pub fn absorb_q(
cfg: &Glm5NextDsaConfig,
q_b: &[f32],
kv_b: &[f32],
full_heads: usize,
) -> Result<Vec<f32>>Expand description
vs nope + v_head_dim = 512). Using one stride for the other still yields a
well-formed 2-D tensor of plausible values.