Time-Series Transformer (Timer)
Update (2025.5): We release a generative time series foundation model Sundial on HuggingFace.
Update (2025.2) We release OpenLTM for pre-training/fine-tuning large time-series models.
Timer is a large time-series model introduced in this paper and enhanced by subsequent work.
This version is univariate pre-trained on 260B time points with 84M parameters, a lightweight generative Transformer for zero-shot point forecasting.
We evaluate the model on the following benchmark: TSLib Dataset.
For more information, please see the GitHub.
There's indeed room for improvement on this model. We are actively working around it and are glad to see constructive suggestions and noteworthy cases :)
Quickstart
pip install transformers==4.40.1 # Use this version and Python 3.10 for stable compatibility
import torch
from transformers import AutoModelForCausalLM
# load pretrain model
model = AutoModelForCausalLM.from_pretrained('thuml/timer-base-84m', trust_remote_code=True)
# prepare input
batch_size, lookback_length = 1, 2880
seqs = torch.randn(batch_size, lookback_length)
# generate forecast
prediction_length = 96
output = model.generate(seqs, max_new_tokens=prediction_length)
print(output.shape)
A notebook example is also provided here. Try it out!
Specification
- Architecture: Causal Transformer (Decoder-only)
- Pre-training Scale: 260B time points
- Context Length: up to 2880
- Parameter Count: 84M
- Patch Length: 96
- Number of Layers: 8
Adaptation
For developers interest in fine-tune this model, we provide model checkpoint and code implementation in OpenLTM.
Acknowledgments
This work was supported by the National Natural Science Foundation of China (62022050 and U2342217), the BNRist Innovation Fund (BNR2024RC01010), and the National Engineering Research Center for Big Data Software.
The model is mostly built from the Internet public time series dataset, which comes from different research teams and providers. We sincerely thank all individuals and organizations who have contributed the data. Without their generous sharing, this model would not have existed.
Citation
If you find Timer or Timer-XL helpful for your research, 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}
}
Contact
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)
License
This model is licensed under the Apache-2.0 License.