Enum DType 

pub enum DType {
    F16,
    Bf16,
    F8E4m3,
    F8E5m2,
}

Element type for Q / K / V tiles. Distinct from atomr-accel-cuda’s future CudaDtype so the FlashAttn crate is self-contained.

Variants

F16

IEEE 754 binary16 — fa2 + fa3.

Bf16

bfloat16 — fa2 + fa3.

F8E4m3

8-bit float, e4m3 — fa3 only, sm_90a+.

F8E5m2

8-bit float, e5m2 — fa3 only, sm_90a+ (used for V in DPA-mixed-precision).

Implementations

impl DType

pub fn size_in_bytes(self) -> usize

Element width in bytes.

pub fn is_fp8(self) -> bool

True iff this dtype is one of the fp8 variants.

pub fn tag(self) -> &'static str

Short tag used inside the kernel-name mangling.

Trait Implementations

impl Clone for DType

fn clone(&self) -> DType

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for DType

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

Formats the value using the given formatter.

impl Hash for DType

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 PartialEq for DType

fn eq(&self, other: &DType) -> 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 Copy for DType

impl Eq for DType

impl StructuralPartialEq for DType