added training progress and bugfixes
This commit is contained in:
Linrador
parent
6e91c352a7
commit
ba4d5ba255
@ -3,6 +3,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"encoding/json"
|
||||
"errors"
|
||||
"fmt"
|
||||
"net/url"
|
||||
@ -17,7 +18,8 @@ type autoStartItem struct {
|
||||
userKey string
|
||||
url string
|
||||
blind bool
|
||||
source string // "watched" | "manual"
|
||||
source string // "watched" | "manual" | "resume"
|
||||
force bool // true = ignoriert Autostart-Pause + Download-Limit
|
||||
}
|
||||
|
||||
func normUser(s string) string {
|
||||
@ -254,7 +256,37 @@ func clearAllPendingAutoStartOnStartup() error {
|
||||
continue
|
||||
}
|
||||
|
||||
if err := os.Remove(filepath.Join(dir, name)); err != nil && !errors.Is(err, os.ErrNotExist) {
|
||||
path := filepath.Join(dir, name)
|
||||
|
||||
raw, err := os.ReadFile(path)
|
||||
if err != nil {
|
||||
if errors.Is(err, os.ErrNotExist) {
|
||||
continue
|
||||
}
|
||||
return err
|
||||
}
|
||||
|
||||
var f pendingAutoStartFile
|
||||
if err := json.Unmarshal(raw, &f); err != nil {
|
||||
_ = os.Remove(path)
|
||||
continue
|
||||
}
|
||||
|
||||
kept := make([]PendingAutoStartItem, 0, len(f.Items))
|
||||
for _, it := range f.Items {
|
||||
if normalizePendingSourceServer(it.Source) == "resume" {
|
||||
kept = append(kept, it)
|
||||
}
|
||||
}
|
||||
|
||||
if len(kept) == 0 {
|
||||
if err := os.Remove(path); err != nil && !errors.Is(err, os.ErrNotExist) {
|
||||
return err
|
||||
}
|
||||
continue
|
||||
}
|
||||
|
||||
if err := savePendingAutoStartItems(strings.TrimSuffix(name, filepath.Ext(name)), kept); err != nil {
|
||||
return err
|
||||
}
|
||||
}
|
||||
@ -301,9 +333,7 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
for {
|
||||
select {
|
||||
case <-scanTicker.C:
|
||||
if isAutostartPaused() {
|
||||
continue
|
||||
}
|
||||
autostartPaused := isAutostartPaused()
|
||||
|
||||
s := getSettings()
|
||||
if !s.UseChaturbateAPI {
|
||||
@ -340,6 +370,15 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
showByUser := map[string]string{}
|
||||
imageByUser := map[string]string{}
|
||||
|
||||
for _, r := range rooms {
|
||||
u := normUser(r.Username)
|
||||
if u == "" {
|
||||
continue
|
||||
}
|
||||
showByUser[u] = normalizePendingShowServer(r.CurrentShow)
|
||||
imageByUser[u] = selectBestRoomImageURL(r)
|
||||
}
|
||||
|
||||
pendingAutoStartMu.Lock()
|
||||
manualItems, err := loadManualPendingAutoStartItemsForProvider("chaturbate")
|
||||
pendingAutoStartMu.Unlock()
|
||||
@ -350,6 +389,16 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
continue
|
||||
}
|
||||
|
||||
pendingAutoStartMu.Lock()
|
||||
resumeItems, err := loadResumePendingAutoStartItemsForProvider("chaturbate")
|
||||
pendingAutoStartMu.Unlock()
|
||||
if err != nil {
|
||||
if verboseLogs() {
|
||||
fmt.Println("⚠️ [autostart] load resume chaturbate pending failed:", err)
|
||||
}
|
||||
continue
|
||||
}
|
||||
|
||||
manualByUser := map[string]PendingAutoStartItem{}
|
||||
manualOrder := make([]string, 0, len(manualItems))
|
||||
|
||||
@ -376,13 +425,30 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
manualOrder = append(manualOrder, key)
|
||||
}
|
||||
|
||||
for _, r := range rooms {
|
||||
u := normUser(r.Username)
|
||||
if u == "" {
|
||||
resumeByUser := map[string]PendingAutoStartItem{}
|
||||
resumeOrder := make([]string, 0, len(resumeItems))
|
||||
|
||||
for _, it := range resumeItems {
|
||||
key := normUser(it.ModelKey)
|
||||
if key == "" {
|
||||
key = chaturbateUserFromURL(it.URL)
|
||||
}
|
||||
if key == "" {
|
||||
continue
|
||||
}
|
||||
showByUser[u] = normalizePendingShowServer(r.CurrentShow)
|
||||
imageByUser[u] = selectBestRoomImageURL(r)
|
||||
|
||||
it.ModelKey = key
|
||||
it.URL = strings.TrimSpace(it.URL)
|
||||
if it.URL == "" {
|
||||
it.URL = fmt.Sprintf("https://chaturbate.com/%s/", key)
|
||||
}
|
||||
|
||||
if _, exists := resumeByUser[key]; exists {
|
||||
continue
|
||||
}
|
||||
|
||||
resumeByUser[key] = it
|
||||
resumeOrder = append(resumeOrder, key)
|
||||
}
|
||||
|
||||
// laufende Jobs sammeln
|
||||
@ -467,7 +533,8 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
continue
|
||||
}
|
||||
|
||||
if it.source == "manual" {
|
||||
switch it.source {
|
||||
case "manual":
|
||||
m, ok := manualByUser[it.userKey]
|
||||
if !ok {
|
||||
continue
|
||||
@ -477,7 +544,21 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
if it.url == "" {
|
||||
continue
|
||||
}
|
||||
} else {
|
||||
|
||||
case "resume":
|
||||
m, ok := resumeByUser[it.userKey]
|
||||
if !ok {
|
||||
continue
|
||||
}
|
||||
|
||||
it.url = strings.TrimSpace(m.URL)
|
||||
if it.url == "" {
|
||||
continue
|
||||
}
|
||||
|
||||
it.force = true
|
||||
|
||||
default:
|
||||
m, ok := watchedByUser[it.userKey]
|
||||
if !ok {
|
||||
continue
|
||||
@ -522,9 +603,98 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
}
|
||||
|
||||
offlineCandidates := make([]autoStartItem, 0, len(watchedOrder))
|
||||
nextPending := make([]PendingAutoStartItem, 0, len(watchedOrder))
|
||||
nextPending := make([]PendingAutoStartItem, 0, len(watchedOrder)+len(resumeOrder)+len(runningByUser))
|
||||
now := time.Now()
|
||||
|
||||
resumePendingThisScan := map[string]bool{}
|
||||
|
||||
// Sichtbare laufende Downloads bei private/hidden/away stoppen
|
||||
// und als "resume" merken. Diese Resume-Einträge starten später
|
||||
// unabhängig von Autostart-Pause und unabhängig vom Download-Limit.
