Struct kn_graph::optimizer::OptimizerSettings
source · pub struct OptimizerSettings {
pub optimize: bool,
pub force_bias_through_conv: bool,
pub fuse_layernorm: bool,
pub div_to_mul: bool,
}Expand description
Settings for the optimizer.
Use Default::default() to get reasonable defaults.
Fields§
§optimize: boolIf false, don’t do any optimization at all.
force_bias_through_conv: boolIf true, convert a bias operation followed by a convolution through the convolution,
even in cases where this requires switching to a non-spatially-broadcasted bias constant.
fuse_layernorm: boolIf true, try fusing the right sequence of operations into a single LayerNorm operation.
div_to_mul: boolIf true, convert a division by a constant into multiplication by the inverse consent.
Trait Implementations§
source§impl Clone for OptimizerSettings
impl Clone for OptimizerSettings
source§fn clone(&self) -> OptimizerSettings
fn clone(&self) -> OptimizerSettings
Returns a copy of the value. Read more
1.0.0 · source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
Performs copy-assignment from
source. Read moresource§impl Debug for OptimizerSettings
impl Debug for OptimizerSettings
source§impl Default for OptimizerSettings
impl Default for OptimizerSettings
impl Copy for OptimizerSettings
Auto Trait Implementations§
impl RefUnwindSafe for OptimizerSettings
impl Send for OptimizerSettings
impl Sync for OptimizerSettings
impl Unpin for OptimizerSettings
impl UnwindSafe for OptimizerSettings
Blanket Implementations§
source§impl<S, D, Swp, Dwp, T> AdaptInto<D, Swp, Dwp, T> for Swhere
T: Real + Zero + Arithmetics + Clone,
Swp: WhitePoint<T>,
Dwp: WhitePoint<T>,
D: AdaptFrom<S, Swp, Dwp, T>,
impl<S, D, Swp, Dwp, T> AdaptInto<D, Swp, Dwp, T> for Swhere T: Real + Zero + Arithmetics + Clone, Swp: WhitePoint<T>, Dwp: WhitePoint<T>, D: AdaptFrom<S, Swp, Dwp, T>,
source§fn adapt_into_using<M>(self, method: M) -> Dwhere
M: TransformMatrix<T>,
fn adapt_into_using<M>(self, method: M) -> Dwhere M: TransformMatrix<T>,
Convert the source color to the destination color using the specified
method.
source§fn adapt_into(self) -> D
fn adapt_into(self) -> D
Convert the source color to the destination color using the bradford
method by default.
source§impl<T, C> ArraysFrom<C> for Twhere
C: IntoArrays<T>,
impl<T, C> ArraysFrom<C> for Twhere C: IntoArrays<T>,
source§fn arrays_from(colors: C) -> T
fn arrays_from(colors: C) -> T
Cast a collection of colors into a collection of arrays.
source§impl<T, C> ArraysInto<C> for Twhere
C: FromArrays<T>,
impl<T, C> ArraysInto<C> for Twhere C: FromArrays<T>,
source§fn arrays_into(self) -> C
fn arrays_into(self) -> C
Cast this collection of arrays into a collection of colors.
source§impl<T> BorrowMut<T> for Twhere
T: ?Sized,
impl<T> BorrowMut<T> for Twhere T: ?Sized,
source§fn borrow_mut(&mut self) -> &mut T
fn borrow_mut(&mut self) -> &mut T
Mutably borrows from an owned value. Read more
source§impl<T, C> ComponentsFrom<C> for Twhere
C: IntoComponents<T>,
impl<T, C> ComponentsFrom<C> for Twhere C: IntoComponents<T>,
source§fn components_from(colors: C) -> T
fn components_from(colors: C) -> T
Cast a collection of colors into a collection of color components.
source§impl<T> FromAngle<T> for T
impl<T> FromAngle<T> for T
source§fn from_angle(angle: T) -> T
fn from_angle(angle: T) -> T
Performs a conversion from
angle.source§impl<T, U> FromStimulus<U> for Twhere
U: IntoStimulus<T>,
impl<T, U> FromStimulus<U> for Twhere U: IntoStimulus<T>,
source§fn from_stimulus(other: U) -> T
fn from_stimulus(other: U) -> T
Converts
other into Self, while performing the appropriate scaling,
rounding and clamping.source§impl<T, U> IntoAngle<U> for Twhere
U: FromAngle<T>,
impl<T, U> IntoAngle<U> for Twhere U: FromAngle<T>,
source§fn into_angle(self) -> U
fn into_angle(self) -> U
Performs a conversion into
T.source§impl<T, U> IntoColor<U> for Twhere
U: FromColor<T>,
impl<T, U> IntoColor<U> for Twhere U: FromColor<T>,
source§fn into_color(self) -> U
fn into_color(self) -> U
Convert into T with values clamped to the color defined bounds Read more
source§impl<T, U> IntoColorUnclamped<U> for Twhere
U: FromColorUnclamped<T>,
impl<T, U> IntoColorUnclamped<U> for Twhere U: FromColorUnclamped<T>,
source§fn into_color_unclamped(self) -> U
fn into_color_unclamped(self) -> U
Convert into T. The resulting color might be invalid in its color space Read more
source§impl<T> IntoStimulus<T> for T
impl<T> IntoStimulus<T> for T
source§fn into_stimulus(self) -> T
fn into_stimulus(self) -> T
Converts
self into T, while performing the appropriate scaling,
rounding and clamping.source§impl<'a, T, C> TryComponentsInto<C> for Twhere
C: TryFromComponents<T>,
impl<'a, T, C> TryComponentsInto<C> for Twhere C: TryFromComponents<T>,
§type Error = <C as TryFromComponents<T>>::Error
type Error = <C as TryFromComponents<T>>::Error
The error for when
try_into_colors fails to cast.source§fn try_components_into(self) -> Result<C, <T as TryComponentsInto<C>>::Error>
fn try_components_into(self) -> Result<C, <T as TryComponentsInto<C>>::Error>
Try to cast this collection of color components into a collection of
colors. Read more
source§impl<T, U> TryIntoColor<U> for Twhere
U: TryFromColor<T>,
impl<T, U> TryIntoColor<U> for Twhere U: TryFromColor<T>,
source§fn try_into_color(self) -> Result<U, OutOfBounds<U>>
fn try_into_color(self) -> Result<U, OutOfBounds<U>>
Convert into T, returning ok if the color is inside of its defined
range, otherwise an
OutOfBounds error is returned which contains
the unclamped color. Read moresource§impl<C, U> UintsFrom<C> for Uwhere
C: IntoUints<U>,
impl<C, U> UintsFrom<C> for Uwhere C: IntoUints<U>,
source§fn uints_from(colors: C) -> U
fn uints_from(colors: C) -> U
Cast a collection of colors into a collection of unsigned integers.