Configuration Reference

Parnassus uses YAML configuration files. This page documents the main user-facing settings.

Dataset

Controls input data loading.

dataset:
  file_path: "input.hepmc"
  num_events: 1000
  entry_start: 0

Field	Type	Default	Description
`file_path`	string	required	Path to the input file (`.hepmc`, `.root`, or `.cmnd`)
`num_events`	integer	`1`	Number of events to process
`entry_start`	integer	`0`	Starting entry index in the input file. Useful for skipping the first N events or for processing a file in parallel chunks by running multiple jobs with non-overlapping `entry_start` + `num_events` ranges.

Generator

Selects and configures the simulation backend.

Common fields

Field	Type	Default	Description
`type`	string	required	Generator type: `"neural"` or `"parametric"`
`name`	string	required	Generator/model name from the registry
`batch_size`	integer	`2000`	Number of events per processing batch
`device`	string	`"cpu"`	Computation device: `"cpu"`, `"cuda"`, or `"mps"`

Neural-specific fields

Field	Type	Default	Description
`num_steps`	integer	`50`	Number of ODE integration steps
`max_particles`	integer	from model	Maximum particles per event. Set from the model metadata; this is not a YAML setting for neural generators. Particles beyond this limit are dropped during preprocessing.
`truth_pt_cut`	float	from model	Minimum truth-particle pT (GeV) for the neural selection cut. Set from the model metadata (`0.25` for `cms_2011_flow_v00`, `0.5` for `aleph_flow_v00`); not a YAML setting.

Parametric-specific fields

Field	Type	Default	Description
`seed`	integer	none	Random seed for reproducibility
`debug`	boolean	`false`	Write intermediate detector-stage collections in addition to the standard parametric output
`pileup`	object	none	Optional pile-up merging configuration (see below)

Pile-up merging (`pileup`)

When the pileup section is present under generator, the parametric pipeline merges pile-up particles from a pre-generated MinBias file with the hard-scatter input before detector simulation. This matches the behaviour of the C++ Delphes PileUpMerger module.

generator:
  type: "parametric"
  name: "cms"
  seed: 42
  pileup:
    file_path: "MinBias.pileup"
    mean_pileup: 50

Field	Type	Default	Description
`file_path`	string	required	Path to a Delphes `.pileup` binary (XDR) file
`mean_pileup`	float	required	Average number of pile-up interactions per bunch crossing (Poisson-sampled)
`max_z_spread`	float	`0.25`	Vertex z truncation bound in meters
`max_t_spread`	float	`800e-12`	Vertex t truncation bound in seconds
`sigma_z`	float	`0.053`	Gaussian sigma for vertex z-smearing in meters
`sigma_t`	float	`160e-12`	Gaussian sigma for vertex t-smearing in seconds
`smear_hs_vertex`	boolean	`true`	Also smear the hard-scatter primary vertex position

Pipelines

Map of pipeline name to pipeline configuration. See Pipelines for full details.

pipelines:
  <PipelineName>:
    type: "cluster" | "isolation"
    batch_size: 2000
    num_processes: 1
    # ... type-specific fields

Common pipeline fields

These execution controls are supported by both cluster and isolation pipelines.

Field	Type	Default	Description
`batch_size`	integer	`2000`	Number of events per postprocessing batch
`num_processes`	integer	`1`	Number of worker processes. Use `1` for synchronous execution; values above `1` use multiprocessing.

Cluster fields

Field	Type	Default	Description
`type`	string	--	`"cluster"`
`collection`	string	`"pflow"`	`"truth"` or `"pflow"`
`algorithm`	string	`"antikt"`	`"antikt"`, `"cambridge"`, `"genkt"`, or `"ee-genkt"`
`dr`	float	`0.5`	Jet radius
`algorithm_param`	float	none	Extra p parameter for `"genkt"` / `"ee-genkt"` (required for those algorithms)
`pt_min`	float	`0`	Minimum jet pT (GeV)
`nconst_min`	integer	`2`	Minimum constituents

Isolation fields

Field	Type	Default	Description
`type`	string	--	`"isolation"`
`collection`	string	`"electrons"`	`"electrons"`, `"muons"`, or `"all"`
`dr`	float	`0.4`	Cone radius

Output

output:
  file_path: "output.root"
  format: default

Field	Type	Default	Description
`file_path`	string	required	Path to the output ROOT file
`format`	string	`"default"`	Reserved for future output-format selection. Currently only ROOT output through the default writer is supported.