|
||||
for user, runningJob := range runningByUser {
|
||||
if runningJob == nil {
|
||||
continue
|
||||
}
|
||||
if runningJob.Hidden {
|
||||
continue
|
||||
}
|
||||
|
||||
show := normalizePendingShowServer(showByUser[user])
|
||||
if show != "private" && show != "hidden" && show != "away" {
|
||||
continue
|
||||
}
|
||||
|
||||
u := strings.TrimSpace(runningJob.SourceURL)
|
||||
if u == "" {
|
||||
u = fmt.Sprintf("https://chaturbate.com/%s/", user)
|
||||
}
|
||||
|
||||
img := strings.TrimSpace(imageByUser[user])
|
||||
|
||||
nextPending = append(nextPending, PendingAutoStartItem{
|
||||
ModelKey: user,
|
||||
URL: u,
|
||||
Mode: "wait_public",
|
||||
CurrentShow: show,
|
||||
ImageURL: img,
|
||||
Source: "resume",
|
||||
})
|
||||
|
||||
resumePendingThisScan[user] = true
|
||||
|
||||
if verboseLogs() {
|
||||
fmt.Println("⏸️ [autostart] stopped because model is no longer public:", user, show)
|
||||
}
|
||||
|
||||
stopJobsInternal([]*RecordJob{runningJob})
|
||||
}
|
||||
|
||||
// Resume hat Vorrang und ignoriert Autostart-Pause.
|
||||
for _, user := range resumeOrder {
|
||||
itm := resumeByUser[user]
|
||||
|
||||
u := strings.TrimSpace(itm.URL)
|
||||
if u == "" {
|
||||
u = fmt.Sprintf("https://chaturbate.com/%s/", user)
|
||||
}
|
||||
|
||||
show := normalizePendingShowServer(showByUser[user])
|
||||
img := strings.TrimSpace(imageByUser[user])
|
||||
|
||||
switch show {
|
||||
case "public":
|
||||
if runningByUser[user] != nil {
|
||||
continue
|
||||
}
|
||||
if queued[user] {
|
||||
continue
|
||||
}
|
||||
|
||||
queue = append(queue, autoStartItem{
|
||||
userKey: user,
|
||||
url: u,
|
||||
blind: false,
|
||||
source: "resume",
|
||||
force: true,
|
||||
})
|
||||
queued[user] = true
|
||||
|
||||
case "private", "hidden", "away", "offline", "unknown":
|
||||
if resumePendingThisScan[user] {
|
||||
continue
|
||||
}
|
||||
|
||||
nextPending = append(nextPending, PendingAutoStartItem{
|
||||
ModelKey: user,
|
||||
URL: u,
|
||||
Mode: "wait_public",
|
||||
CurrentShow: show,
|
||||
ImageURL: img,
|
||||
Source: "resume",
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
for _, user := range watchedOrder {
|
||||
m := watchedByUser[user]
|
||||
|
||||
@ -538,6 +708,9 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
|
||||
switch show {
|
||||
case "public":
|
||||
if autostartPaused {
|
||||
continue
|
||||
}
|
||||
if runningByUser[user] != nil {
|
||||
continue
|
||||
}
|
||||
@ -566,11 +739,10 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
Source: "watched",
|
||||
})
|
||||
|
||||
if runningJob := runningByUser[user]; runningJob != nil {
|
||||
stopJobsInternal([]*RecordJob{runningJob})
|
||||
}
|
||||
|
||||
default:
|
||||
if autostartPaused {
|
||||
continue
|
||||
}
|
||||
if runningByUser[user] != nil {
|
||||
continue
|
||||
}
|
||||
@ -606,6 +778,9 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
|
||||
switch show {
|
||||
case "public":
|
||||
if autostartPaused {
|
||||
continue
|
||||
}
|
||||
if runningByUser[user] != nil {
|
||||
continue
|
||||
}
|
||||
@ -629,6 +804,9 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
continue
|
||||
|
||||
default:
|
||||
if autostartPaused {
|
||||
continue
|
||||
}
|
||||
if runningByUser[user] != nil {
|
||||
continue
|
||||
}
|
||||
@ -648,7 +826,8 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
}
|
||||
}
|
||||
|
||||
// Nur EIN Offline-Kandidat gleichzeitig in die Queue
|
||||
// Nur EIN Offline-Kandidat gleichzeitig in die Queue.
