pipeline-run-checklist

Pipeline Run Checklist

When to Use

• Starting a full or partial pipeline run for a configured galaxy instance
• Resuming after an error at a specific step
• Deciding whether step 7 (DrizzledInpainter) is needed
• Verifying outputs before proceeding to the next step

Pre-Flight Checks (before step 1)

• ./configure has been run — no [GALAXY] placeholders remain in notebooks
• All three conda environments exist: conda env list | grep -E 'stenv|astroba|dcr'
• CRDS environment variables are set: echo $CRDS_PATH
• JupyterLab is running with nb_conda_kernels so each notebook can select its own kernel

Step 1 — Image Download

Notebook: Images/ImageDownloader.ipynb Environment: stenv What it does: Queries MAST for HST observations of [GALAXY] and downloads raw FLT/FLC files.

Validation:

• FLT or FLC FITS files appear in Images/
• File count is non-zero and matches expected observations

Common failures:

• MAST query returns 0 results → check GALAXY_WILDCARD constant matches the MAST target name
• Download incomplete → re-run the download cell; MAST supports resuming

Step 2 — NED Info Download

Notebook: Data/NED/NED_InfoDownloader.ipynb Environment: stenv What it does: Downloads galaxy metadata (distance, morphology, redshift) from NED.

Validation:

• Output files appear in Data/NED/
• Distance and morphology values look reasonable for the target

Step 3 — GAIA Catalog Download

Notebook: Data/GAIA/GAIA_Downloader.ipynb Environment: stenv What it does: Downloads GAIA star catalog for astrometric alignment.

Validation:

• FITS or CSV catalog file appears in Data/GAIA/
• Source count is non-zero

Step 4 — Update CRDS References

Script: Images/update_crds.sh Environment: shell What it does: Downloads/updates HST reference files to ~/Data/CRDS.

bash Images/update_crds.sh

Validation:

• Script completes without errors
• $CRDS_PATH/references/hst/ directories (iref/, jref/, etc.) are populated

Common failures:

• CRDS_SERVER_URL not set → set in ~/.bashrc and source ~/.bashrc
• Disk space low → CRDS mirrors can be several GB; free space before running

Step 5 — Cosmic Ray Removal (DeepCR)

Notebook: Images/DeepCR-Remover.ipynb Environment: dcr What it does: Runs DeepCR neural network to identify and mask cosmic rays in each FLT/FLC.

Validation:

• Cosmic-ray-masked files (e.g., *_crc.fits) appear in Images/
• Mask fraction per image is plausible (< ~5% of pixels)

Common failures:

• CUDA/GPU not available → DeepCR falls back to CPU; much slower but works
• deepCR package not found → confirm the dcr environment is selected as the kernel

Step 6 — Image Reduction (Drizzle)

Notebook: Images/ImageReducer.ipynb Environment: stenv What it does: Runs AstroDrizzle to align, combine, and drizzle all exposures into final science images.

Validation:

• Drizzled science mosaic (*_drz_sci.fits) appears in Images/ProcessedImages/HST/
• Weight map (*_drz_wht.fits) is present alongside the science mosaic
• No large NaN/blank regions in the science image (open in DS9 or matplotlib to check)

Common failures:

• iref / jref variables not set → CRDS reference lookup fails; check step 4 was run
• Poor alignment → tweak ASTRODRIZZLE_PARAMS or check GAIA catalog coverage from step 3

Step 7 — NaN Inpainting (optional)

Notebook: Images/ProcessedImages/HST/PythonNotebooks/DrizzledInpainter.ipynb Environment: astroba

Decision — skip or run?

Open Images/ProcessedImages/HST/DS9/FOVs/ and inspect FOV region files in DS9:

• No blank/NaN regions within the science FOV → skip step 7
• Blank edges or chip gaps intersect the galaxy or science region → run step 7

Validation (if run):

• Inpainted mosaic is written to Images/ProcessedImages/HST/
• NaN regions are filled; pixel values at boundaries look smooth

Step 8 — Photometry Check

Notebook: Images/ProcessedImages/HST/PythonNotebooks/PhotometryChecker.ipynb Environment: stenv What it does: Compares source photometry from the drizzled image against catalog values as a quality check.

Validation:

• Photometry comparison plot is generated
• Residuals / zero-point offset are within acceptable range for the instrument/filter

Pipeline Complete

All 8 steps done. Final data products live in Images/ProcessedImages/HST/. Science analysis notebooks go in Science/.

Resuming After a Failure

1. Identify the failing step from the error message.
2. Fix the root cause (see common failures above, or ask the Pipeline Explorer agent).
3. Re-run only the failed step and all subsequent steps — earlier outputs are still valid unless you changed input files or constants.