MNISTDataset

@Serializable

data class MNISTDataset(val images: List<MNISTImage>, executionContext: ExecutionContext = DefaultDataExecutionContext()) : Dataset<MNISTImage, Float> (source)

MNIST dataset implementation using Dataset/DataBatch API.

Provides batching as tensors FP32 with shapes: x: batch, 1, 28, 28 (normalized 0..1) y: batch (labels as floats)

Constructors

constructor(images: List<MNISTImage>, executionContext: ExecutionContext = DefaultDataExecutionContext())

open override val xSize: Int

open override fun getX(idx: Int): MNISTImage

open override fun getY(idx: Int): Float

fun subset(fromIndex: Int, toIndex: Int): MNISTDataset

Returns a subset of the dataset.