|
||||
// Bei pausiertem Autostart werden oben keine normalen Offline-Kandidaten erzeugt.
|
||||
if !blindQueued && !hasHiddenProbeRunningForProvider("chaturbate") && len(offlineCandidates) > 0 {
|
||||
n := len(offlineCandidates)
|
||||
start := 0
|
||||
@ -675,7 +854,7 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
}
|
||||
}
|
||||
|
||||
if selectedBlindUser != "" {
|
||||
if selectedBlindUser != "" && !autostartPaused {
|
||||
if m, ok := watchedByUser[selectedBlindUser]; ok {
|
||||
u := resolveChaturbateURL(m)
|
||||
if u != "" {
|
||||
@ -705,10 +884,6 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
pendingAutoStartMu.Unlock()
|
||||
|
||||
case <-startTicker.C:
|
||||
if isAutostartPaused() {
|
||||
continue
|
||||
}
|
||||
|
||||
s := getSettings()
|
||||
if !s.UseChaturbateAPI {
|
||||
continue
|
||||
@ -736,18 +911,34 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
showByUser[u] = strings.ToLower(strings.TrimSpace(r.CurrentShow))
|
||||
}
|
||||
|
||||
it := queue[0]
|
||||
show := strings.TrimSpace(showByUser[it.userKey])
|
||||
isPublic := strings.Contains(show, "public")
|
||||
paused := isAutostartPaused()
|
||||
|
||||
// Nicht-public nur einzeln nacheinander prüfen
|
||||
if it.blind && hasHiddenProbeRunningForProvider("chaturbate") {
|
||||
startIdx := -1
|
||||
for i, candidate := range queue {
|
||||
if paused && !candidate.force {
|
||||
continue
|
||||
}
|
||||
|
||||
startIdx = i
|
||||
break
|
||||
}
|
||||
|
||||
if startIdx < 0 {
|
||||
continue
|
||||
}
|
||||
|
||||
queue = queue[1:]
|
||||
it := queue[startIdx]
|
||||
queue = append(queue[:startIdx], queue[startIdx+1:]...)
|
||||
delete(queued, it.userKey)
|
||||
|
||||
show := strings.TrimSpace(showByUser[it.userKey])
|
||||
isPublic := strings.Contains(show, "public")
|
||||
|
||||
// Nicht-public nur einzeln nacheinander prüfen.
|
||||
if it.blind && hasHiddenProbeRunningForProvider("chaturbate") {
|
||||
continue
|
||||
}
|
||||
|
||||
if isJobRunningForURL(it.url) {
|
||||
continue
|
||||
}
|
||||
@ -756,8 +947,9 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
lastTry[it.userKey] = time.Now()
|
||||
|
||||
job, err := startRecordingInternal(RecordRequest{
|
||||
URL: it.url,
|
||||
Cookie: lastCookieHdr,
|
||||
URL: it.url,
|
||||
Cookie: lastCookieHdr,
|
||||
IgnoreConcurrentLimit: it.force,
|
||||
})
|
||||
if err != nil {
|
||||
if verboseLogs() {
|
||||
@ -767,10 +959,34 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
}
|
||||
|
||||
if verboseLogs() {
|
||||
fmt.Println("▶️ [autostart] started:", it.url)
|
||||
if it.source == "resume" {
|
||||
fmt.Println("▶️ [autostart] resumed:", it.url)
|
||||
} else {
|
||||
fmt.Println("▶️ [autostart] started:", it.url)
|
||||
}
|
||||
}
|
||||
|
||||
if job != nil {
|
||||
if it.source == "resume" {
|
||||
pendingAutoStartMu.Lock()
|
||||
_ = removeResumePendingAutoStartItemForProvider("chaturbate", it.userKey)
|
||||
pendingAutoStartMu.Unlock()
|
||||
|
||||
go chaturbateAbortIfNoOutput(job.ID, outputProbeMax, nil, func() {
|
||||
pendingAutoStartMu.Lock()
|
||||
_ = saveResumePendingAutoStartItemForProvider("chaturbate", PendingAutoStartItem{
|
||||
ModelKey: it.userKey,
|
||||
URL: it.url,
|
||||
Mode: "wait_public",
|
||||
CurrentShow: "unknown",
|
||||
Source: "resume",
|
||||
})
|
||||
pendingAutoStartMu.Unlock()
|
||||
})
|
||||
|
||||
continue
|
||||
}
|
||||
|
||||
if it.source == "manual" {
|
||||
go chaturbateAbortIfNoOutput(job.ID, outputProbeMax, func() {
|
||||
pendingAutoStartMu.Lock()
|
||||
@ -786,12 +1002,17 @@ func startChaturbateAutoStartWorker(store *ModelStore) {
|
||||
}
|
||||
}
|
||||
} else {
|
||||
if paused && !it.force {
|
||||
continue
|
||||
}
|
||||
|
||||
lastBlindTry[it.userKey] = time.Now()
|
||||
|
||||
job, err := startRecordingInternal(RecordRequest{
|
||||
URL: it.url,
|
||||
Cookie: lastCookieHdr,
|
||||
Hidden: true,
|
||||
URL: it.url,
|
||||
Cookie: lastCookieHdr,
|
||||
Hidden: true,
|
||||
IgnoreConcurrentLimit: it.force,
|
||||
})
|
||||
if err != nil || job == nil {
|
||||
if verboseLogs() {
|
||||
|
||||
@ -1,55 +1,245 @@
|
||||
# backend\ml\train_detector_model.py
|
||||
