Enum Approximation 

pub enum Approximation {
    Exact,
    Close,
    Approximate,
    VeryApproximate,
    SuperApproximate,
    UltraApproximate,
    Custom(f32, f32, f32),
    Ulp(u64),
}

Variants

Exact

Close

Approximate

VeryApproximate

SuperApproximate

UltraApproximate

Custom(f32, f32, f32)

Ulp(u64)

Compare by integer ULP distance in the reference tensor’s own float type, accepting a distance up to the given bound.

Unlike the tolerance-based variants this does not go through an f32 cast, so an f16 comparison stays an f16 comparison. Use it to assert that two implementations of a kernel agree to within a known number of rounding steps.

Trait Implementations

impl Clone for Approximation

fn clone(&self) -> Approximation

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for Approximation

impl Debug for Approximation

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

Formats the value using the given formatter.

impl Default for Approximation

fn default() -> Approximation

Returns the “default value” for a type.

impl Eq for Approximation

impl From<bool> for Approximation

fn from(b: bool) -> Self

Converts to this type from the input type.

impl PartialEq for Approximation

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

Equality operator ==.

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

Inequality operator !=.