OpenLTM

OpenLTM is an open codebase aiming to provide a pipeline to develop and evaluate large time-series models.

For deep time series models, we recommend Time-Series-Library and this comprehensive Survey.

Try out our out-of-the-box time series foundation models in Large-Time-Series-Model or HuggingFace.

Updates

:triangular_flag_on_post: News (2025.05) We release 260B pre-trained Timer in PyTorch. This version is equivalent to Huggingface Model, but more user-friendly for fine-tuning. This notebook shows how to load and use the checkpoint.

:triangular_flag_on_post: News (2025.04) Many thanks for the implementation of TTMs and other LLM4TS methods from frndtls.

:triangular_flag_on_post: News (2024.12) Many thanks for the implementation of GPT4TS from khairulislam.

:triangular_flag_on_post: News (2024.10) We include several large time-series models, release pre-training code, and provide scripts.

What is LTM?

LTM (Large Time-Series Model, aka Time Series Foundation Model, TSFM) is a deep time series models built on scalable backbones (e.g., Transformers) via large-scale pre-training, which will be applied to a variety of time series data (e.g., zero-shot forecasting) and downstream tasks (e.g., general feature extraction of time series). For more information, here we list some tutorials: [CN], [EN].

Model Checklist

• Timer-XL - Timer-XL: Long-Context Transformer for Unified Time Series Forecasting. [ICLR 2025], [Code]
• Moirai - Unified Training of Universal Time Series Forecasting Transformers. [ICML 2024], [Code]
• Timer - Timer: Generative Pre-trained Transformers Are Large Time Series Models. [ICML 2024], [Code]
• Moment - MOMENT: A Family of Open Time-series Foundation Model. [ICML 2024], [Code]
• TTMs - Tiny Time Mixers (TTMs): Fast Pre-trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate Time Series. [Arxiv 2024], [Code]
• GPT4TS - One Fits All: Power General Time Series Analysis by Pretrained LM. [NeurIPS 2023], [Code]
• Time-LLM: . Time-LLM: Time Series Forecasting by Reprogramming Large Language Models. [ICLR 2024], [Code]
• AutoTimes: Autoregressive Time Series Forecasters via Large Language Models. [NeurIPS 2024], [Code]
• Sundial: A Family of Highly Capable Time Series Foundation Models. [ICML 2025], [Code]
• LLMTime: Large Language Models Are Zero-Shot Time Series Forecasters. [NeurIPS 2023], [Code]
• Chronos: Learning the Language of Time Series. [TMLR 2024], [Code]
• Time-MoE: Billion-Scale Time Series Foundation Models With Mixture Of Experts. [ICLR 2025], [Code]
• A Decoder-Only Foundation Model for Time-Series Forecasting. [ICML 2024], [Code]

Usage

1. Install Python 3.11 For convenience, execute the following command.

pip install -r requirements.txt

2. Place downloaded data in the folder ./dataset. Here is a dataset summary.

• For univariate pre-training (skip this step if you use a pre-trained checkpoint):

  • UTSD contains 1 billion time points for large-scale pre-training (in numpy format): [Download].
  • ERA5-Family (40-year span, thousands of variables) for domain-specific model: [Download].

• For supervised training or modeling adaptation

  • Datasets from TSLib : [Download].

3. Training from scratch or using our pre-trained checkpoint in the folder ./checkpoints

• We provide the checkpoint pre-trained on 260B time points [Download].

4. We provide some supervised training, pre-training, and adaptation scripts under the folder ./scripts/:

# Supervised training
# (a) one-for-one forecasting
bash ./scripts/supervised/forecast/moirai_ecl.sh
# (b) one-for-all (rolling) forecasting
bash ./scripts/supervised/rolling_forecast/timer_xl_ecl.sh

# Large-scale pre-training
# (a) pre-training on UTSD
bash ./scripts/pretrain/timer_xl_utsd.sh
# (b) pre-training on ERA5
bash ./scripts/pretrain/timer_xl_era5.sh

# Model adaptation
# (a) full-shot fine-tune
bash ./scripts/adaptation/full_shot/timer_xl_etth1.sh
# (b) few-shot fine-tune
bash ./scripts/adaptation/few_shot/timer_xl_etth1.sh

5. Develop your large time-series model.

• Add the model file to the folder ./models. You can follow the ./models/timer_xl.py.
• Include the newly added model in the Exp_Basic.model_dict of ./exp/exp_basic.py.
• Create the corresponding scripts under the folder ./scripts.

Recommended Resources

Here we list some resources of LTMs, which support out-of-box usage (e.g., zero-shot forecasting):

• Sundial: https://huggingface.co/thuml/sundial-base-128m
• Timer: https://huggingface.co/thuml/timer-base-84m
• Chronos: https://huggingface.co/amazon/chronos-t5-base
• Moirai: https://huggingface.co/Salesforce/moirai-1.0-R-base
• TimesFM: https://huggingface.co/google/timesfm-1.0-200m
• Time-MoE: https://huggingface.co/Maple728/TimeMoE-50M
• TTMs: https://huggingface.co/ibm-research/ttm-research-r2

[!NOTE] LTMs are still small compared to foundation models of other modalities. For example, it is okay to use CPUs for inference, RTX 4090s for adaptation, A100s for pre-training.

Citation

If you find this repo helpful, please cite our paper.

@inproceedings{liutimer,
  title={Timer: Generative Pre-trained Transformers Are Large Time Series Models},
  author={Liu, Yong and Zhang, Haoran and Li, Chenyu and Huang, Xiangdong and Wang, Jianmin and Long, Mingsheng},
  booktitle={Forty-first International Conference on Machine Learning}
}

@article{liu2024timer,
  title={Timer-XL: Long-Context Transformers for Unified Time Series Forecasting},
  author={Liu, Yong and Qin, Guo and Huang, Xiangdong and Wang, Jianmin and Long, Mingsheng},
  journal={arXiv preprint arXiv:2410.04803},
  year={2024}
}

@article{liu2025sundial,
  title={Sundial: A Family of Highly Capable Time Series Foundation Models},
  author={Liu, Yong and Qin, Guo and Shi, Zhiyuan and Chen, Zhi and Yang, Caiyin and Huang, Xiangdong and Wang, Jianmin and Long, Mingsheng},
  journal={arXiv preprint arXiv:2502.00816},
  year={2025}
}

Acknowledgment

We appreciate the following GitHub repos a lot for their valuable code and efforts:

• Time-Series-Library (https://github.com/thuml/Time-Series-Library)
• Large-Time-Series-Model (https://github.com/thuml/Large-Time-Series-Model)
• AutoTimes (https://github.com/thuml/AutoTimes)

Contributors

If you have any questions or want to use the code, feel free to contact:

• Yong Liu (liuyong21@mails.tsinghua.edu.cn)
• Guo Qin (qinguo24@mails.tsinghua.edu.cn)
• Haixuan Liu (liuhaixu21@mails.tsinghua.edu.cn)