# backend/ml/train_detector_model.py
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import shutil
|
||||
from pathlib import Path
|
||||
|
||||
from ultralytics import YOLO
|
||||
|
||||
|
||||
def emit_progress(stage, progress, message="", **extra):
|
||||
out = {
|
||||
"type": "progress",
|
||||
"stage": stage,
|
||||
"progress": max(0.0, min(1.0, float(progress))),
|
||||
"message": message,
|
||||
}
|
||||
out.update(extra)
|
||||
print(json.dumps(out, ensure_ascii=False), flush=True)
|
||||
|
||||
|
||||
def count_yolo_samples(dataset_root: Path, split: str) -> int:
|
||||
images_dir = dataset_root / "images" / split
|
||||
labels_dir = dataset_root / "labels" / split
|
||||
|
||||
if not images_dir.exists() or not labels_dir.exists():
|
||||
return 0
|
||||
|
||||
image_exts = {".jpg", ".jpeg", ".png", ".webp"}
|
||||
count = 0
|
||||
|
||||
for image_path in images_dir.iterdir():
|
||||
if not image_path.is_file():
|
||||
continue
|
||||
|
||||
if image_path.suffix.lower() not in image_exts:
|
||||
continue
|
||||
|
||||
label_path = labels_dir / f"{image_path.stem}.txt"
|
||||
if label_path.exists() and label_path.stat().st_size > 0:
|
||||
count += 1
|
||||
|
||||
return count
|
||||
|
||||
|
||||
def safe_int(value, fallback):
|
||||
try:
|
||||
return int(value)
|
||||
except Exception:
|
||||
return fallback
|
||||
|
||||
|
||||
def main():
|
||||
parser = argparse.ArgumentParser()
|
||||
parser.add_argument("--root", required=True)
|
||||
parser.add_argument("--base", default="yolo11n.pt")
|
||||
parser.add_argument("--epochs", default="80")
|
||||
parser.add_argument("--imgsz", default="640")
|
||||
parser.add_argument("--device", default="cpu")
|
||||
parser.add_argument("--workers", default="2")
|
||||
parser.add_argument("--patience", default="20")
|
||||
args = parser.parse_args()
|
||||
|
||||
root = Path(args.root).resolve()
|
||||
yaml_path = root / "detector" / "dataset" / "dataset.yaml"
|
||||
dataset_root = root / "detector" / "dataset"
|
||||
yaml_path = dataset_root / "dataset.yaml"
|
||||
runs_dir = root / "detector" / "runs"
|
||||
out_dir = root / "detector" / "model"
|
||||
|
||||
epochs = max(1, safe_int(args.epochs, 80))
|
||||
imgsz = max(64, safe_int(args.imgsz, 640))
|
||||
workers = max(0, safe_int(args.workers, 2))
|
||||
patience = max(0, safe_int(args.patience, 20))
|
||||
|
||||
if not yaml_path.exists():
|
||||
raise SystemExit(f"dataset.yaml not found: {yaml_path}")
|
||||
|
||||
train_count = count_yolo_samples(dataset_root, "train")
|
||||
val_count = count_yolo_samples(dataset_root, "val")
|
||||
|
||||
emit_progress(
|
||||
"detector",
|
||||
0.01,
|
||||
"YOLO-Dataset wird geprüft…",
|
||||
trainSamples=train_count,
|
||||
valSamples=val_count,
|
||||
epochs=epochs,
|
||||
imgsz=imgsz,
|
||||
device=args.device,
|
||||
)
|
||||
|
||||
if train_count <= 0:
|
||||
raise SystemExit("no YOLO train samples found")
|
||||
|
||||
if val_count <= 0:
|
||||
raise SystemExit("no YOLO val samples found")
|
||||
|
||||
emit_progress(
|
||||
"detector",
|
||||
0.03,
|
||||
"YOLO-Basismodell wird geladen…",
|
||||
base=args.base,
|
||||
)
|
||||
|
||||
model = YOLO(args.base)
|
||||
|
||||
best_epoch = 0
|
||||
last_epoch = 0
|
||||
|
||||
def on_train_epoch_start(trainer):
|
||||
epoch = int(getattr(trainer, "epoch", 0)) + 1
|
||||
total = int(getattr(trainer, "epochs", epochs) or epochs)
|
||||
|
||||
emit_progress(
|
||||
"detector",
|
||||
0.04 + 0.90 * ((epoch - 1) / max(1, total)),
|
||||
f"Object Detector trainiert… Epoche {epoch}/{total}",
|
||||
epoch=epoch,
|
||||
epochs=total,
|
||||
trainSamples=train_count,
|
||||
valSamples=val_count,
|
||||
)
|
||||
|
||||
def on_train_epoch_end(trainer):
|
||||
nonlocal last_epoch
|
||||
|
||||
epoch = int(getattr(trainer, "epoch", 0)) + 1
|
||||
total = int(getattr(trainer, "epochs", epochs) or epochs)
|
||||
last_epoch = max(last_epoch, epoch)
|
||||
|
||||
emit_progress(
|
||||
"detector",
|
||||
0.04 + 0.90 * (epoch / max(1, total)),
|
||||
f"Object Detector trainiert… Epoche {epoch}/{total}",
|
||||
epoch=epoch,
|
||||
epochs=total,
|
||||
trainSamples=train_count,
|
||||
valSamples=val_count,
|
||||
)
|
||||
|
||||
def on_fit_epoch_end(trainer):
