Skip to content

SCPN Fusion Core v1.0.2 — Codex Hardening Sprint: 60 Test Files, AMR, FNO, PCE UQ

Choose a tag to compare

View all tags
@anulum anulum released this 14 Feb 14:46
· 1607 commits to main since this release
v1.0.2
ef09657

v1.0.2 — Codex Hardening Sprint: 60 Test Files, AMR, FNO, PCE UQ

Major hardening release driven by 130+ commits since v1.0.1. Test suite expanded from 4 to 60 test files. Significant new physics modules added.

License Metadata Fix

All package manifests now carry consistent AGPL-3.0-or-later license metadata across pyproject.toml, setup.py, and 4 Rust crates (fusion-math, fusion-core, fusion-engineering, fusion-ml).

Solvers and Numerics

Feature Description
Analytical Jacobian Inverse reconstruction with selectable JacobianMode
Chebyshev SOR Chebyshev-accelerated SOR with benchmark vs fixed SOR
Phase-space memory kernel Transport solver with memory-dependent diffusion
Patch-based AMR amr.rs hierarchy with refinement criterion on `

Physics and Transport

  • EPED-like pedestal model (fusion-core/src/pedestal.rs): pedestal width scaling Delta_ped ~ sqrt(beta_p,ped) * (rho_s / R), ELM trigger and crash application
  • Pedestal model integrated into transport solver H-mode path

Machine Learning and Uncertainty Quantification

Model Description
FNO turbulence Multi-layer training pipeline + Rust weight loading
Neural transport surrogate Rust implementation with .npz loading
Polynomial Chaos Expansion Multivariate Hermite basis, Latin Hypercube sampling, fit/predict, first-order Sobol indices

Engineering and Bindings

  • fusion-engineering modules: tritium systems, blanket design, magnet engineering, layout optimization
  • Expanded PyO3 bindings: inverse, transport, and plant APIs exposed to Python
  • Benchmark suites for inverse Jacobians and neural transport throughput

Validation

  • 42 validation configs — ITER, SPARC, JET, DIII-D, ITPA regression suites
  • 30+ input hardening commits — runtime guards across all physics and control modules
  • ITER/SPARC regression reference data and validation tests
  • GPU acceleration roadmap (docs/GPU_ACCELERATION_ROADMAP.md)

Test Suite

  • 60 test files (up from 4 in v1.0.1)
  • Property-based testing via Hypothesis across all subpackages

CI Fix

  • TestPyPI publish job marked continue-on-error so releases are no longer blocked by unconfigured TestPyPI trusted publisher

Install

pip install scpn-fusion==1.0.2

Full Changelog: v1.0.1...v1.0.2

Assets 2
Loading