pipeline

Class(es) used for analysis in PISA.

class graphnet.data.pipeline.InSQLitePipeline(module_dict, features, truth, device, retro_table_name, outdir, batch_size, n_workers, pipeline_name)[source]

Bases: ABC, Logger

Create a SQLite database for PISA analysis.

The database will contain truth and GNN predictions and, if available, RETRO reconstructions.

Initialise the pipeline.

Parameters:

  • module_dict (Dict) – A dictionary with GNN modules from GraphNet. E.g. {‘energy’: gnn_module_for_energy_regression}

  • features (List[str]) – List of input features for the GNN modules.

  • truth (List[str]) – List of truth for the GNN ModuleList.

  • device (device) – The device used for computation.

  • retro_table_name (str, default: 'retro') – Name of the retro table for.

  • outdir (Optional[str], default: None) – the directory in which the pipeline database will be stored.

  • batch_size (int, default: 100) – Batch size for inference.

  • n_workers (int, default: 10) – Number of workers used in dataloading.

  • pipeline_name (str, default: 'pipeline') – Name of the pipeline. If such a pipeline already exists, an error will be prompted to avoid overwriting.