|
||||
nonlocal best_epoch
|
||||
|
||||
epoch = int(getattr(trainer, "epoch", 0)) + 1
|
||||
metrics = getattr(trainer, "metrics", None) or {}
|
||||
|
||||
# Ultralytics nutzt je nach Version unterschiedliche Keys.
|
||||
map50 = (
|
||||
metrics.get("metrics/mAP50(B)")
|
||||
or metrics.get("metrics/mAP50")
|
||||
or metrics.get("mAP50")
|
||||
)
|
||||
map5095 = (
|
||||
metrics.get("metrics/mAP50-95(B)")
|
||||
or metrics.get("metrics/mAP50-95")
|
||||
or metrics.get("mAP50-95")
|
||||
)
|
||||
|
||||
if map50 is not None or map5095 is not None:
|
||||
best_epoch = epoch
|
||||
|
||||
emit_progress(
|
||||
"detector",
|
||||
0.04 + 0.90 * (epoch / max(1, epochs)),
|
||||
f"Object Detector validiert… Epoche {epoch}/{epochs}",
|
||||
epoch=epoch,
|
||||
epochs=epochs,
|
||||
mAP50=map50,
|
||||
mAP5095=map5095,
|
||||
)
|
||||
|
||||
model.add_callback("on_train_epoch_start", on_train_epoch_start)
|
||||
model.add_callback("on_train_epoch_end", on_train_epoch_end)
|
||||
model.add_callback("on_fit_epoch_end", on_fit_epoch_end)
|
||||
|
||||
emit_progress(
|
||||
"detector",
|
||||
0.05,
|
||||
"Object Detector Training startet…",
|
||||
trainSamples=train_count,
|
||||
valSamples=val_count,
|
||||
epochs=epochs,
|
||||
)
|
||||
|
||||
result = model.train(
|
||||
data=str(yaml_path),
|
||||
epochs=int(args.epochs),
|
||||
imgsz=int(args.imgsz),
|
||||
epochs=epochs,
|
||||
imgsz=imgsz,
|
||||
project=str(runs_dir),
|
||||
name="detect",
|
||||
exist_ok=True,
|
||||
device="cpu",
|
||||
workers=2,
|
||||
patience=20,
|
||||
device=args.device,
|
||||
workers=workers,
|
||||
patience=patience,
|
||||
)
|
||||
|
||||
emit_progress(
|
||||
"detector",
|
||||
0.96,
|
||||
"Bestes YOLO-Modell wird übernommen…",
|
||||
lastEpoch=last_epoch,
|
||||
bestEpoch=best_epoch,
|
||||
)
|
||||
|
||||
best = runs_dir / "detect" / "weights" / "best.pt"
|
||||
if not best.exists():
|
||||
raise SystemExit(f"best.pt not found after training: {best}")
|
||||
last = runs_dir / "detect" / "weights" / "last.pt"
|
||||
|
||||
if not best.exists():
|
||||
if last.exists():
|
||||
best = last
|
||||
else:
|
||||
raise SystemExit(f"best.pt not found after training: {best}")
|
||||
|
||||
out_dir = root / "detector" / "model"
|
||||
out_dir.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
final_model = out_dir / "best.pt"
|
||||
final_model.write_bytes(best.read_bytes())
|
||||
shutil.copy2(best, final_model)
|
||||
|
||||
print(json.dumps({
|
||||
result_path = runs_dir / "detect"
|
||||
status = {
|
||||
"ok": True,
|
||||
"model": str(final_model),
|
||||
"runs": str(runs_dir / "detect"),
|
||||
}))
|
||||
"sourceModel": str(best),
|
||||
"runs": str(result_path),
|
||||
"trainSamples": train_count,
|
||||
"valSamples": val_count,
|
||||
"epochs": epochs,
|
||||
"imgsz": imgsz,
|
||||
"device": str(args.device),
|
||||
}
|
||||
|
||||
with (out_dir / "status.json").open("w", encoding="utf-8") as f:
|
||||
json.dump(status, f, ensure_ascii=False, indent=2)
|
||||
|
||||
emit_progress(
|
||||
"detector",
|
||||
1.0,
|
||||
"Object Detector fertig.",
|
||||
**status,
|
||||
)
|
||||
|
||||
print(json.dumps(status, ensure_ascii=False), flush=True)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
|
||||
@ -51,6 +51,15 @@ def target_from_annotation(annotation):
|
||||
|
||||
return correction_target(annotation)
|
||||
|
||||
def emit_progress(stage, progress, message="", **extra):
|
||||
out = {
|
||||
"type": "progress",
|
||||
"stage": stage,
|
||||
"progress": max(0.0, min(1.0, float(progress))),
|
||||
"message": message,
|
||||
}
|
||||
out.update(extra)
|
||||
print(json.dumps(out, ensure_ascii=False), flush=True)
|
||||
|
||||
def load_clip():
|
||||
device = "cuda" if torch.cuda.is_available() else "cpu"
|
||||
@ -157,31 +166,48 @@ def main():
|
||||
model_dir.mkdir(parents=True, exist_ok=True)
|
||||
|
||||
rows = read_jsonl(feedback_path)
|
||||
total = max(1, len(rows))
|
||||
|
||||
emit_progress(
|
||||
"scene",
|
||||
0.02,
|
||||
"CLIP-Modell wird geladen…",
|
||||
totalSamples=len(rows),
|
||||
)
|
