rope_yarn_scaled

Function rope_yarn_scaled 

Source
pub fn rope_yarn_scaled(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    q: DevicePtr,
    k: DevicePtr,
    positions: DevicePtr,
    seq_len: u32,
    num_q_heads: u32,
    num_kv_heads: u32,
    head_dim: u32,
    rotary_dim: u32,
    inv_freq: DevicePtr,
    attention_factor: f32,
    stream: u64,
) -> Result<()>
Expand description

YaRN RoPE for standard attention with explicit cosine/sine amplitude.