Struct KVCacheConfig 

pub struct KVCacheConfig {
    pub dim: usize,
    pub key_bits: u8,
    pub value_bits: u8,
    pub key_strategy: QuantStrategy,
    pub seed: u64,
    pub max_tokens: usize,
}

Configuration for a quantized KV cache.

Fields

dim: usize

Vector dimension (must match model’s head dimension).

key_bits: u8

Bits per coordinate for key quantization.

value_bits: u8

Bits per coordinate for value quantization.

key_strategy: QuantStrategy

Quantization strategy for keys.

seed: u64

Random seed for quantizer initialization.

max_tokens: usize

Maximum number of tokens. When exceeded, oldest tokens are evicted (sliding window). 0 means unlimited.

Implementations

impl KVCacheConfig

pub fn new(dim: usize) -> Self

Create a default config for a given dimension.

Defaults: 4-bit keys with Prod strategy, 4-bit values with MSE strategy.

pub fn with_key_bits(self, bits: u8) -> Self

Set key bit width.

pub fn with_value_bits(self, bits: u8) -> Self

Set value bit width.

pub fn with_key_strategy(self, strategy: QuantStrategy) -> Self

Set key quantization strategy.

pub fn with_seed(self, seed: u64) -> Self

Set random seed.

pub fn with_max_tokens(self, max_tokens: usize) -> Self

Set maximum token capacity (0 = unlimited). When the cache exceeds this limit, oldest tokens are evicted.

Trait Implementations

impl Clone for KVCacheConfig

fn clone(&self) -> KVCacheConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for KVCacheConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for KVCacheConfig

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for KVCacheConfig

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

where __S: Serializer,

Serialize this value into the given Serde serializer.