||||
|
||||
clip_model, processor, device = load_clip()
|
||||
|
||||
emit_progress(
|
||||
"scene",
|
||||
0.08,
|
||||
"CLIP-Embeddings werden vorbereitet…",
|
||||
totalSamples=len(rows),
|
||||
device=device,
|
||||
)
|
||||
|
||||
embeddings = []
|
||||
labels = []
|
||||
targets = []
|
||||
used = 0
|
||||
skipped = 0
|
||||
|
||||
for row in rows:
|
||||
for idx, row in enumerate(rows, start=1):
|
||||
sample_id = str(row.get("sampleId") or "").strip()
|
||||
if not sample_id:
|
||||
skipped += 1
|
||||
continue
|
||||
|
||||
image_path = frames_dir / f"{sample_id}.jpg"
|
||||
if not image_path.exists():
|
||||
skipped += 1
|
||||
continue
|
||||
|
||||
label = target_from_annotation(row)
|
||||
if not label:
|
||||
label = "unknown"
|
||||
|
||||
try:
|
||||
if not sample_id:
|
||||
skipped += 1
|
||||
continue
|
||||
|
||||
image_path = frames_dir / f"{sample_id}.jpg"
|
||||
if not image_path.exists():
|
||||
skipped += 1
|
||||
continue
|
||||
|
||||
label = target_from_annotation(row)
|
||||
if not label:
|
||||
label = "unknown"
|
||||
|
||||
emb = embed_image(clip_model, processor, device, image_path)
|
||||
|
||||
embeddings.append(emb)
|
||||
@ -191,13 +217,33 @@ def main():
|
||||
"sexPosition": label,
|
||||
})
|
||||
used += 1
|
||||
|
||||
except Exception as e:
|
||||
print(f"skip {sample_id}: {repr(e)}")
|
||||
print(f"skip {sample_id or '<missing>'}: {repr(e)}", flush=True)
|
||||
skipped += 1
|
||||
|
||||
finally:
|
||||
emit_progress(
|
||||
"scene",
|
||||
0.08 + 0.78 * (idx / total),
|
||||
f"Scene-Samples werden verarbeitet… {idx}/{len(rows)}",
|
||||
currentSample=idx,
|
||||
totalSamples=len(rows),
|
||||
usedSamples=used,
|
||||
skippedSamples=skipped,
|
||||
)
|
||||
|
||||
if used < 5:
|
||||
raise SystemExit(f"too few usable samples: {used}")
|
||||
|
||||
emit_progress(
|
||||
"scene",
|
||||
0.88,
|
||||
"Scene-Embeddings werden gespeichert…",
|
||||
usedSamples=used,
|
||||
skippedSamples=skipped,
|
||||
)
|
||||
|
||||
x = np.stack(embeddings).astype("float32")
|
||||
y = np.array(labels)
|
||||
|
||||
@ -210,9 +256,25 @@ def main():
|
||||
with (model_dir / "scene_clip_targets.json").open("w", encoding="utf-8") as f:
|
||||
json.dump(targets, f, ensure_ascii=False, indent=2)
|
||||
|
||||
emit_progress(
|
||||
"scene",
|
||||
0.93,
|
||||
"Scene-KNN wird trainiert…",
|
||||
usedSamples=used,
|
||||
skippedSamples=skipped,
|
||||
)
|
||||
|
||||
knn = train_knn(x, y)
|
||||
joblib.dump(knn, model_dir / "scene_clip_knn.joblib")
|
||||
|
||||
emit_progress(
|
||||
"scene",
|
||||
0.96,
|
||||
"Scene-Logistic-Regression wird trainiert…",
|
||||
usedSamples=used,
|
||||
skippedSamples=skipped,
|
||||
)
|
||||
|
||||
lr_status = "skipped_single_class"
|
||||
lr = train_lr_if_possible(x, y)
|
||||
if lr is not None:
|
||||
@ -246,6 +308,19 @@ def main():
|
||||
with (model_dir / "status.json").open("w", encoding="utf-8") as f:
|
||||
json.dump(status, f, ensure_ascii=False, indent=2)
|
||||
|
||||
emit_progress(
|
||||
"scene",
|
||||
1.0,
|
||||
"CLIP-Scene-Positionsmodell fertig.",
|
||||
usedSamples=used,
|
||||
skippedSamples=skipped,
|
||||
classes=sorted(counts.keys()),
|
||||
logisticRegression=lr_status,
|
||||
knn="trained",
|
||||
)
|
||||
|
||||
print(json.dumps(status, ensure_ascii=False), flush=True)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@ -41,11 +41,106 @@ func normalizePendingSourceServer(v string) string {
|
||||
switch strings.TrimSpace(strings.ToLower(v)) {
|
||||
case "watched":
|
||||
return "watched"
|
||||
case "resume":
|
||||
return "resume"
|
||||
default:
|
||||
return "manual"
|
||||
}
|
||||
}
|
||||
|
||||
func loadResumePendingAutoStartItemsForProvider(provider string) ([]PendingAutoStartItem, error) {
|
||||
provider = strings.ToLower(strings.TrimSpace(provider))
|
||||
if provider == "" {
|
||||
return nil, errors.New("missing provider")
|
||||
}
|
||||
|
||||
items, err := loadPendingAutoStartItems(pendingAutoStartGlobalUserKey)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
out := make([]PendingAutoStartItem, 0, len(items))
