|
# GRU Sequence Anomaly Detector |
|
|
|
This project provides an open-source, bidirectional GRU-based deep learning model to detect anomalies in time-series transactional data. It is designed to be general-purpose and supports transfer learning. |
|
|
|
## π Project Structure |
|
|
|
- `models/` β Contains the model architecture and trained weights |
|
- `pipeline/` β Core training, evaluation, and export logic |
|
- `utils/` β Logging and utility functions |
|
- `notebooks/` β Example usage and exploration |
|
- `tests/` β Unit and integration tests |
|
- `main.py` β Entry script to run training/evaluation |
|
- `fine_tune_template.py` β Script for model fine-tuning on external datasets |
|
- `model_card.md` β Model documentation and expected usage |
|
- `requirements.txt` β All required dependencies |
|
|
|
## π¦ Pretrained Models |
|
|
|
- `models/txn_anomaly_model.pt` β PyTorch model file for fine-tuning or loading |
|
- `models/txn_anomaly_model.onnx` β ONNX model file for deployment in other runtimes |
|
|
|
## π Quick Start |
|
|
|
```bash |
|
pip install -r requirements.txt |
|
python main.py |
|
``` |
|
|
|
To fine-tune: |
|
|
|
```bash |
|
python fine_tune_template.py --data your_dataset.csv |
|
``` |
|
|
|
|