Enum Format 

pub enum Format {
    Bf16,
    F32,
    I32,
    Fp8E4m3 {
        scale: Scale,
    },
    Nvfp4 {
        group: u32,
    },
}

2026-09-28: A tensor format.

Variants

Bf16

2026-09-28: bfloat16.

F32

2026-09-28: float32.

I32

2026-09-28: int32 (token and expert ids).

Fp8E4m3

2026-09-28: FP8 E4M3 with the given scale layout.

Fields

scale: Scale

2026-09-28: Scale sharing.

Nvfp4

2026-09-28: NVFP4: E2M1 values, one E4M3 scale per group values, one F32 global.

Fields

group: u32

2026-09-28: Values per scale.

Implementations

impl Format

pub fn parse(s: &str) -> Result<Self, FormatError>

2026-09-28: Parse the canonical spelling, e.g. bf16, fp8/token, fp8/block128x128, nvfp4/g16.

pub fn name(&self) -> String

2026-09-28: The canonical spelling.

pub fn is_plain(&self) -> bool

2026-09-28: True for an unquantized format (16/32-bit floats and int32).

pub fn is_edge_format(&self) -> bool

2026-09-28: True when the scale layout is one an activation edge can carry.

pub fn weight_bytes(&self, n: u64, k: u64) -> Option<u64>

2026-09-30: Bytes of a [n, k] weight in this format, scales included: as Format::bytes, except that an NVFP4 weight has one F32 global scale for the whole tensor (weight_scale_2), where an activation has one per row.

pub fn bytes(&self, rows: u64, dim: u64) -> Option<u64>

2026-09-28: Bytes of a rows x dim tensor in this format, scales included. Used by the buffer planner. None when dim is not a multiple of the scale group.

Trait Implementations

impl Clone for Format

fn clone(&self) -> Format

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for Format

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

Formats the value using the given formatter.

impl Display for Format

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

Formats the value using the given formatter.

impl Hash for Format

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl Ord for Format

fn cmp(&self, other: &Format) -> Ordering

This method returns an Ordering between self and other.

fn max(self, other: Self) -> Self

where Self: Sized,

Compares and returns the maximum of two values.

fn min(self, other: Self) -> Self

where Self: Sized,

Compares and returns the minimum of two values.

fn clamp(self, min: Self, max: Self) -> Self

where Self: Sized,

Restrict a value to a certain interval.

impl PartialEq for Format

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

Tests for self and other values to be equal, and is used by ==.

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

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl PartialOrd for Format

fn partial_cmp(&self, other: &Format) -> Option<Ordering>

This method returns an ordering between self and other values if one exists.

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

Tests less than (for self and other) and is used by the < operator.

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

Tests less than or equal to (for self and other) and is used by the <= operator.

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

Tests greater than (for self and other) and is used by the > operator.

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

Tests greater than or equal to (for self and other) and is used by the >= operator.

impl Copy for Format

impl Eq for Format

impl StructuralPartialEq for Format