def Q8.quantize

bendtorch@1.3.0.0 · dtype.bend:383

import Base
import bend-ml-tensor-array@0.1.18.0/main.bend as TA

383# W quantized by rows (the matrix is consumed)
384def Q8.quantize(+r: Nat, +c: Nat, m: TA.Mat<r, c>) -> Q8<r, c>:
385  q8_fin(r, c, Nat.mul(r, c), r, q8_rows(r, c, ml_l(r, c, TA.Mat.to_list(r, c, m))))