Enum ProtoTensor 

pub enum ProtoTensor {
    Quantized {
        protos: Array3<i8>,
        quantization: Quantization,
    },
    Float(Array3<f32>),
}

Prototype tensor variants for fused decode+render pipelines.

Carries either raw quantized data (to skip CPU dequantization and let the GPU shader dequantize) or dequantized f32 data (from float models or legacy paths).

Variants

Quantized

Raw int8 protos with quantization parameters — skip CPU dequantization. The GPU fragment shader will dequantize per-texel using the scale and zero_point.

Fields

protos: Array3<i8>

quantization: Quantization

Float(Array3<f32>)

Dequantized f32 protos (from float models or legacy path).

Implementations

impl ProtoTensor

pub fn is_quantized(&self) -> bool

Returns true if this is the quantized variant.

pub fn dim(&self) -> (usize, usize, usize)

Returns the spatial dimensions (height, width, num_protos).

pub fn as_f32(&self) -> Cow<'_, Array3<f32>>

Returns dequantized f32 protos. For the Float variant this is a no-copy reference; for Quantized it allocates and dequantizes.

Trait Implementations

impl Clone for ProtoTensor

fn clone(&self) -> ProtoTensor

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ProtoTensor

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

Formats the value using the given formatter.