Struct LoraAttnWeights 

pub struct LoraAttnWeights {
    pub layer_idx: usize,
    pub q: Option<LoraPair>,
    pub k: Option<LoraPair>,
    pub v: Option<LoraPair>,
    pub o: Option<LoraPair>,
    pub kernels: LoraKernels,
    pub q_route: Option<LoraRoute>,
    pub k_route: Option<LoraRoute>,
    pub v_route: Option<LoraRoute>,
    pub o_route: Option<LoraRoute>,
}

2026-09-25: Per-layer attention LoRA weights, installed by copy onto Qwen3AttentionLayer.

The q pair folds its delta into the raw q_proj output at offset 0 over the full q_proj width (on a gated model the interleaved [Q|gate] row), before deinterleave_qg splits it (apply_q_lora in qwen3_attention/decode/attention_forward.rs).

Fields

layer_idx: usize

2026-09-25: The global layer index (0..num_hidden_layers), set at install. Prefill uses it to index a request slot’s per-layer pairs; attn_layer_idx counts attention layers only and differs from it on hybrid GDN/attention models.

q: Option<LoraPair>

k: Option<LoraPair>

v: Option<LoraPair>

o: Option<LoraPair>

kernels: LoraKernels

q_route: Option<LoraRoute>

2026-09-25: Per-request routing tables, per module. None means no routing, and the module’s pair above is applied. With Some, a step that carries a seq_slot buffer applies the routed delta through apply_lora_bgmv.

k_route: Option<LoraRoute>

v_route: Option<LoraRoute>

o_route: Option<LoraRoute>

Trait Implementations

impl Clone for LoraAttnWeights

fn clone(&self) -> LoraAttnWeights

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for LoraAttnWeights