pub struct SharedV41 {
pub compress_kv: Option<DevicePtr>,
pub compress_len: usize,
pub index_k: Option<DevicePtr>,
pub topk_idxs: Vec<i32>,
pub topk: usize,
pub candidates: Vec<bool>,
pub cand_width: usize,
}Expand description
The SharedAttentionRuntime: what source layers publish for the layers
below them.
Fields§
§compress_kv: Option<DevicePtr>the current kv source’s latent cache and how many groups it holds
compress_len: usize§index_k: Option<DevicePtr>the current kv source’s index-key cache
topk_idxs: Vec<i32>[queries][topk], offset for the window rows, -1 = absent
topk: usize§candidates: Vec<bool>[queries][cand_width]
cand_width: usizeTrait Implementations§
Auto Trait Implementations§
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