Struct QuantizedFp8 

pub struct QuantizedFp8 {
    pub qvalues: Vec<u8>,
    pub scales: Vec<f32>,
    pub groups: Vec<u32>,
    pub variant: Fp8Variant,
}

An f32 tensor quantized to f8 bytes, one f32 scale per group (the tensor is scaled by 1 / scale before conversion, matching the affine schemes’ convention of storing a multiplicative dequantization scale).

Fields

qvalues: Vec<u8>

One f8-encoded byte per input element, same order as the input.

scales: Vec<f32>

One scale per group.

groups: Vec<u32>

qvalues[i]’s group is groups[i].

variant: Fp8Variant

Which variant qvalues was encoded with.

Trait Implementations

impl Clone for QuantizedFp8

fn clone(&self) -> QuantizedFp8

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuantizedFp8

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

Formats the value using the given formatter.