Type Alias MaxOp<Input, Axes, KeepDims, OutputDType>

MaxOp: ValidateReduction<Input["__shape"], Axes> extends true
    ? StorageTransformation<
        "max",
        TensorStorage<
            OutputDType,
            ComputeReductionShape<Input["__shape"], Axes, KeepDims>,
            ComputeStrides<ComputeReductionShape<Input["__shape"], Axes, KeepDims>>,
            ReductionOpLayout<Input["__layout"]>,
        >,
        readonly [Input],
    > & { __keepDims: KeepDims; __maxAxes: Axes }
    : never

Max reduction operation Computes the maximum of tensor elements along specified axes

Type Parameters

Input extends AnyTensorStorage
Input tensor storage type
Axes extends readonly number[] | undefined
Axes to reduce along (undefined means reduce all)
KeepDims extends boolean = false
Whether to keep reduced dimensions as size 1
OutputDType extends AnyDType = Input["__dtype"]
Output data type (defaults to same as input)

Example

// Max along axis 1, remove dimension
type Input = TensorStorage<Float32, [2, 3, 4], [12, 4, 1], DefaultLayoutFlags>;
type Result = MaxOp<Input, [1], false>; // Shape: [2, 4]

Example

// Global max (all elements)
type GlobalMax = MaxOp<Input, undefined, false>; // Shape: []