Struct PromptCacheDescriptor 

pub struct PromptCacheDescriptor { /* private fields */ }

Caller-supplied identity and geometry for a reusable prefix cache.

Implementations

impl PromptCacheDescriptor

pub fn new( model_family: impl Into<String>, effective_model_type: impl Into<String>, checkpoint_fingerprint: impl Into<String>, prefix_content_fingerprint: impl Into<String>, architecture_fingerprint: impl Into<String>, layer_count: usize, global_layer_start: usize, global_layer_end: usize, batch_size: usize, layer_layout: LayerSchedule<LayerCachePolicy>, layer_prefix_offsets: Vec<i32>, state_segments: Vec<PromptCacheStateSegment>, sink_tokens: usize, topology: PromptCacheTopology, ) -> Result<Self, PromptCacheError>

Creates and validates a complete reusable prefix-cache descriptor.

pub fn model_family(&self) -> &str

Stable architecture family.

pub fn effective_model_type(&self) -> &str

Effective normalized model type.

pub fn checkpoint_fingerprint(&self) -> &str

Caller-verified checkpoint identity.

pub fn prefix_content_fingerprint(&self) -> &str

Prefix-content identity.

pub fn architecture_fingerprint(&self) -> &str

Cache-relevant architecture identity.

pub const fn layer_count(&self) -> usize

Total model layer count.

pub const fn global_layer_start(&self) -> usize

Inclusive first global layer stored by this rank.

pub const fn global_layer_end(&self) -> usize

Exclusive global layer boundary stored by this rank.

pub const fn batch_size(&self) -> usize

Prefix batch size.

pub const fn layer_layout(&self) -> &LayerSchedule<LayerCachePolicy>

Ordered cache layout.

pub fn layer_prefix_offsets(&self) -> &[i32]

Per-layer processed-token deltas.

pub fn state_segments(&self) -> &[PromptCacheStateSegment]

Named state-layout ranges.

pub const fn sink_tokens(&self) -> usize

Attention sink token count.

pub const fn topology(&self) -> &PromptCacheTopology

Distributed rank-local layout.

pub fn with_topology( self, topology: PromptCacheTopology, ) -> Result<Self, PromptCacheError>

Replaces the distributed topology and revalidates the descriptor.

pub fn with_architecture_fingerprint( self, architecture_fingerprint: impl Into<String>, ) -> Result<Self, PromptCacheError>

Replaces the cache-relevant architecture fingerprint.

pub fn with_layer_count( self, layer_count: usize, ) -> Result<Self, PromptCacheError>

Replaces the total model layer count while preserving the owned range.

pub fn from_model_identity( model: PromptCacheModelIdentity, checkpoint_fingerprint: impl Into<String>, prefix_content_fingerprint: impl Into<String>, batch_size: usize, ) -> Result<Self, PromptCacheError>

Derives every model-owned field from a prepared model identity.

The checkpoint and prefix-content fingerprints remain caller-owned because they identify the concrete weights and processed input rather than model structure.

pub fn validate(&self) -> Result<(), PromptCacheError>

Validates the complete portable identity and cache geometry.

Trait Implementations

impl Clone for PromptCacheDescriptor

fn clone(&self) -> PromptCacheDescriptor

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for PromptCacheDescriptor

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Eq for PromptCacheDescriptor

impl Hash for PromptCacheDescriptor

fn hash<__H: Hasher>(&self, state: &mut __H)

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl PartialEq for PromptCacheDescriptor

fn eq(&self, other: &PromptCacheDescriptor) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for PromptCacheDescriptor