inversepde 0.1.0

Production package for inverse PDE surrogate modeling

inversepde

A production-ready package for inverse PDE surrogate modeling.

Installation

pip install -e .

Quickstart

import torch
from inversepde import DiffusionFamily, InverseSolver

family = DiffusionFamily()
solver = InverseSolver(family=family)

coords = torch.rand(1, 120, 2)
values = torch.rand(1, 120)
pred = solver.predict(coords, values)
print(pred.shape)

API

• PDEFamily: abstract base class for PDE families.
• DiffusionFamily: diffusion data-generation wrapper.
• ReactionDiffusionFamily: reaction-diffusion wrapper.
• InverseSolver: high-level training and inference API.
• load_pretrained and list_available_pretrained: checkpoint helpers.
• plot_prediction: visualization utility.

Development

pip install -e .[dev]
pytest -q

Reproducibility

• Python: 3.14.x (validated in this workspace)
• Install: pip install -e .[dev]
• Tests: python -m pytest -q
• Lock file: requirements.lock.txt

Release Checks

ruff check .
mypy inversepde/inversepde
python -m build
python -m twine check dist/*