Skip to main content

compressed_sparse_attention_shared_mqa_f32

singe_kernel::cpu::attention

Function compressed_sparse_attention_shared_mqa_f32 

Source 
pub fn compressed_sparse_attention_shared_mqa_f32(
    query: &[f32],
    kv_entries: &[f32],
    valid_mask: &[i32],
    sink: &[f32],
    config: CsaSharedMqaConfig,
) -> Vec<f32>