|
||||
for _, it := range items {
|
||||
if normalizePendingSourceServer(it.Source) != "resume" {
|
||||
continue
|
||||
}
|
||||
if pendingProviderFromURL(it.URL) != provider {
|
||||
continue
|
||||
}
|
||||
|
||||
out = append(out, it)
|
||||
}
|
||||
|
||||
return out, nil
|
||||
}
|
||||
|
||||
func removeResumePendingAutoStartItemForProvider(provider, modelKey string) error {
|
||||
provider = strings.ToLower(strings.TrimSpace(provider))
|
||||
modelKey = strings.ToLower(strings.TrimSpace(modelKey))
|
||||
|
||||
if provider == "" || modelKey == "" {
|
||||
return nil
|
||||
}
|
||||
|
||||
items, err := loadPendingAutoStartItems(pendingAutoStartGlobalUserKey)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
next := make([]PendingAutoStartItem, 0, len(items))
|
||||
for _, it := range items {
|
||||
if normalizePendingSourceServer(it.Source) == "resume" &&
|
||||
pendingProviderFromURL(it.URL) == provider &&
|
||||
strings.ToLower(strings.TrimSpace(it.ModelKey)) == modelKey {
|
||||
continue
|
||||
}
|
||||
|
||||
next = append(next, it)
|
||||
}
|
||||
|
||||
return savePendingAutoStartItems(pendingAutoStartGlobalUserKey, next)
|
||||
}
|
||||
|
||||
func saveResumePendingAutoStartItemForProvider(provider string, item PendingAutoStartItem) error {
|
||||
provider = strings.ToLower(strings.TrimSpace(provider))
|
||||
if provider == "" {
|
||||
return errors.New("missing provider")
|
||||
}
|
||||
|
||||
items, err := loadPendingAutoStartItems(pendingAutoStartGlobalUserKey)
|
||||
if err != nil {
|
||||
return err
|
||||
}
|
||||
|
||||
item.ModelKey = strings.ToLower(strings.TrimSpace(item.ModelKey))
|
||||
item.URL = strings.TrimSpace(item.URL)
|
||||
item.Mode = normalizePendingModeServer(item.Mode)
|
||||
item.CurrentShow = normalizePendingShowServer(item.CurrentShow)
|
||||
item.ImageURL = strings.TrimSpace(item.ImageURL)
|
||||
item.Source = "resume"
|
||||
|
||||
if item.ModelKey == "" || item.URL == "" {
|
||||
return nil
|
||||
}
|
||||
|
||||
replaced := false
|
||||
for i := range items {
|
||||
if normalizePendingSourceServer(items[i].Source) == "resume" &&
|
||||
pendingProviderFromURL(items[i].URL) == provider &&
|
||||
strings.ToLower(strings.TrimSpace(items[i].ModelKey)) == item.ModelKey {
|
||||
items[i] = item
|
||||
replaced = true
|
||||
break
|
||||
}
|
||||
}
|
||||
|
||||
if !replaced {
|
||||
items = append(items, item)
|
||||
}
|
||||
|
||||
return savePendingAutoStartItems(pendingAutoStartGlobalUserKey, items)
|
||||
}
|
||||
|
||||
func normalizePendingModeServer(v string) string {
|
||||
if strings.TrimSpace(strings.ToLower(v)) == "probe_retry" {
|
||||
return "probe_retry"
|
||||
|
||||
@ -3,6 +3,7 @@ package main
|
||||
|
||||
import (
|
||||
"context"
|
||||
"os"
|
||||
"strings"
|
||||
"sync"
|
||||
"time"
|
||||
@ -11,10 +12,11 @@ import (
|
||||
// Eine Nacharbeit (kann ffmpeg, ffprobe, thumbnails, rename, etc. enthalten)
|
||||
type PostWorkTask struct {
|
||||
Key string // z.B. Dateiname oder Job-ID, zum Deduplizieren
|
||||
Path string // Datei, die während queued/running gegen Explorer-Löschen gelockt wird
|
||||
Run func(ctx context.Context) error
|
||||
Added time.Time
|
||||
SortBucket int // 0 = /done, 1 = /done/keep
|
||||
SortName string // Dateiname lower-case
|
||||
SortBucket int
|
||||
SortName string
|
||||
}
|
||||
|
||||
type PostWorkQueue struct {
|
||||
@ -33,6 +35,8 @@ type PostWorkQueue struct {
|
||||
waitingKeys []string // sortierte wartende Keys
|
||||
runningKeys map[string]struct{} // Keys, die gerade wirklich laufen (Semaphor gehalten)
|
||||
cancelByKey map[string]context.CancelFunc
|
||||
|
||||
fileLocks map[string]*os.File
|
||||
}
|
||||
|
||||
func NewPostWorkQueue(queueSize int, maxParallelFFmpeg int) *PostWorkQueue {
|
||||
@ -53,6 +57,8 @@ func NewPostWorkQueue(queueSize int, maxParallelFFmpeg int) *PostWorkQueue {
|
||||
waitingKeys: make([]string, 0, queueSize),
|
||||
runningKeys: make(map[string]struct{}),
|
||||
cancelByKey: make(map[string]context.CancelFunc),
|
||||
|
||||
fileLocks: make(map[string]*os.File),
