valska_hera_beam.external_tools.bayeseor.sweep

Sweep orchestration for BayesEoR perturbation studies.

Functions

`archive_jobs_json`(run_dir)	Archive run_dir/jobs.json -> run_dir/jobs_<UTCSTAMP>.json
`run_fwhm_sweep`(*, template_yaml, install, ...)	Orchestrate a sweep over multiple perturbation values.
`sweep_point_dir`(results_root, beam_model, ...)	Per-point run directory for sweeps.
`sweep_root`(results_root, beam_model, ...)	Central sweep output location.
`write_sweep_manifest`(*, results_root, ...[, ...])	Write sweep_manifest.json under the sweep directory.

Classes

`SweepPoint`(perturb_parameter, perturb_frac, ...)	Metadata for a single sweep point.
`SweepResult`(results_root, beam_model, ...)	Summary of a completed sweep run (prepare + optional submit).

class valska_hera_beam.external_tools.bayeseor.sweep.SweepPoint(perturb_parameter: Literal['fwhm_deg', 'antenna_diameter'], perturb_frac: float, run_label: str, run_dir: Path, manifest_json: Path)

Metadata for a single sweep point.

property fwhm_perturb_frac: float | None: Backward-compatible alias for legacy FWHM-only callers.

manifest_json: Path

perturb_frac: float

perturb_parameter: Literal['fwhm_deg', 'antenna_diameter']

run_dir: Path

run_label: str

class valska_hera_beam.external_tools.bayeseor.sweep.SweepResult(results_root: Path, beam_model: str, sky_model: str, variant: str, run_id: str, perturb_parameter: Literal['fwhm_deg', 'antenna_diameter'], data_path: Path, created_utc: str, sweep_dir: Path, sweep_manifest_json: Path, template_yaml: Path, points: list[SweepPoint], submit_results: list[dict[str, Any]])

Summary of a completed sweep run (prepare + optional submit).

beam_model: str

created_utc: str

data_path: Path

perturb_parameter: Literal['fwhm_deg', 'antenna_diameter']

points: list[SweepPoint]

results_root: Path

run_id: str

sky_model: str

submit_results: list[dict[str, Any]]

sweep_dir: Path

sweep_manifest_json: Path

template_yaml: Path

variant: str

valska_hera_beam.external_tools.bayeseor.sweep._default_fwhm_fracs() → list[float]

valska_hera_beam.external_tools.bayeseor.sweep._format_run_label(*, perturb_parameter: Literal['fwhm_deg', 'antenna_diameter'], frac: float) → str

valska_hera_beam.external_tools.bayeseor.sweep._format_run_label_from_antenna_diameter_frac(frac: float) → str: Must match cli_prepare.py formatting so directory layout is predictable.

valska_hera_beam.external_tools.bayeseor.sweep._format_run_label_from_fwhm_frac(frac: float) → str: Must match cli_prepare.py formatting so directory layout is predictable.

valska_hera_beam.external_tools.bayeseor.sweep._jobs_path(run_dir: Path) → Path

valska_hera_beam.external_tools.bayeseor.sweep._utc_now_compact() → str

valska_hera_beam.external_tools.bayeseor.sweep._utc_now_iso() → str

valska_hera_beam.external_tools.bayeseor.sweep.archive_jobs_json(run_dir: Path) → Path | None

Archive run_dir/jobs.json -> run_dir/jobs_<UTCSTAMP>.json

Returns the archived path if an archive was created, else None.

valska_hera_beam.external_tools.bayeseor.sweep.run_fwhm_sweep(*, template_yaml: Path, install: BayesEoRInstall, runner: CondaRunner | ContainerRunner, results_root: Path, beam_model: str, sky_model: str, variant: str, run_id: str, data_path: Path, slurm_cpu: dict[str, object] | None = None, slurm_gpu: dict[str, object] | None = None, overrides: dict[str, str] | None = None, perturb_parameter: Literal['fwhm_deg', 'antenna_diameter'] = 'fwhm_deg', perturb_fracs: Iterable[float] | None = None, fwhm_fracs: Iterable[float] | None = None, unique: bool = False, submit: Literal['none', 'cpu', 'gpu', 'all'] = 'none', hypothesis: Literal['signal_fit', 'no_signal', 'both'] = 'both', depend_afterok: str | None = None, sbatch_exe: str = 'sbatch', submit_dry_run: bool = False, force: bool = False, resubmit: bool = False, record: Literal['jobs.json', 'manifest'] = 'jobs.json', dry_run: bool = False) → SweepResult

Orchestrate a sweep over multiple perturbation values.

