IBM releases SOTA Granite Time Series PatchTST-FM-r2 model with commercial-friendly license
Source
huggingface.co
Date
Context
IBM's PatchTST-FM-r2 takes a history of regularly sampled measurements and predicts future values without requiring a separate training run for that dataset. It can return both a central forecast and quantiles, estimates describing different points in the predicted distribution. That lets an application represent uncertainty alongside its forecast. Release article.
The release includes weights, implementation and an inferenceRunning a trained model to get answers — the phase where AI is actually used, as opposed to trained.Full definition → pipeline, with a choice of Apache 2.0 or OpenMDW 1.0 licensing. Its concrete value is an available forecasting system that teams can evaluate on their own demand, energy or telemetry data, including the spread of predicted outcomes rather than only one number.
Terms in this piece · Glossary
inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
