def gdw_blk
import bendtorch-gpu@1.3.0.0/main.bend as BendtorchGpuimport Base
515# block bl's mean and 1/sqrt(var + eps), then its rows
516def gdw_blk(+hw: Nat, +i: U32, +bl: U32, +c: U32, s: GS) -> GS:
517 match s:
518 case GS{x, dy, mr, sg, sb}: gdw_blk1(hw, i, bl, c, sg, sb, x, dy, Array.get(F32, mr, (bl * 2 : U32)))