Struct QRow 

pub struct QRow {
    pub rows: usize,
    pub cols: usize,
    pub bits: u32,
    /* private fields */
}

Per-row (per-output-channel) quantized 2D matrix at bits ∈ {4,8}, packed 32/bits per word, with one scale per row — more accurate than a single per-tensor scale, and int4 is 1/8 the memory.

Fields

rows: usize

cols: usize

bits: u32

Implementations

impl QRow

pub fn nbytes(&self) -> usize

pub fn dequant(&self) -> Tensor

Dequantize back to an f32 [rows, cols] tensor, on-device.