application) operation on a deployment
  • Model - the result of successful training. A model is associated with the algorithm and the job that created it

The module workflow is as follows: register an algorithm - deploy its image to an environment - start training - obtain a model - apply the model.

Documentation Sections​

  • ML Studio Interface - a description of the module screens, forms, and fields
  • ML Studio Step-by-Step Guide - an end-to-end workflow from environment configuration to model training and application, including the train and predict SAFL commands
  • SM-ML Settings - installation and configuration of sm-ml-service