67. TrainMode — 29 named updates

Part: IV · Runtime
Package: github.com/openfluke/welvet/layers/parallel
Status: ok — ✅ 29 modes

When

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

Where

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

cd 67-train-modes && source ../env.sh && go run .

Why

Backprop is one update, not the only one. Credit assignment (broadcast gap, head proxy, sparse duty clock) has to be a named axis you can race — not a comment in a notebook. Cameral Mix also needs one TrainMode per hemisphere on the same loss.

What

parallel.TrainMode: AllNamedTrainModes() = 29 (Inherit omitted). Stack-local Split/Alt plus Step* 1D pipe twins and Mesh* grid schedulers. TrainStackMSE / TrainStackCE honour BranchModes. Display names use Short() / ShortTrainMode. Rival metric is hard Acc vs StepBP; Lucy Score is Tput × Avail × Acc — do not mix those sentences.

Sample output (captured)

named 31 linestep 11
stepfastproxy Step[T][S][FP] <nil>
[T]=Tween  [S]=Split  [FP]=FastProxy  [L]=Linear  [HP]=HeadProxy  [F]=Freeze  [Sh]=Shadow  [A]=Adv  [M]=Memory

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/67-train-modes).

package main

import (
	"fmt"

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

func main() {
	named := parallel.AllNamedTrainModes()
	line := 0
	for _, m := range named {
		if m.IsLineStep() {
			line++
		}
	}
	fp, err := parallel.ParseTrainMode("stepfastproxy")
	fmt.Println("named", len(named), "linestep", line)
	fmt.Println("stepfastproxy", fp.Short(), err)
	fmt.Println(parallel.ShortTrainModeLegend)
}