Data and notebook downloads
The Python distribution contains only the pymhm library and its typing and
release metadata. Notebook sources, example helpers, case configurations,
datasets, recorded fields and figures are separate repository and website files.
No case collection or resource registry is installed with the library.
The download catalogue lists individual notebooks, their companion ZIPs, available records and figures with byte sizes, SHA-256 checksums, source attribution and scope. A notebook's first cell identifies its companion URL and checksum. The companion contains its importable Python helpers, acquisition recipes, small configurations and selected input catalogue. It contains no datasets, computed fields or images. Acquiring and extracting it verifies its bytes without executing its code; the subsequent notebook cells import the declared helpers.
The notebook uses a writable workspace for acquired inputs and generated outputs. Selected data downloads are verified against their SHA-256 identities; valid cached files can be reused. Install the native backend identified by the notebook when required; see installation. Opening a notebook or displaying a recorded figure does not recompute its fields.
Primary datasets
| Dataset | Public source and identity | Acquisition scope |
|---|---|---|
| SPE10 Model 2 | OPM/opm-data, pinned revision | Full 60×220×85 permeability/porosity, or the separately downloaded layers 1, 36, 85 |
| Marmousi II | Madagascar primary SEG-Y files | Two verified files, about 310 MB total; the declared 2048×512 material crop retains its coordinates and units |
| HPC4e | labmec/MHM, pinned Data_13_Set | Three verified material arrays with the article's physical bounds and axis conventions |
The separately downloadable dataset identity records contain the source revisions, checksums, units and array conventions. The corresponding public acquisition helpers reuse valid caches and reject mismatched downloads. They keep primary dataset files outside the Python installation.
The compact SPE10 downloads are layer 1, layer 36 and layer 85. Their catalogue entries link the accompanying metadata and dataset notices.
SPE10 is attributed to Mike Christie and Martin Blunt. The two pinned property include-file headers state that their data are public domain. OPM's repository also supplies ODbL 1.0 database/deck and DbCL 1.0 content notices unless otherwise specified. Those notices accompany the downloadable slices and are separate from PyMHM's LGPL license. The porosity convention, including OPM's recorded replacement of most zeros by 1e-7, is preserved. Downloading or redistributing other datasets requires observing the terms at their stated primary source; the package license does not grant additional third-party rights.
Current studies and historical comparisons
Current computed fields use the notebook's public acquisition recipe. After acquiring its companion, use its runner from the companion workspace. Inspect its discretization, numerical scope and resource requirements before a full study:
cd /path/to/extracted-companion # the notebook's ROOT directory
python -m scripts.run_notebooks /path/to/72_marmousi.ipynb --study --plan
The command performs no acquisition in planning mode. Complete studies retain their stated computational cost; a small current control is labeled separately from a matched article reproduction.
Original external field archives and article image crops are optional historical inputs. The catalogue records unavailable payloads and links their primary sources. Article crops without verified redistribution permission are not exported as downloads. Supply the exact attributed payload when requesting historical replay; a fresh PyMHM field cannot replace an original external comparison.