type HO
import bend-ml-llama@0.2.1.0/main.bend as BendMlLlamaimport Base
766# one query head: softmax(q·K · scale) · V, and the two arrays handed back
767type HO is Type:
768 HO{k: Array<F32>, v: Array<F32>, o: List<&2, F32>}
import bend-ml-llama@0.2.1.0/main.bend as BendMlLlamaimport Base
766# one query head: softmax(q·K · scale) · V, and the two arrays handed back
767type HO is Type:
768 HO{k: Array<F32>, v: Array<F32>, o: List<&2, F32>}