TorchLean API

Docs HomeGuideExamplesGraphs

NN.API.Models.Fno1d

Fourier Neural Operator Model Helpers (API) #

Reusable configuration for the real-valued 1D FNO path used by the runnable Burgers example.

The public example is intentionally about data loading, training, evaluation, and plotting. The model constructor lives here so other examples can reuse the same typed FNO shape contract without copying the wiring.

source
structure NN.API.nn.models.Fno1dConfig :
Type

Configuration for a real-valued 1D FNO on a fixed grid.

Instances For
    source
    def NN.API.nn.models.instReprFno1dConfig.repr :
    Fno1dConfig → ℕ → Std.Format
    Instances For
      source
      @[implicit_reducible]
      instance NN.API.nn.models.instReprFno1dConfig :
      Repr Fno1dConfig
      source
      @[reducible, inline]
      abbrev NN.API.nn.models.fno1dInShape (cfg : Fno1dConfig) :
      Shape

      Input shape for a scalar field sampled on cfg.grid points.

      Instances For
        source
        @[reducible, inline]
        abbrev NN.API.nn.models.fno1dOutShape (cfg : Fno1dConfig) :
        Shape

        Output shape for a scalar field sampled on cfg.grid points.

        Instances For
          source
          def NN.API.nn.models.fno1dReal (cfg : Fno1dConfig) (hModesFit : 2 * cfg.modes ≤ cfg.grid := by decide) :
          M (Sequential (fno1dInShape cfg) (fno1dOutShape cfg))

          Real-valued FNO1D constructor.

          The backend may later choose a dense-DFT or cuFFT-backed spectral implementation, but the model shape and parameter contract are the same from the API side.

          Instances For