TrackFormer: Particle Trackfitting with Transformer

Overview

TrackFormer is a unique solution to particle trajectory reconstruction that uses transformer-inspired design. Using Transformers' self-attention mechanism, this model performs track fitting, resulting in better accuracy and efficiency.

Features

• Transformer-based Architecture: For fast efficient and accurate particle track fitting.
• Built with Lightning Integration:
• Modular Design:
• Logging and CLI integration:

Getting Started

1. Clone the repository:
2. download datasets to specific directory
3. cd to appropriate script to train, test split dataset: ./split_dataset.sh /path/to/downloaded/dataset 80 10 10
4. example usage python main.py fit --config configs/tformer.yaml --data ActsDataModule --config configs/trainer.yaml

Important links

• slides

• report