A Fabric function for scoring MLflow models in batches from a notebook, provided the models have signatures; you can call it via MLFlowTransformer, Spark SQL or a PySpark UDF. Fabric Warehouse does not support the T-SQL PREDICT statement.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains PREDICT in context, with comparison tables and the common traps.
Terms in this definition
- MLflow
An open-source framework for experiment tracking and model management. Azure Machine Learning adopts it as its tracking API, so each workspace acts as a compatible MLflow tracking server.
- Notebook
Combines runnable code cells, in Python, SQL, Scala or R, with their results in one interactive Azure Databricks document; it runs on attached compute, and a job can run it as a task.
- Serverless
Compute tier for single Azure SQL databases that scales automatically, pauses when idle and charges by the second. It is offered in General Purpose and Hyperscale, not Business Critical, and reserved capacity does not apply.
- UDF
Custom logic written in JavaScript and registered with a Cosmos DB container, callable only from queries. Calling one raises the request unit charge.
- Fabric Warehouse
A relational data warehouse in the classic style, offered as a Fabric item, that supports T-SQL fully, transactions included.
- T-SQL
The dialect of SQL that Microsoft uses for Azure SQL and SQL Server. Azure Monitor logs are queried with KQL instead.
Related terms
- LLM
Large language model, usually a transformer network with billions of parameters that learnt to predict the next token from vast amounts of text. Compared with a small language model it is more capable across tasks, but slower and costlier.
- ML
Building models that infer patterns from data so they can predict outcomes, for instance through regression or classification, using tools like Azure Machine Learning or Synapse Spark pools. Predictive ML is distinct from generative AI.
- Supervised learning
Training a model on examples whose correct output is labelled so it learns to predict that output; regression and classification are typical cases.
- System Insights
A Windows Server capability that uses machine-learning models running on the server itself to predict future CPU, storage and network usage from performance data.
- Time-series forecasting
A machine learning workload, not a generative or NLP one, that uses historical time-stamped data to predict future values like demand or sales.