Struct InterpolateOptions 

pub struct InterpolateOptions {
    pub mode: InterpolateMode,
    pub coordinate_transform: CoordinateTransformMode,
    pub nearest_mode: NearestRoundingMode,
    pub cubic_options: CubicOptions,
    pub output_size: Option<[usize; 2]>,
    pub scale_factor: Option<[f32; 2]>,
}

Interpolation options.

Flex supports all coordinate transforms, rounding modes and cubic options. CubeCL supports the default nearest and nearest-exact combinations, plus bilinear, bicubic and Lanczos3 with AlignCorners or HalfPixel and default cubic options. With CubeCL, explicit scales must equal the resolved output/input ratios exactly, unless the transform is AlignCorners. Unsupported combinations panic.

Fields

mode: InterpolateMode

Algorithm used.

coordinate_transform: CoordinateTransformMode

Coordinate transformation mode.

nearest_mode: NearestRoundingMode

Nearest rounding mode.

cubic_options: CubicOptions

Cubic interpolation options.

output_size: Option<[usize; 2]>

Output spatial size [height, width].

Exactly one of output_size or scale_factor must be set when calling burn_tensor::module::interpolate.

scale_factor: Option<[f32; 2]>

Scale factor [scale_height, scale_width] applied to the input spatial size. The output size is floor(input_size * scale_factor). Scale-dependent transforms use these supplied factors for coordinates rather than recomputing the output/input ratio. AlignCorners and TfCropAndResize map coordinates from the resolved sizes instead.

Implementations

impl InterpolateOptions

pub fn new(mode: InterpolateMode) -> Self

Create new interpolate options with mode-dependent defaults.

pub fn with_coordinate_transform(self, mode: CoordinateTransformMode) -> Self

Set coordinate transformation mode.

pub fn with_nearest_mode(self, mode: NearestRoundingMode) -> Self

Set nearest rounding mode.

pub fn with_cubic_options(self, options: CubicOptions) -> Self

Set cubic options.

pub fn with_align_corners(self, align_corners: bool) -> Self

Set corner alignment for bilinear, bicubic and Lanczos3 interpolation. true selects AlignCorners; false selects HalfPixel. Nearest modes are unchanged; use Self::with_coordinate_transform to set their mapping.

pub fn align_corners(&self) -> bool

Returns true if coordinate transform is AlignCorners.

pub fn with_output_size(self, output_size: [usize; 2]) -> Self

Set the output size [height, width], clearing any scale factor.

pub fn with_scale_factor(self, scale_factor: [f32; 2]) -> Self

Set the scale factor [scale_height, scale_width], clearing any output size.

Trait Implementations

impl Clone for InterpolateOptions

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for InterpolateOptions

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for InterpolateOptions

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for InterpolateOptions

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl Serialize for InterpolateOptions

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for InterpolateOptions