The shortest route depends on the artifact you already have.
| Goal | What you need | Recommended route |
|---|---|---|
| Check the installation | A clean checkout | Run one quickstart smoke test |
| Understand the method | Nothing | Read Particle flow and MLPF |
| Evaluate a published model | A compatible checkpoint and ROOT file | Run the Key4HEP standalone evaluator |
| Train a model | A prepared TFDS dataset | Download and verify a dataset, then train a model |
| Reproduce a dataset | Generator configuration and detector software | Follow dataset generation |
| Validate CMS reconstruction | CMSSW output | Use the CMS validation workflow |
| Validate CLD or CLIC reconstruction | EDM4hep ROOT plus a checkpoint | Use the Key4HEP validation workflow |
| Compare deployment formats | A checkpoint, model_kwargs.pkl, and TFDS | Run PyTorch-to-ONNX validation |
Three levels of use¶
1. Use published artifacts¶
This is the recommended starting point. Download a dataset or checkpoint and avoid detector simulation. It is the quickest route to model evaluation and training experiments.
Use the dataset catalog to choose a compatible detector, input representation, version, and model recipe.
2. Run a local smoke workflow¶
The scripts in scripts/local_test_*.sh download two small input files, build a small dataset, and run a short CPU training. They establish that the software components fit together. Physics performance requires the dedicated validation workflows.
3. Produce data at scale¶
The Pixi and Snakemake workflow covers simulation, postprocessing, TFDS creation, training, and validation. A full campaign can use many batch jobs and substantial storage. It requires a reviewed site configuration and detector software access.
The generation guide explains the shared stages. The CMS and Key4HEP pages document only the detector-specific differences.
Detector choice¶
Choose CMS for proton-proton collisions in the CMS Run-3 software and detector environment.
Choose CLD for electron-positron studies with the CLD detector concept and Key4HEP/EDM4hep data.
Choose CLIC for electron-positron studies with CLICdet and Key4HEP/EDM4hep data.
CLD and CLIC support both reconstructed track/cluster inputs and lower-level tracker/calorimeter hit inputs. The hit workflows have much larger events and are best treated as research workflows. See Current capabilities.