pub struct MoeDef {
pub experts: usize,
pub top_k: usize,
pub shared_experts: usize,
pub load_balance_loss_weight: f64,
pub router_z_loss_weight: f64,
}Expand description
Configurable dropless token-choice mixture of experts.
experts are routed experts; shared_experts are always active. Router
regularization belongs to the architecture config so every training entry
point applies the same stable objective.
Fields§
§experts: usize§top_k: usize§load_balance_loss_weight: f64§router_z_loss_weight: f64Trait Implementations§
Source§impl<'de> Deserialize<'de> for MoeDef
impl<'de> Deserialize<'de> for MoeDef
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 MoeDef
impl RefUnwindSafe for MoeDef
impl Send for MoeDef
impl Sync for MoeDef
impl Unpin for MoeDef
impl UnsafeUnpin for MoeDef
impl UnwindSafe for MoeDef
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