Struct ExecutorAttentionConfig 

pub struct ExecutorAttentionConfig {
    pub attention_type: AttentionType,
    pub enable_flash_attention: bool,
    pub enable_paged_attention: bool,
    pub block_size: Option<usize>,
    pub sliding_window_size: Option<usize>,
}

Runtime attention configuration for model executor

Note: This is different from ferrum_types::AttentionConfig which describes the model architecture’s attention configuration from config.json. This type describes the runtime execution settings.

Fields

attention_type: AttentionType

Type of attention to use

enable_flash_attention: bool

Enable flash attention if available

enable_paged_attention: bool

Enable paged attention

block_size: Option<usize>

Block size for paged attention

sliding_window_size: Option<usize>

Sliding window size (if using sliding window attention)

Trait Implementations

impl Clone for ExecutorAttentionConfig

fn clone(&self) -> ExecutorAttentionConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ExecutorAttentionConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for ExecutorAttentionConfig

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for ExecutorAttentionConfig

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

where __S: Serializer,

Serialize this value into the given Serde serializer.