23. layers/gdn — gated delta net

Part: III · Layers
Package: github.com/openfluke/welvet/layers/gdn
Status: ok — ✅

When

Building or training a net that needs the layers/gdn Op (also usable as a Parallel cam).

Where

import "github.com/openfluke/welvet/layers/gdn"

cd 23-gdn && source ../env.sh && go run .

Why

Linear attention / decode-first mixers (Gated DeltaNet) need a first-class package under KindLinearAttn.

What

Exec CPU/SIMD/WebGPU; ForwardDecode; truncated BPTT. Timed matrix is Float32-primary: PermutationOK is f32 × FormatNone/BinaryPacked × CPU/SIMD/WebGPU; other cells GAP (declared), not FAIL.

Sample output (captured)

<nil>

Live capture from go run ./cmd/runall on local ../../welvet (exit 0).

Source

Copied from the Welvet feature book examples (openfluke.github.io/welvet/examples/23-gdn).

package main

import (
	"fmt"

	"github.com/openfluke/welvet/layers/gdn"
)

func main() {
	l, err := gdn.New(gdn.Config{
		HiddenSize: 64, NumKeyHeads: 4, NumValueHeads: 4,
		KeyHeadDim: 16, ValueHeadDim: 16, ConvKernel: 4,
	})
	if err != nil {
		fmt.Println("config rejected:", err)
		return
	}
	l.Reset()
	x := make([]float32, 64)
	y := make([]float32, 64)
	fmt.Println(l.ForwardDecode(x, y))
}