midas-pipeline
End-to-end MIDAS HEDM orchestrator. FF is the single-scan degeneracy of PF. One package, one CLI, two scan modes.
Status
0.1.0 — end-to-end PF and FF paths live. The scanning indexer matches the C IndexerScanningOMP reference on its 1-voxel C-parity gate (seed identity, solution counts, voxel-center positions exact; orientation matrices within mrad-scale, the refiner closes the gap downstream). Real-data validation: datasetC CP-Ti consolidation_pf reproduces the legacy pf_MIDAS.py grain count (770 == 770, all common). Park22 P5c parity gate now runs in ~6.7s vs the original 790s after the scanning-indexer position-grid fix.
Stages call in-process Python kernels via midas-index / midas-fit-grain / midas-transforms / midas-stress. FF mode shells out to python -m midas_index and python -m midas_fit_grain (same kernels, subprocess for the FF parity-preserving pattern). No CUDA C; GPU is torch-only.
Install
pip install -e packages/midas_pipeline
Notebooks
Worked-example Jupyter notebooks live in notebooks/. They are not shipped with pip install — get them by cloning the MIDAS repository.
CLI
midas-pipeline run --scan-mode {ff,pf,auto} --params Parameters.txt --result rundir/
midas-pipeline status rundir/
midas-pipeline resume rundir/ --from <stage>
midas-pipeline reprocess rundir/
midas-pipeline inspect rundir/LayerNr_1/
midas-pipeline simulate --out simdir/ --params Parameters.txt
midas-pipeline seed --params ... --output UniqueOrientations.csv
When --scan-mode is omitted (default auto), the CLI sniffs the parameter file: nScans > 1 or presence of BeamSize / scanning keys → pf, otherwise ff. For PF mode, --n-scans, --scan-step, --beam-size, and --scan-pos-tol default to values in the params file (CLI flags override).
Indexer backend
midas-pipeline run --indexer-backend {python,c-omp} ...
python (default) — in-process numba/torch indexer. Portable (CPU/CUDA/MPS), differentiable, slower on large PF datasets.
c-omp — bundled unified C binary (midas_indexer) from midas-index. Requires midas-index installed with a working OpenMP toolchain (macOS: brew install libomp). ~290× faster than the Python path on real PF datasets (per the datasetC CP-Ti benchmark in packages/midas_index/dev/). Output is bit-identical to the Python path on the PF parity gate.
Coexistence with midas-ff-pipeline
The legacy midas-ff-pipeline console-script is preserved as an independent FF orchestrator (its own kernels, its own CLI). It is not deprecated by midas-pipeline run --scan-mode ff — both paths invoke the same midas-index / midas-fit-grain kernels under the hood, and both stay green on the FF parity gate. Pick whichever is more convenient for your workflow.
Architecture
- One orchestrator with a mode-dependent
STAGE_ORDER. - Shared kernel packages (
midas-index,midas-fit-grain,midas-transforms, etc.) extended in place; FF behavior preserved by parity gates. - PF-only modules live inside
midas_pipeline(find_grains/,sinogen,recon/,fuse,potts,em_refine,seeding/). - Differentiability + multi-device mandatory on every new compute path (CPU / CUDA / MPS via torch).
Constraints
- No CUDA C; GPU support is torch-only.
- No deletions of legacy code in this effort.
midas-process-grainsis FF-only; PF consolidation is fresh pure-Python.utils/calcMiso.pyis not imported by this package; all orientation math comes frommidas-stress.TOMO/midas_tomo_python.pyis imported in place, not relocated.
1 maintainer
Metadata
1 maintainer
Release files for midas-pipeline 0.18.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| midas_pipeline-0.18.0.tar.gz | 237.3 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| midas_pipeline-0.18.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 509.3 kB
Release files / midas_pipeline-0.18.0.tar.gz
| Download URL | midas_pipeline-0.18.0.tar.gz |
|---|---|
| Size | 237.3 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
efa2f836329ea1af46b7f4e6f301de9b3910f9dd071ac2d2350a521851dcb32e
|
|
BLAKE2b-256 checksum How to use checksums |
d48c6c24a539017682de177872df10bedded078a0be7941ef1f07b964fc522a2
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 6, 2026.
Transparency logRelease files / midas_pipeline-0.18.0-py3-none-any.whl
| Download URL | midas_pipeline-0.18.0-py3-none-any.whl |
|---|---|
| Size | 272.0 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
e8cf2d1fe8acaecbb83258c795169afc20dd7d73edbcd6f86fb8f1a15bb6cafe
|
|
BLAKE2b-256 checksum How to use checksums |
90a1866438a63fcab0e1b1eb6b3582a3e8e3d770bd74927163cf758abecbc1c7
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 6, 2026.
Transparency log
