Fit a Parquet sample

Prerequisites

Each row should contain vector wavelength and flux columns plus uncertainty, redshift, RA, Dec, and an object identifier.

batch = qsospec.fit_batch(
    "spectra.parquet",
    "runs/sample",
    row_indices=range(20),
    n_workers="auto",
)

print(batch.n_completed, batch.n_failed)

Expected outputs

The run bundle contains canonical resumable Parquet datasets. Finalization validates them without producing duplicate compact tables. Repeating the command with the same configuration resumes the run.

Common failures

• Dust-map preflight fails: configure the external data directory.

• Individual missing coordinates: inspect the failures table.

• Resume rejects configuration: start a new run directory.

Next: Batch fitting and Run-bundle reference.