MPPT v0.2.0-rc.6#
Release candidate: this release focuses on observable HEFFSO results, reproducible managed runs, and reliable installation with the current supported toolchain.
Release date: 2026-09-03
Previous candidate: v0.2.0-rc.5
At a glance#
Area |
What changed |
|---|---|
HEFFSO |
Properties schema v3, solver diagnostics, determinant identity, and orbital analysis |
Reports |
A manifest-driven bundle with three CSVs, one Markdown file, and four PNGs |
Managed runs |
Strict artifact inventories, hashes, archive support, and restart validation |
PySCF integration |
A complete guide for exporting finished SCF, CASCI, and CASSCF references |
Build |
Miniforge-first setup, clean CMake/GNU builds, and one-prefix native/Python installation |
HEFFSO results you can inspect#
HEFFSO property analysis is enabled by default with
mppt.heffso.properties: true.<prefix>heffso_properties.h5uses the strictmppt_heffso_properties_v3schema.Every returned state carries persisted solver convergence evidence.
Determinant identities and vacuum-relative occupations remain traceable in the final HEFFSO row order.
Caspy exports state tables and line-strength or oscillator-strength spectra as a deterministic report bundle.
Runs you can reproduce#
run.jsoninventories required inputs, outputs, logs, software versions, and artifact hashes.Restart checks reject stale or incompatible outputs instead of silently reusing them.
pyscf2mppt archivepublishes a validated, no-replace snapshot of a completed managed run.Current schema-v3 reports and archived schema-v2 reports remain separate; mixed contracts are rejected.
PySCF references you can reuse#
Finished PySCF SCF, CASCI, and CASSCF objects can be exported without rebuilding their final one-electron operators.
External core Hamiltonians and explicit spin-orbit or property tensors have documented validation and tensor conventions.
Exported HDF5 references can enter the downstream workflow through either the Python API or
pyscf2mppt --from-hdf5.Full and selected-orbital routes, restart behavior, and common failures are covered in the hosted external-reference guide.
Build and selected-route fixes#
Clean builds support the current documented CMake and GNU compiler path.
The release environment updates Python, PySCF, HDF5, NumPy, and build-tool constraints together.
cmake --installplaces native libgrpp,libgrpppy, andpyscf2mpptin the active environment prefix.The selected route stores the vacuum energy once as
e_vide_fulland keeps its compact J/K and pair-packed ERI contract.The bundled CIPSIXX revision is pinned and verified in release snapshots.
Action required#
Replace the retired
effective_operatorssetting withmppt.heffso.properties.Use the canonical DIAGPT-to-HEFFSO handoff
<prefix>heffso.h5; the retired<prefix>heffso.1.h5fallback is not supported.Regenerate current HEFFSO reports instead of treating archived schema-v2 files as schema-v3 inputs.
Resume completed schema-v1 managed runs before archiving. Migration marks them
faileduntil a rerun records schema-v2 artifact and software evidence.Rebuild and install the source tree so native libgrpp,
libgrpppy, andpyscf2mpptuse the same prefix. Reinstall Caspy when managed HEFFSO reports are required.
Run these commands from the release source root in the active mppt
environment created through the installation guide:
cmake --build build --parallel 1
cmake --install build
python -m pip install --no-cache-dir ./caspy
Regenerate a current report while rerunning the enabled workflow stages:
pyscf2mppt run input.yaml --output-dir run --resume --force-workflow --heffso-report
Component versions#
Component |
Version |
|---|---|
MPPT release line |
|
|
|
Caspy |
|
HEFFSO properties schema |
|
Validation#
All 341
pyscf2mppt, 58 Caspy, and 131 release-snapshot tests pass.Wheels for
pyscf2mppt 0.5.0andcaspy 0.4.0build and install without editable mode.The release pipeline rebuilds the sanitized snapshot, runs
test-main, and builds the hosted documentation before publication.The public snapshot verifies its exact file set, executable modes, hashes, policy identity, and bundled CIPSIXX revision.
Known limitations#
Determinant selection remains the main scaling bottleneck for large selected-route calculations.
The selected route avoids full dense two-electron storage, but it does not remove the cost of determinant selection.
Publisher DOI sites can reject automated link checks with HTTP 403 even when the corresponding internal documentation links are valid.