---
license: other
pipeline_tag: time-series-forecasting
library_name: pytorch
tags:
- time-series
- anomaly-detection
- LLM
datasets:
- NIPS-TS-SWAN
- MSL
- SMD
- SWaT
- PSM
---

# TriP-LLM

This is the official checkpoints release for the **TriP-LLM**, a novel framework for unsupervised anomaly detection in multivariate time-series data using pretrained Large Language Models (LLMs).

The model was presented in the paper: [TriP-LLM: A Tri-Branch Patch-wise Large Language Model Framework for Time-Series Anomaly Detection](https://huggingface.co/papers/2508.00047)

## Model Description
- **Name**: TriP-LLM  
- **Task**: Time-Series Anomaly Detection  
- **Framework**: PyTorch 
- **Repository**: [GitHub – YYZStart/TriP-LLM](https://github.com/YYZStart/TriP-LLM)


## Usage

To get started with TriP-LLM, you can follow the installation and usage instructions from the [GitHub repository](https://github.com/YYZStart/TriP-LLM).

### Installation

We conducted our experiments using PyTorch 2.4.1, Python 3.11 and CUDA 12.4.
```bash
pip install torch==2.4.1 torchvision==0.19.1 torchaudio==2.4.1 --index-url https://download.pytorch.org/whl/cu124
```
To install the required dependencies and set up the environment, run the following commands:

```bash
git clone https://github.com/YYZStart/TriP-LLM.git
cd TriP-LLM
pip install -r requirements.txt
```

### Reproduce Experiments

You can reproduce our main experiments with:

```bash
python main.py
```

## 📎 Citation

If you find this repository useful for your research, please cite our paper:

```bibtex
@misc{TriPLLM,
      title={TriP-LLM: A Tri-Branch Patch-wise Large Language Model Framework for Time-Series Anomaly Detection}, 
      author={Yuan-Cheng Yu and Yen-Chieh Ouyang and Chun-An Lin},
      year={2025},
      eprint={2508.00047},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2508.00047}, 
}
```