YAML Configuration

AssimilationSuite.from_yaml() is the recommended entry point for production workflows. A single YAML file describes all benchmarks, applications, and the covariance library.

Minimal example

benchmarks:
  - title: HMI-001
    m: 1.0000
    dm: 0.0005
    sens0_path: data/hmi001_sens0.m
    results_path: data/hmi001_res.m

applications:
  - title: my-reactor
    sens0_path: data/reactor_sens0.m
    results_path: data/reactor_res.m

covariances:
  file_path: data/covariances.h5

Load with:

from andalus import AssimilationSuite

suite = AssimilationSuite.from_yaml("assimilation.yaml")
posterior = suite.glls()

---

benchmarks section

Each entry in the benchmarks list describes one integral experiment. Two source formats are supported: Serpent files and HDF5.

From Serpent files

Required keys:

Key	Type	Description
title	string	Unique identifier
m	float	Measured value (e.g. \(k_{\text{eff}}\))
dm	float	Measurement uncertainty (\(1\sigma\))
sens0_path	string	Path to the Serpent _sens0.m file
results_path	string	Path to the Serpent _res.m file

Optional keys:

Key	Type	Default	Description
kind	string	"keff"	Observable type
zailist	list[int]	all	Subset of nuclides (ZAI numbers) to load
pertlist	list[int]	all	Subset of reactions (MT numbers) to load
materials	list	all	Subset of Serpent materials
flux_det_name	string	—	Detector name for the flux spectrum
flux_det_path	string	—	Path to the Serpent _det0.m file

Example:

benchmarks:
  - title: HMI-001
    m: 1.0000
    dm: 0.0005
    sens0_path: data/hmi001_sens0.m
    results_path: data/hmi001_res.m
    kind: keff
    zailist: [922350, 922380]
    pertlist: [18, 102]
    flux_det_name: FLUX
    flux_det_path: data/hmi001_det0.m

From HDF5

Use hdf5_path instead of sens0_path / results_path. You may load a single entry, a named subset, or all entries from the file:

benchmarks:
  # Single benchmark by title
  - title: HMI-001
    hdf5_path: data/benchmarks.h5

  # Named subset
  - titles: [HMI-001, HMI-002, HMI-003]
    hdf5_path: data/benchmarks.h5

  # All benchmarks in the file (omit titles / set to null)
  - hdf5_path: data/benchmarks.h5

Optional key:

Key	Type	Default	Description
kind	string	"keff"	HDF5 group key

---

applications section

Identical structure to benchmarks except that m and dm are optional.

From Serpent files

applications:
  - title: my-reactor
    sens0_path: data/reactor_sens0.m
    results_path: data/reactor_res.m
    # Optional: provide a measurement if available
    m: 1.0023
    dm: 0.0010

From HDF5

applications:
  - title: my-reactor
    hdf5_path: data/applications.h5

---

covariances section

The covariances section points to a single HDF5 file that contains pre-processed nuclear data covariance matrices (produced via Covariance.to_hdf5()).

covariances:
  file_path: data/covariances.h5

AssimilationSuite.from_yaml() automatically determines which nuclides (ZAIs) to load from the union of all ZAIs present in the benchmark and application sensitivities. The reactions loaded are currently fixed to MT 2, 4, 18, 102, 456, and 35018.

---

Full annotated example

# benchmarks: integral experiments used to constrain nuclear data
benchmarks:
  # --- loaded from Serpent output ---
  - title: HMI-001
    m: 1.0000
    dm: 0.0005
    sens0_path: serpent/hmi001_sens0.m
    results_path: serpent/hmi001_res.m
    zailist: [922350, 922380, 942390]

  - title: PU-MET-FAST-001
    m: 1.0000
    dm: 0.0003
    sens0_path: serpent/pmf001_sens0.m
    results_path: serpent/pmf001_res.m
    flux_det_name: FLUX
    flux_det_path: serpent/pmf001_det0.m

  # --- loaded from HDF5 library ---
  - titles: [LEU-COMP-THERM-001, LEU-COMP-THERM-002]
    hdf5_path: db/icsbep_benchmarks.h5

# applications: systems to predict
applications:
  - title: reactor-core
    sens0_path: serpent/core_sens0.m
    results_path: serpent/core_res.m

  - title: waste-storage
    sens0_path: serpent/waste_sens0.m
    results_path: serpent/waste_res.m

# covariances: nuclear data covariance library
covariances:
  file_path: db/jeff40_covariances.h5