pub struct FullAttentionScratch {Show 16 fields
pub x_norm: DevicePtr,
pub q_full: DevicePtr,
pub q_split: DevicePtr,
pub gate_split: DevicePtr,
pub k: DevicePtr,
pub v: DevicePtr,
pub q_norm_out: DevicePtr,
pub k_norm_out: DevicePtr,
pub attn_out: DevicePtr,
pub gated_attn: DevicePtr,
pub o: DevicePtr,
pub x_resid: DevicePtr,
pub x_norm2: DevicePtr,
pub gate_act: DevicePtr,
pub up_act: DevicePtr,
pub x_out: DevicePtr,
}Expand description
Per-call scratch buffers for the full-attention forward.
Fields§
§x_norm: DevicePtr§q_full: DevicePtr§q_split: DevicePtr§gate_split: DevicePtr§k: DevicePtr§v: DevicePtr§q_norm_out: DevicePtr§k_norm_out: DevicePtr§attn_out: DevicePtr§gated_attn: DevicePtr§o: DevicePtr§x_resid: DevicePtr§x_norm2: DevicePtr§gate_act: DevicePtr§up_act: DevicePtr§x_out: DevicePtrAuto Trait Implementations§
impl Freeze for FullAttentionScratch
impl RefUnwindSafe for FullAttentionScratch
impl Send for FullAttentionScratch
impl Sync for FullAttentionScratch
impl Unpin for FullAttentionScratch
impl UnwindSafe for FullAttentionScratch
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