REST interfaces have become popular for web application development in particular as they provide a lightweight approach for interacting with HTTP data and services. Developers increasingly rely on REST endpoints to provide advanced analytics functionality in applications. With the introduction of Oracle Machine Learning Services with Oracle Autonomous Database, Oracle makes it easy for data science teams and application developers to manage and deploy machine learning models using a REST API for ease of application integration. As such, OML Services facilitates collaborations across the data science team.
MLOps, that is machine learning plus operations, is in part about collaboration and communication between data scientists and operations professionals, including application developers to manage the machine learning lifecycle. A key part of this lifecycle involves the deployment, update, versioning, and use of machine learning models.
With Oracle Machine Learning in Oracle Database and Oracle Autonomous Database, data scientists build in-database models that reside as first-class objects in the database, where deployment is immediate. For example, users invoke SQL queries used for accessing model details and scoring data, both in batch and for more real-time applications. However, with OML Services, those same models can be deployed through REST endpoints hosted through Oracle Autonomous Database to enable real-time and mini-batch scoring.
Since data science teams will also build models using open source frameworks such as scikit-learn, TensorFlow, MXNet, among others, OML Services also support import of third-party classification and regression models exported in ONNX format. This enables MLOps for your in-database and third-party models in the Oracle Autonomous Database environment. Users benefit from the performance and scalability of Oracle Cloud Infrastructure for real-time scoring within a common framework.
In addition, OML Services supports proprietary cognitive text capabilities – also from REST endpoints – including topic discovery, keywords, summary, and similarity with support for English, Spanish, and French, and sentiment for English. Cognitive text is enabled using an OML-provided pre-built model based on millions of Wikipedia articles. OML Services also supports import of third-party ONNX-format cognitive image models for the scoring of images and tensors.
Features of Oracle Machine Learning Services include:
OML Services is integrated with other Oracle Machine Learning components and the broader Oracle ecosystem. Through OML AutoML UI – a no-code user interface supporting automated machine learning, users immediately deploy models to OML Services with just a few clicks. Similarly, the OML Models interface with Autonomous Database enables deploying models produced using the OML4SQL or OML4Py APIs with just a few clicks as well. In addition, applicable OML in-database models produced in any Oracle Database can be exported and imported into OML Services. Applicable models produced using Oracle Cloud Infrastructure Data Science and exported using ONNX format can also be used with OML Services. Other Oracle components such as Oracle APEX, a low-code development platform for building scalable, secure enterprise applications, can readily uptake OML model results.
See our OML Services GitHub repository for examples using OML Services REST API and the OML Services User's Guide. Learn more about Oracle Machine Learning and try it today using your Always Free Services from Oracle.
OML Services is available with Oracle Autonomous Database 19c and 21c, with cost directly related to compute usage and subject to standard Autonomous Database pricing terms, including auto scaling.