Python API

The pages below are generated from the actual Python sources, including signatures, abstract methods, dataclass fields, and existing docstrings. AutoAPI parses the files without importing the application. Installing CUDA, PyTorch, or NNPD itself is not required to build these pages.

Start with the extension surface

Module	Main responsibilities
nnpd.core.api	Experiment, Analysis, Context, Product, Metric, Plot, load_experiment, load_analysis.
nnpd.core.settings	Choice, Variant, expand_sweep, load_settings and selectors.
nnpd.core.runner	plan, execute, restore, analyze_run, expand_jobs, verify_store.
nnpd.core.storage	Store, Artifact, Writer, manifests and atomic writes.
nnpd.core.runtime	Runtime and process/device coordination.
nnpd.results	runs, export_csv, comparison_plot.

The usual convenience imports are:

from nnpd import (
    Analysis, Choice, Experiment, Metric, Plot, Product,
    execute, expand_sweep, load_analysis, load_experiment, load_settings, plan, restore,
)

Context, Artifact, Writer, and Runtime are imported from their respective nnpd.core modules rather than the top-level package.

Source API

• nnpd
  • Submodules
    • nnpd.cli
    • nnpd.core
    • nnpd.nre
    • nnpd.results
• gaussian
  • Submodules
    • gaussian.analysis
    • gaussian.diagnostics
    • gaussian.evaluation
    • gaussian.experiment
    • gaussian.extensions
    • gaussian.metrics
    • gaussian.plots
    • gaussian.problem
    • gaussian.sampling

Use the extension guide for full hook contracts and the analysis guide for the Gaussian product/metric mapping. Private implementation helpers are not the main extension interface.