pub struct ExpertWeights<'a> {
pub gate_proj: &'a MlxBuffer,
pub up_proj: &'a MlxBuffer,
pub down_proj: &'a MlxBuffer,
}Expand description
A single expert’s weight matrices (float32, pre-dequantized or float).
Each expert has three projection matrices:
gate_proj:[input_dim, intermediate_dim]row-majorup_proj:[input_dim, intermediate_dim]row-majordown_proj:[intermediate_dim, input_dim]row-major
Fields§
§gate_proj: &'a MlxBuffer§up_proj: &'a MlxBuffer§down_proj: &'a MlxBufferAuto Trait Implementations§
impl<'a> Freeze for ExpertWeights<'a>
impl<'a> RefUnwindSafe for ExpertWeights<'a>
impl<'a> Send for ExpertWeights<'a>
impl<'a> Sync for ExpertWeights<'a>
impl<'a> Unpin for ExpertWeights<'a>
impl<'a> UnsafeUnpin for ExpertWeights<'a>
impl<'a> UnwindSafe for ExpertWeights<'a>
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