Configuration

Analyses use YAML configuration layered on top of a compact canonical default file. This keeps ordinary analyses reproducible while allowing a variation to change only the relevant options.

CM2026 analyses

The observational defaults live in configs/cm2026/cm2026_defaults.yaml. The standard baseline overlay is configs/cm2026/cm2026_base.yaml and named variations live in configs/cm2026/variations.

Create arguments for the baseline or a named variation with:

from cup1d import Analysis, Args

baseline_args = Args.from_baseline()
variation_args = Args.from_variation("zmin")
analysis = Analysis(variation_args)

Each variation YAML is sparse: its values override the baseline defaults. Redshift nodes, covariance grids, fiducial model values, priors, and emulator training-set choices are derived internally where appropriate.

Other analyses

Configurations that are not CM2026 variations are grouped by analysis type:

configs/forecasts

Synthetic forecasts, named by the covariance data set and emulator.

configs/mocks/<simulation>

Simulation mocks, with separate files for each covariance and emulator combination.

configs/hires

Joint fits that add high-resolution P1D measurements to DESI DR1.

configs/tan2026

Paper-specific configurations for the TAN2026 analysis.

Load an observational configuration explicitly with:

args = Args.from_yaml("configs/tan2026/DESIY1_FFT3_dir_DLA_TAN.yaml")

Synthetic configurations must request the synthetic defaults:

forecast_args = Args.from_yaml(
    "configs/forecasts/DESIY1_QMLE3_CH24_mpgcen_gpr.yaml",
    synthetic=True,
)
mock_args = Args.from_yaml(
    "configs/mocks/mpg_central/DESIY1_QMLE3_CH24_mpgcen_gpr.yaml",
    synthetic=True,
)

Multiple data sets are selected with a YAML list. For example, configs/hires/hires.yaml combines DESI DR1 with Karacayli et al. (2022):

data_label:
  - DESIY1_QMLE3
  - Karacayli2022

scripts/data/data_sampler.py accepts either a CM2026 variation name or a path to a YAML configuration file. Relative YAML paths are interpreted from the repository root.