colliderml 0.3.0

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

ColliderML

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

Features

• Efficient parallel data downloading with resume capability
• Support for common HEP data formats
• Machine learning utilities for particle physics
• Visualization tools for physics data

Installation

For Users

# Create and activate environment
conda create -n collider-env python=3.11  # 3.10 or 3.11 recommended
conda activate collider-env

# Install package
pip install colliderml

For Developers

# Create and activate environment
conda create -n collider-dev python=3.11  # 3.10 or 3.11 recommended
conda activate collider-dev

# Clone repository
git clone https://github.com/OpenDataDetector/ColliderML.git
cd ColliderML

# Install in development mode with extra dependencies
pip install -e ".[dev]"

Quick Start

CLI

# Download 100 events from the taster campaign into ./data
colliderml get -c taster -e 100 -O data

from colliderml.core.data.manifest import ManifestClient
from colliderml.core.io import DataDownloader

manifest = ManifestClient()
files = manifest.select_files(campaign=None, datasets=["ttbar"], objects=["tracks"], max_events=1000)

downloader = DataDownloader()
results = downloader.download_files([f.path for f in files], local_dir="data", max_workers=4, resume=True)

for path, result in results.items():
    print(path, result.success, result.error)

Features

• Manifest-driven: Always selects files from the latest portal manifest
• Parallel Downloads: Download multiple files concurrently
• Resume Capability: Optionally resume interrupted downloads
• Progress Tracking: Real-time progress bars
• Clear Errors: Helpful failure messages and HEAD checks

Development

1. Activate your environment:

   conda activate collider-dev
   

2. Run tests:

   # Run unit tests only
   pytest -v -m "not integration"
   
   # Run all tests including integration tests
   pytest -v
   
   # Run with coverage report
   pytest --cov=colliderml
   

3. Build documentation:

   mkdocs build
   mkdocs serve  # View at http://127.0.0.1:8000
   

License

MIT License