TSFM: Time Series Foundation Models

Public notebooks, utilities, and serving components for working with Time Series Foundation Models (TSFM).

The core TSFM time series models have been made available on Hugging Face -- check out the granite time series collection here. Additional details can be found in our wiki. Information on the services component can be found here.

Python Version

The current Python versions supported are 3.10, 3.11, 3.12, 3.13.

Initial Setup

First clone the repository:

git clone "https://github.com/ibm-granite/granite-tsfm.git" 
cd granite-tsfm

📕 Notebooks Installation

Several notebooks are provided in the notebooks folder. They allow you to perform pre-training and finetuning on the models. To install use pip:

pip install ".[notebooks]"

🔗 Links to the notebooks

• Getting started with PatchTSMixer [Try it out]
• Transfer learning with PatchTSMixer [Try it out]
• Transfer learning with PatchTST [Try it out]
• Getting started with TinyTimeMixer (TTM) [Try it out]
• TTM full benchmarking scripts and results are available [here]
• FlowState example notebook is available [here]
• PatchTST-FM example notebook is available [here]

📗 Google Colab Tutorials

Run the TTM tutorial in Google Colab, and quickly build a forecasting application with the pre-trained TSFM models.

• TTM Colab Tutorial

💻 Demos Installation

The demo presented at NeurIPS 2023 is available in tsfmhfdemos. This demo requires you to have pre-trained and finetuned models in place (we plan to release these at a later date). To install the requirements use pip:

pip install ".[demos]"

🪲 Issues

If you encounter an issue with this project, you are welcome to submit a bug report. Before opening a new issue, please search for similar issues. It's possible that someone has already reported it.

🌏 Wiki

Wiki Page

Notice

The intention of this repository is to make it easier to use and demonstrate Granite TimeSeries components that have been made available in the Hugging Face transformers library. As we continue to develop these capabilities we will update the code here.

IBM Public Repository Disclosure: All content in this repository including code has been provided by IBM under the associated open source software license and IBM is under no obligation to provide enhancements, updates, or support. IBM developers produced this code as an open source project (not as an IBM product), and IBM makes no assertions as to the level of quality nor security, and will not be maintaining this code going forward.