pub enum QuantizationGrouping {
Fixed {
size: u32,
},
WholeAxis,
Block2d {
block_shape: [NonZeroU32; 2],
},
}Expand description
How values are partitioned by a quantized physical layout.
WholeAxis is shape-relative by design: all values on the group axis
share one scale. This represents channelwise quantization without making a
matrix dimension part of the otherwise stable quantization-format ABI.
Block2d carries the exact rectangular source block shape; the physical
layout separately binds its two dimensions to ordered logical axes.
Variants§
Implementations§
Source§impl QuantizationGrouping
impl QuantizationGrouping
pub const fn fixed(size: u32) -> Self
pub const fn fixed_size(self) -> Option<u32>
pub const fn block_2d(block_shape: [NonZeroU32; 2]) -> Self
pub const fn block_shape_2d(self) -> Option<[NonZeroU32; 2]>
pub const fn resolved_size(self, axis_extent: u64) -> u64
Trait Implementations§
Source§impl Clone for QuantizationGrouping
impl Clone for QuantizationGrouping
Source§fn clone(&self) -> QuantizationGrouping
fn clone(&self) -> QuantizationGrouping
Returns a duplicate of the value. Read more
1.0.0 (const: unstable) · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
Performs copy-assignment from
source. Read moreimpl Copy for QuantizationGrouping
Source§impl Debug for QuantizationGrouping
impl Debug for QuantizationGrouping
Source§impl<'de> Deserialize<'de> for QuantizationGrouping
impl<'de> Deserialize<'de> for QuantizationGrouping
Source§fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>where
__D: Deserializer<'de>,
fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>where
__D: Deserializer<'de>,
Deserialize this value from the given Serde deserializer. Read more
impl Eq for QuantizationGrouping
Source§impl PartialEq for QuantizationGrouping
impl PartialEq for QuantizationGrouping
Source§impl Serialize for QuantizationGrouping
impl Serialize for QuantizationGrouping
impl StructuralPartialEq for QuantizationGrouping
Auto Trait Implementations§
impl Freeze for QuantizationGrouping
impl RefUnwindSafe for QuantizationGrouping
impl Send for QuantizationGrouping
impl Sync for QuantizationGrouping
impl Unpin for QuantizationGrouping
impl UnsafeUnpin for QuantizationGrouping
impl UnwindSafe for QuantizationGrouping
Blanket Implementations§
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