xma.functional.swiglu.cuda_implementation.forward

swiglu_forward_cuda_jit(mG: cutlass.cute.Tensor, mU: cutlass.cute.Tensor, mY: cutlass.cute.Tensor) → None

swiglu_forward_cuda_kernel(gG: cutlass.cute.Tensor, gU: cutlass.cute.Tensor, gY: cutlass.cute.Tensor, gID: cutlass.cute.Tensor, copy_atom: cutlass.cute.CopyAtom, tiled_copy: cutlass.cute.TiledCopy, shape: cutlass.cute.Shape) → None