FMTK: A Modular Toolkit for Composable Time Series Foundation Model Pipelines

2025-11-30

AI Summary

Existing time-series foundation models (TSFMs) suffer from non-modular, non-reproducible adaptation pipelines that rely on ad-hoc, model-specific implementations. Method: We propose a lightweight, open-source TSFM toolkit featuring standardized backbone interfaces and pluggable abstractions for encoders, decoders, and adapters—enabling a highly modular fine-tuning framework. This design supports flexible cross-model and multi-task composition, significantly improving pipeline reusability, maintainability, and development efficiency. Results: Experiments demonstrate that high-performance TSFM pipelines can be constructed with only ~7 lines of code while maintaining state-of-the-art performance across long-term forecasting, anomaly detection, and other downstream tasks. Our core contribution is the first unified, decoupled componentization paradigm for TSFMs—establishing foundational infrastructure to support industrial-scale adaptation of time-series foundation models.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Natural Language Processing: (Large) Language Models
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Foundation models (FMs) have opened new avenues for machine learning applications due to their ability to adapt to new and unseen tasks with minimal or no further training. Time-series foundation models (TSFMs) -- FMs trained on time-series data -- have shown strong performance on classification, regression, and imputation tasks. Recent pipelines combine TSFMs with task-specific encoders, decoders, and adapters to improve performance; however, assembling such pipelines typically requires ad hoc, model-specific implementations that hinder modularity and reproducibility. We introduce FMTK, an open-source, lightweight and extensible toolkit for constructing and fine-tuning TSFM pipelines via standardized backbone and component abstractions. FMTK enables flexible composition across models and tasks, achieving correctness and performance with an average of seven lines of code. https://github.com/umassos/FMTK

Problem

• Addresses modularity in time-series foundation model pipelines
• Reduces ad hoc implementations for model-task composition
• Enables flexible, reproducible pipeline construction with minimal code

Innovation

• Modular toolkit for time series foundation models
• Standardized abstractions for flexible pipeline composition
• Lightweight library enabling few-line code implementations