AVX for phonemizer

cmd/train_phonemizer/main.go

@@ -9,7 +9,7 @@ import "flag"
 import "github.com/neurlang/classifier/datasets/phonemizer"
 import "github.com/neurlang/classifier/layer/majpool2d"
 import "github.com/neurlang/classifier/datasets"
-import "github.com/neurlang/classifier/learning"
+import "github.com/neurlang/classifier/learning/avx"
 import "github.com/neurlang/classifier/net/feedforward"
 
 func error_abs(a, b uint32) uint32 {
@@ -22,6 +22,7 @@ func error_abs(a, b uint32) uint32 {
 func main() {
 	cleantsv := flag.String("cleantsv", "", "clean tsv dataset for the language")
 	dstmodel := flag.String("dstmodel", "", "model destination .json.lzw file")
+	resume := flag.Bool("resume", false, "resume training")
 	flag.Parse()
 
 	var improved_success_rate = 0
@@ -49,7 +50,7 @@ func main() {
 	net.NewCombiner(majpool2d.MustNew(fanout2, 1, fanout1, 1, fanout2, 1, 1))
 	net.NewLayer(1, 0)
 
-	//net.ReadCompressedWeightsFromFile("output.94.json.t.lzw")
+
 
 	trainWorst := func(worst int) {
 		var tally = new(datasets.Tally)
@@ -76,7 +77,7 @@ func main() {
 			wg.Wait()
 		}
 
-		var h learning.HyperParameters
+		var h avx.HyperParameters
 		h.Threads = runtime.NumCPU()
 		h.Factor = 1 // affects the solution size
 
@@ -100,6 +101,9 @@ func main() {
 
 		h.Name = fmt.Sprint(worst)
 		h.SetLogger("solutions11.txt")
+
+		h.AvxLanes = 16
+		h.AvxSkip = 4
 
 		fmt.Println("hashtron position:", worst, "(job size:", tally.Len(), ")")
 
@@ -131,25 +135,31 @@ func main() {
 		success := percent * 100 / len(datakeys)
 		println("[success rate]", success, "%", "with", errsum, "errors")
 
-		err := net.WriteCompressedWeightsToFile("output." + fmt.Sprint(success) + ".json.t.lzw")
-		if err != nil {
-			println(err.Error())
+		if dstmodel == nil {
+			err := net.WriteCompressedWeightsToFile("output." + fmt.Sprint(success) + ".json.t.lzw")
+			if err != nil {
+				println(err.Error())
+			}
 		}
 
 		if dstmodel != nil && len(*dstmodel) > 0 && improved_success_rate < success {
-			improved_success_rate = success
-			err := net.WriteCompressedWeightsToFile(*dstmodel)
-			if err != nil {
-				println(err.Error())
+			if improved_success_rate > 0 {
+				err := net.WriteCompressedWeightsToFile(*dstmodel)
+				if err != nil {
+					println(err.Error())
+				}
 			}
+			improved_success_rate = success
 		}
 
 		if success == 100 {
 			println("Max accuracy or wrong data. Exiting")
 			os.Exit(0)
 		}
 	}
-
+	if resume != nil && *resume && dstmodel != nil {
+		net.ReadCompressedWeightsFromFile(*dstmodel)
+	}
 	for {
 		shuf := net.Shuffle(true)
 		evaluate()