Enum PrecisionMode 

pub enum PrecisionMode {
    FP32,
    FP16,
    BF16,
    Mixed {
        compute: ComputePrecision,
        accumulate_fp32: bool,
    },
}

Precision mode for inference

Variants

FP32

Full precision (FP32)

FP16

Half precision (FP16) - good for NVIDIA GPUs

BF16

Brain float 16 (BF16) - good for modern accelerators

Mixed

Mixed precision - compute in FP16/BF16 but accumulate in FP32

Fields

compute: ComputePrecision

Compute precision

accumulate_fp32: bool

Whether to accumulate in FP32

Implementations

impl PrecisionMode

pub fn is_reduced_precision(&self) -> bool

Check if this mode uses reduced precision

pub fn memory_reduction_factor(&self) -> f32

Get the memory reduction factor compared to FP32

pub fn name(&self) -> &str

Get human-readable name

Trait Implementations

impl Clone for PrecisionMode

fn clone(&self) -> PrecisionMode

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PrecisionMode

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

Formats the value using the given formatter.

impl Default for PrecisionMode

fn default() -> PrecisionMode

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for PrecisionMode

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for PrecisionMode

fn eq(&self, other: &PrecisionMode) -> 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 Serialize for PrecisionMode

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl Copy for PrecisionMode

impl Eq for PrecisionMode

impl StructuralPartialEq for PrecisionMode