Struct KvTensorShape 

pub struct KvTensorShape {
    pub attention_type: AttentionType,
    pub num_layers: u32,
    pub num_heads: u32,
    pub num_kv_heads: u32,
    pub head_dim: usize,
    pub hidden_size: usize,
}

Describes the tensor shape of a KV cache for a specific model family.

This is the logical layout, not the physical memory layout.

Fields

attention_type: AttentionType

The attention mechanism type.

num_layers: u32

Number of transformer layers.

num_heads: u32

Number of attention heads.

num_kv_heads: u32

Number of key/value heads (for GQA: smaller than num_heads).

head_dim: usize

Dimension of each attention head (usually 64 or 128).

hidden_size: usize

Hidden size / model dimension (d_model).

Implementations

impl KvTensorShape

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

Validate that the shape is internally consistent.

pub fn kv_elements_per_token_per_layer(&self) -> usize

Total KV elements per token (key + value) for one layer.

pub fn kv_bytes_per_token_per_layer(&self) -> usize

Total KV bytes per token (f32) for one layer.

pub fn total_kv_elements(&self, num_tokens: usize) -> usize

Total KV elements for N tokens across all layers.

pub fn total_kv_bytes(&self, num_tokens: usize) -> usize

Total raw bytes for N tokens across all layers.

pub fn head_vector_dim(&self) -> usize

Dimension for a single key or value vector for one KV head.

pub fn kv_vectors_per_token_per_layer(&self) -> usize

Number of KV head vectors per token per layer.

Trait Implementations

impl Clone for KvTensorShape

fn clone(&self) -> KvTensorShape

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for KvTensorShape

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for KvTensorShape

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for KvTensorShape

impl PartialEq for KvTensorShape

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for KvTensorShape

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for KvTensorShape