Python-first StaMPS runtime

Process and inspect pySTAMPS workflows without losing the engineering details.

pySTAMPS is a static-dataset processing toolkit for StaMPS-style workflows. It helps you inspect a dataset, run selected processing stages, and understand each stage-level output before moving to output comparison.

Read the pipeline guide See the command flow Read the function reference

Inspect

Run this first and confirm that dataset layout and stage metadata are what you expect.

uv run pystamps status --dataset /path/to/dataset

Run

Process only what you need, and optionally dry-run before expensive execution.

uv run pystamps run --dataset /path/to/dataset --start-step 1 --end-step 4 --dry-run

uv run pystamps run --dataset /path/to/dataset --start-step 6 --end-step 8

Optimize

Inspect native Rust/CPU and optional CUDA kernel coverage before tuning a run config.

uv run pystamps describe-backends

New to interferometry?

Read Pipeline Guide for the combined science and program tutorial, or Getting Started for the shorter beginner path.

Need the shortest path?

Go to Quick Start for the minimum set of commands to inspect a dataset copy and dry-run the pipeline.

Need a formal map?

Use Function Reference to see which modules own configuration, orchestration, dataset discovery, and runtime execution.

What pySTAMPS actually does

It works on a dataset folder

pySTAMPS expects a StaMPS-style dataset root with patch directories such as PATCH_1 and stage artifacts inside those patches or at the merged dataset level.

It orchestrates stages 1 through 8

The pipeline can run only the range you need. This matters because later stages often depend on outputs created by earlier stages, and some stages are expensive.

It is artifact driven

Existing stage artifacts are inspected and can be skipped. That lets you resume or study a partial run without throwing away all intermediate results.

Start here if you want to run the package

Install the Python environment and external tools from Installation.
Read Pipeline Guide if you want the science, stage outputs, kernel switching, verification, and benchmarking explained together.
Follow Quick Start on a copy of a dataset, not your only original dataset.
Use Usage once you need stage selection and backend tuning.
Use Verification when you need output comparison and parity gate checks.