Hit Identification inside SuperFGD with Transformers

Basic use

• To run the baseline configuration, run: python main.py -f "config/baseline.yaml"

-To run in test mode, simply add a -t argument above.

Overview

• config holds the different configuration files for training and testing.
• data holds several files for data handling, Lightning DataModule creation and preprocessing.
• models contains the main engine in engine_nodecl.py and the different variants of transformer encoders that have been tested in transformer_encoder.py
• notebooks gives notebooks to illustrate the work. In sfgd_eda.ipynb you'll basic data analysis on the dataset. model_inference.ipynb displays attention maps and the model predictions compared to the ground truth. umap_embedding.ipynb applies the UMAP dimensionality reduction on the last layer output of the model (before the linear classifier).