Chronos T5 Tiny is a time series forecasting model hosted on Hugging Face. It forms part of the Chronos family of pretrained foundation models for time series data and is based on the T5 architecture.
The model supports probabilistic forecasting by generating predictions at specified quantile levels. It accepts historical data in pandas DataFrame format and produces forecasts for a user-defined number of future steps. Integration occurs through the dedicated Chronos library, which provides a BaseChronosPipeline class for loading the model and running inference.
Usage begins with installation of the chronos-forecasting package followed by loading the model with a device mapping such as CUDA for GPU acceleration. An example workflow reads a CSV file containing time series records, then calls predict_df with parameters for prediction length, quantile levels, and an identifier column. The model card lists it under the Apache 2.0 license and references two arXiv papers that describe its development.
It is delivered as a downloadable model repository on the Hugging Face platform, with associated Safetensors weights and instructions for libraries, inference providers, notebooks, and local applications. The repository is maintained under the Amazon organization on Hugging Face.
In the Data science & ML workbench space, Chronos T5 Tiny takes a focused approach. It enables accurate time series forecasting for predictive analytics and business planning. It is built as an open-source project for data scientists and analysts. Chronos T5 Tiny is open source under the Open Source license. It runs on the web, the command line, and API, and it can be self-hosted.
Behind Chronos T5 Tiny is Amazon, and the product first shipped in 2024. Key capabilities include time series forecasting, predictive analytics, and T5 architecture.
Latest indexed changes and source events
amazon/chronos-t5-tiny verified by the PulseGate indexer
Other apps tracked under the same category.
