Watch MindsDB AI Research team presenting at Pytorch Ecosystem Day.
Pytorch via SQL commands: A flexible, modular AutoML framework that democratizes ML for database users
Data is the most important ingredient of machine learning (ML), so building models at the data layer can enable seamless pipelines from model creation to inference. Not only does this streamline ML workflows, but also democratizes machine learning by bringing it to the hands of those who work with data, i.e. database users. So, what if we could make PyTorch models available inside any databases on the market?
To address this situation, we are developing MindsDB, an open-source, PyTorch-based ML platform that abstracts models as virtual database tables so they can be operated via SQL commands. PyTorch enables building models with complex inputs and outputs, including time-series data, text, and audiovisual data. However, such models require expertise and time to build. Therefore, in order to win the hearts and minds of regular database users, we have to automate this workflow.
For most data types that PyTorch supports, we can automatically generate a “good enough” model and data-processing pipeline from just the raw data and the endpoint. In this presentation, we would focus on the core component of MindsDB called Lightwood, which performs exactly this on top of the PyTorch framework.