colliderml 0.3.1

A modern machine learning library for high-energy physics data analysis

ColliderML

A modern machine learning library for high-energy physics data analysis.

Installation

pip install colliderml

For development: pip install -e ".[dev]"

Getting the data

Option 1 — CLI (download to local cache, then load with the library):

colliderml download --channels ttbar --pileup pu0 --objects particles,tracker_hits,calo_hits,tracks --max-events 200

Cache location: default ~/.cache/colliderml, or set COLLIDERML_DATA_DIR. List downloaded configs: colliderml list-configs.

Option 2 — HuggingFace only:

from datasets import load_dataset
dataset = load_dataset("CERN/ColliderML-Release-1", "ttbar_pu0_particles", split="train")

Using the library

The notebook notebooks/colliderml_loader_exploration.ipynb shows how to use the convenience functions: loading local Parquet with load_tables, exploding event tables, pileup subsampling, calibration, and plotting.

Full docs: https://opendatadetector.github.io/ColliderML

Development

pytest -v -m "not integration"

Docs are built with VitePress: npm ci --prefix docs && npm run --prefix docs docs:build.

License

MIT License