pub struct AttnMasters {
pub wqkv: Vec<f32>,
pub wo: Vec<f32>,
pub gproj: Option<Vec<f32>>,
pub iln: Vec<f32>,
pub pln: Vec<f32>,
}Expand description
Trained attention + norm masters of a full-layer FCD layer (f32).
Fields§
§wqkv: Vec<f32>Fused [qrows + 2·kvdim, hidden] (q rows first, then k, then v).
wo: Vec<f32>[hidden, nh·hd].
gproj: Option<Vec<f32>>Per-head gate projection [nh, hidden] (Spark-X2.5), if the layer has one.
iln: Vec<f32>§pln: Vec<f32>Trait Implementations§
Auto Trait Implementations§
impl Freeze for AttnMasters
impl RefUnwindSafe for AttnMasters
impl Send for AttnMasters
impl Sync for AttnMasters
impl Unpin for AttnMasters
impl UnsafeUnpin for AttnMasters
impl UnwindSafe for AttnMasters
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