pub struct AttentionDef {
pub name: String,
pub num_heads: Option<usize>,
pub num_kv_heads: Option<usize>,
pub head_dim: Option<usize>,
pub dropout: f64,
pub bias: bool,
pub position_encoding: PositionEncoding,
pub window_size: Option<usize>,
pub causal: bool,
pub qk_norm: bool,
}Expand description
Attention mechanism definition
Fields§
§name: String§num_heads: Option<usize>§num_kv_heads: Option<usize>§head_dim: Option<usize>§dropout: f64§bias: bool§position_encoding: PositionEncoding§window_size: Option<usize>§causal: bool§qk_norm: boolPer-head RMSNorm on Q and K before RoPE (OLMo2/Gemma-style stabilizer)
Trait Implementations§
Source§impl Clone for AttentionDef
impl Clone for AttentionDef
Source§fn clone(&self) -> AttentionDef
fn clone(&self) -> AttentionDef
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 moreSource§impl Debug for AttentionDef
impl Debug for AttentionDef
Source§impl Default for AttentionDef
impl Default for AttentionDef
Source§impl<'de> Deserialize<'de> for AttentionDef
impl<'de> Deserialize<'de> for AttentionDef
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
Auto Trait Implementations§
impl Freeze for AttentionDef
impl RefUnwindSafe for AttentionDef
impl Send for AttentionDef
impl Sync for AttentionDef
impl Unpin for AttentionDef
impl UnsafeUnpin for AttentionDef
impl UnwindSafe for AttentionDef
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