pub enum MoeProviderRole {
Replica,
Router,
ExpertHolder,
PrefillDecode,
Prefill,
Decode,
}Expand description
MoE pipeline roles a provider can play in distributed serving.
Variants§
Replica
Holds the full model; serves single-peer inference. The default.
Router
Runs the gating-network step and fans out batched expert calls to the appropriate expert holders.
ExpertHolder
Holds one or more experts declared in moe_holdings.
PrefillDecode
Runs both prefill and decode phases co-located (SOTA central pattern; the Tenzro fallback when only one provider can fit the model).
Prefill
Runs only the prefill phase; hands off KV cache to a decode
peer over iroh. Pairs with Decode.
Decode
Runs only the decode phase; accepts KV cache from a prefill peer over iroh.
Trait Implementations§
Source§impl Clone for MoeProviderRole
impl Clone for MoeProviderRole
Source§fn clone(&self) -> MoeProviderRole
fn clone(&self) -> MoeProviderRole
Returns a duplicate of the value. Read more
1.0.0 (const: unstable) · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
Performs copy-assignment from
source. Read moreimpl Copy for MoeProviderRole
Source§impl Debug for MoeProviderRole
impl Debug for MoeProviderRole
Source§impl<'de> Deserialize<'de> for MoeProviderRole
impl<'de> Deserialize<'de> for MoeProviderRole
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
impl Eq for MoeProviderRole
Source§impl Hash for MoeProviderRole
impl Hash for MoeProviderRole
Source§impl PartialEq for MoeProviderRole
impl PartialEq for MoeProviderRole
Source§fn eq(&self, other: &MoeProviderRole) -> bool
fn eq(&self, other: &MoeProviderRole) -> bool
Tests for
self and other values to be equal, and is used by ==.Source§impl Serialize for MoeProviderRole
impl Serialize for MoeProviderRole
impl StructuralPartialEq for MoeProviderRole
Auto Trait Implementations§
impl Freeze for MoeProviderRole
impl RefUnwindSafe for MoeProviderRole
impl Send for MoeProviderRole
impl Sync for MoeProviderRole
impl Unpin for MoeProviderRole
impl UnsafeUnpin for MoeProviderRole
impl UnwindSafe for MoeProviderRole
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