Retail Sales Forecasting using the M5 dataset with Granite Time Series - Few-shot finetuning, evaluation, and visualization

In this tutorial, we will explore time series forecasting using the IBM Granite Timeseries model to predict retail sales. We will cover key techniques such as few-shot forecasting and fine-tuning. We are using M5 datasets from the official M-Competitions repository to forecast future sales aggregated by state. The aim of this recipe is to showcase how to use a pre-trained time series foundation model for multivariate forecasting and explores various features available with Granite Time Series Foundation Models.

This lab uses TinyTimeMixers (TTMs), which are compact pre-trained models for Multivariate Time-Series Forecasting, open-sourced by IBM Research. With less than 1 Million parameters, TTM introduces the notion of the first-ever "tiny" pre-trained models for Time-Series Forecasting. TTM outperforms several popular benchmarks demanding billions of parameters in zero-shot and few-shot forecasting and can easily be fine-tuned for multivariate forecasts.

Prerequisites

This lab is a Jupyter notebook. Please follow the instructions in pre-work to run the lab.

Lab

To run the notebook from your command line in Jupyter using the active virtual environment from the pre-work, run:

jupyter notebook notebooks/Bike_Sharing_Finetuning_with_Exogenous.ipynb

The path of the notebook file above is relative to the granite-workshop folder from the git clone in the pre-work.

Credits

This notebook is a modified version of the IBM Granite Community Retail Forecasting using M5 Sales Data Few-shot, Fine-tuning, Evaluation, and Visualization notebook. Refer to the IBM Granite Community for the official notebooks.