TorchLean API

Docs HomeGuideExamplesGraphs

NN.API.Samples.Bands

Small Band-Image Dataset (4×4) #

Several TorchLean tutorials use a compact 4×4 image classification task:

This module packages that synthetic dataset in one place so examples can stay focused on models and training rather than data-loading infrastructure.

Class Spec #

source
def NN.API.Samples.Bands.classes :
List Image2D.BandClass

Canonical label set for the band dataset: vertical ↦ 0, horizontal ↦ 1.

Instances For

    Typed Tensors (Tensor-First) #

    source
    @[reducible, inline]
    abbrev NN.API.Samples.Bands.shape :
    Spec.Shape

    Canonical image shape for the band dataset (single-channel 4×4).

    Instances For
      source
      def NN.API.Samples.Bands.trainCHWFloat :
      List (Spec.Tensor Float shape × ℕ)

      Training set samples: a small list of (x, label) pairs.

      Instances For
        source
        def NN.API.Samples.Bands.probesCHWFloat :
        List (String × Spec.Tensor Float shape × ℕ)

        Probe set for reporting: (name, x, expectedLabel) triples.

        Instances For
          source
          def NN.API.Samples.Bands.trainCHW {α : Type} [Context α] (cast : Float → α) :
          List (Spec.Tensor α shape × ℕ)

          Cast trainCHWFloat into an arbitrary scalar backend α.

          Instances For
            source
            def NN.API.Samples.Bands.probesCHW {α : Type} [Context α] (cast : Float → α) (probes : List (String × Spec.Tensor Float shape × ℕ) := probesCHWFloat) :
            List (String × Spec.Tensor α shape × ℕ)

            Cast probesCHWFloat into an arbitrary scalar backend α.

            Instances For