Build a partial-observation benchmark

Choose the physical system and observation view, then generate reproducible dataset and quality-control code.

7 PDE losses Local + Linux SeaWulf Slurm Deep-linkable choices

Choose a benchmark slice, inspect its size, and copy the exact dataset-generation code. Ground truth is generated once; observation masks are deterministic views applied afterward.

Scientific gate: every run produces PDE-specific quality diagnostics. Bundled reference solvers are not paper-grade ground truth until the documented numerical-validation and release gates pass.

1. Choose the benchmark

Advanced paths and resources
Optional: future model campaign planning

Future model campaign planning

Primary matched-mask comparison: every normal learned baseline gets an independent checkpoint for each PDE and each of the nine observation masks. The model trained on random 3% is reused only for the separate mask-transfer/OOD table; it never replaces matched-mask training in the primary IID table.

Open the complete observation-training protocol

Execution boundary: this planner counts the complete paper matrix, including methods that are not integrated in PDE-OBS. Its campaign manifest is planning-only. It never invents commands for Gappy POD, DeepONet, PINN/PINO, Transolver/GNOT, DiffusionPDE, or FunDPS.

Campaign preset

Planning status
    Method rowFit policySeedsObservationsPreparationEvaluationsBuilder status

    2. Copy the generated code

    The dataset YAML is the executable source of truth. Use the environment tab that matches your machine.

    Loading benchmark contract...

    3. Quality report for every PDE

    Every generated sample receives finite-value, geometry, initial/boundary, and family-specific physics diagnostics. Dataset aggregation reports each selected PDE separately; the complete benchmark reports all seven losses together.

    PDEReported normalized lossInterpretation

    Always produced

      Gate levels

      1. Report: report scientific losses without applying an unfrozen PDE threshold; malformed array/geometry contracts are still quarantined.
      2. Strict: reject structural/BC/IC failures, log them to *.quality-failures.jsonl, and apply a PDE limit only when calibrated.
      3. Publication candidate: expert-only and intentionally blocked by this Builder until validated solver evidence and a per-stratum threshold table exist. It does not by itself set publication_ready.
      Important interpretation: saved-frame residuals are not the same as integrator replay error. Helmholtz nominal residual and legacy regularized transfer defect stay separate. Bounded Navier-Stokes uses a versioned vorticity/streamfunction residual plus divergence and boundary diagnostics.