Parameters

template_yaml: BayesEoR template YAML to render.
install: BayesEoR installation metadata.
runner: Runner configuration (conda or container).
results_root: Root directory for results.
beam_model, sky_model, variant, run_id: Taxonomy fields and sweep identifier.
data_path: Input data path.
slurm_cpu, slurm_gpu: SLURM settings for CPU and GPU stages.
overrides: Template overrides applied to each run.
perturb_parameter: Which config key to perturb: fwhm_deg or antenna_diameter.
perturb_fracs: Iterable of perturbation fractions to apply to perturb_parameter.
fwhm_fracs: Backward-compatible alias for perturb_fracs when using perturb_parameter='fwhm_deg'.
unique: If True, append a UTC timestamp to each run directory.
submit: Stage(s) to submit after prepare (“none”, “cpu”, “gpu”, “all”).
hypothesis: Which GPU hypothesis to submit (“signal_fit”, “no_signal”, “both”).
depend_afterok: Optional sbatch job id to depend on for GPU submissions.
sbatch_exe: sbatch executable to invoke.
submit_dry_run: If True, do not submit; return the commands that would run.
force: If True, allow resubmission even if jobs.json indicates prior submissions.
resubmit: If True, archive jobs.json before submitting.
record: Where to record submission metadata. Currently only “jobs.json” is supported.
dry_run: If True, do not prepare or submit; only compute intended run paths.

Returns

SweepResult: Summary of the sweep, including prepared points and submit results.

Sweep directory layout

For sweeps we intentionally colocate all points so they can be archived/removed as a unit:

<results_root>/bayeseor/<beam_model>/<sky_model>/_sweeps/<run_id>/<variant>/<run_label>/

Behaviour

Prepares one run_dir per perturbation frac (stable unless unique=True).
Optionally submits per run_dir via submit_bayeseor_run with stage cpu/gpu/all.
Writes/updates sweep_manifest.json under:
<results_root>/bayeseor/<beam_model>/<sky_model>/_sweeps/<run_id>/

Resubmission behaviour

If resubmit=True and submit_dry_run=False, for each point we archive jobs.json (if present) to jobs_<timestamp>.json before submitting.
This enables “GPU-only resubmit across sweep” safely:
submit=”gpu”, resubmit=True

valska_hera_beam.external_tools.bayeseor.sweep.sweep_point_dir(results_root: Path, beam_model: str, sky_model: str, run_id: str, *, variant: str, run_label: str) → Path

Per-point run directory for sweeps.

Layout:: <results_root>/bayeseor/<beam_model>/<sky_model>/_sweeps/<run_id>/<variant>/<run_label>/

valska_hera_beam.external_tools.bayeseor.sweep.sweep_root(results_root: Path, beam_model: str, sky_model: str, run_id: str) → Path

Central sweep output location.

Layout:: <results_root>/bayeseor/<beam_model>/<sky_model>/_sweeps/<run_id>/

valska_hera_beam.external_tools.bayeseor.sweep.write_sweep_manifest(*, results_root: Path, beam_model: str, sky_model: str, variant: str, run_id: str, perturb_parameter: Literal['fwhm_deg', 'antenna_diameter'], data_path: Path, template_yaml: Path, sweep_dir: Path, points: list[SweepPoint], submit_results: list[dict[str, Any]] | None = None) → Path

Write sweep_manifest.json under the sweep directory.

Parameters

results_root, beam_model, sky_model, variant, run_id: Identify the sweep location and taxonomy.
data_path: Data path used for the sweep.
template_yaml: Template used for the sweep points.
sweep_dir: Target sweep directory.
points: Prepared sweep points.
submit_results: Optional submission results to include.

Returns

Path: Path to the written sweep_manifest.json.