|
||||
}
|
||||
|
||||
pq.cond = sync.NewCond(&pq.mu)
|
||||
@ -110,6 +116,20 @@ func removeQueuedTaskLocked(tasks []PostWorkTask, key string) []PostWorkTask {
|
||||
return tasks
|
||||
}
|
||||
|
||||
func postWorkTaskLockPath(task PostWorkTask) string {
|
||||
return strings.TrimSpace(task.Path)
|
||||
}
|
||||
|
||||
func (pq *PostWorkQueue) unlockFileLocked(key string) {
|
||||
f := pq.fileLocks[key]
|
||||
if f == nil {
|
||||
return
|
||||
}
|
||||
|
||||
_ = f.Close()
|
||||
delete(pq.fileLocks, key)
|
||||
}
|
||||
|
||||
// Enqueue dedupliziert nach Key (damit du nicht durch Events doppelt queue-st)
|
||||
func (pq *PostWorkQueue) Enqueue(task PostWorkTask) bool {
|
||||
if task.Key == "" || task.Run == nil {
|
||||
@ -123,15 +143,30 @@ func (pq *PostWorkQueue) Enqueue(task PostWorkTask) bool {
|
||||
defer pq.mu.Unlock()
|
||||
|
||||
if _, ok := pq.inflight[task.Key]; ok {
|
||||
return false // schon queued oder läuft
|
||||
return false
|
||||
}
|
||||
if pq.maxQueued > 0 && len(pq.queue) >= pq.maxQueued {
|
||||
return false
|
||||
}
|
||||
|
||||
lockPath := postWorkTaskLockPath(task)
|
||||
var lockFile *os.File
|
||||
|
||||
if lockPath != "" {
|
||||
f, err := lockPostWorkFile(lockPath)
|
||||
if err != nil {
|
||||
return false
|
||||
}
|
||||
lockFile = f
|
||||
}
|
||||
|
||||
pq.inflight[task.Key] = struct{}{}
|
||||
pq.queued++
|
||||
|
||||
if lockFile != nil {
|
||||
pq.fileLocks[task.Key] = lockFile
|
||||
}
|
||||
|
||||
insertAt := len(pq.queue)
|
||||
for i, existing := range pq.queue {
|
||||
if lessPostWorkTask(task, existing) {
|
||||
@ -204,6 +239,8 @@ func (pq *PostWorkQueue) RemoveQueued(key string) bool {
|
||||
}
|
||||
|
||||
delete(pq.inflight, key)
|
||||
pq.unlockFileLocked(key)
|
||||
|
||||
if pq.queued > 0 {
|
||||
pq.queued--
|
||||
}
|
||||
@ -247,6 +284,7 @@ func (pq *PostWorkQueue) workerLoop(id int) {
|
||||
delete(pq.runningKeys, task.Key)
|
||||
delete(pq.inflight, task.Key)
|
||||
delete(pq.cancelByKey, task.Key)
|
||||
pq.unlockFileLocked(task.Key)
|
||||
if pq.queued > 0 {
|
||||
pq.queued--
|
||||
}
|
||||
@ -273,6 +311,10 @@ func (pq *PostWorkQueue) workerLoop(id int) {
|
||||
pq.mu.Lock()
|
||||
pq.removeWaitingKeyLocked(task.Key)
|
||||
pq.runningKeys[task.Key] = struct{}{}
|
||||
|
||||
// Wichtig: Lock vor der eigentlichen Nacharbeit freigeben,
|
||||
// damit moveToDoneDir/removeWithRetry unter Windows nicht blockiert werden.
|
||||
pq.unlockFileLocked(task.Key)
|
||||
pq.mu.Unlock()
|
||||
|
||||
if task.Run != nil {
|
||||
|
||||
13
backend/postwork_lock_other.go
Normal file
13
backend/postwork_lock_other.go
Normal file
@ -0,0 +1,13 @@
|
||||
// backend\postwork_lock_other.go
|
||||
|
||||
//go:build !windows
|
||||
|
||||
package main
|
||||
|
||||
import "os"
|
||||
|
||||
func lockPostWorkFile(path string) (*os.File, error) {
|
||||
// Auf Unix verhindert ein offenes Handle kein Löschen.
|
||||
// Der Fallback hält nur ein Handle, damit der Code plattformübergreifend baut.
|
||||
return os.Open(path)
|
||||
}
|
||||
34
backend/postwork_lock_windows.go
Normal file
34
backend/postwork_lock_windows.go
Normal file
@ -0,0 +1,34 @@
|
||||
// backend\postwork_lock_windows.go
|
||||
|
||||
//go:build windows
|
||||
|
||||
package main
|
||||
|
||||
import (
|
||||
"os"
|
||||
"syscall"
|
||||
)
|
||||
|
||||
const fileShareDelete = 0x00000004
|
||||
|
||||
func lockPostWorkFile(path string) (*os.File, error) {
|
||||
ptr, err := syscall.UTF16PtrFromString(path)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
handle, err := syscall.CreateFile(
|
||||
ptr,
|
||||
0, // desired access: nur Handle halten, kein Lesen/Schreiben nötig
|
||||
syscall.FILE_SHARE_READ|syscall.FILE_SHARE_WRITE, // bewusst OHNE FILE_SHARE_DELETE
|
||||
nil,
|
||||
syscall.OPEN_EXISTING,
|
||||
syscall.FILE_ATTRIBUTE_NORMAL,
|
||||
0,
|
||||
)
|
||||
if err != nil {
|
||||
return nil, err
|
||||
}
|
||||
|
||||
return os.NewFile(uintptr(handle), path), nil
|
||||
}
|
||||
@ -39,6 +39,9 @@ type RecordRequest struct {
|
||||
Cookie string `json:"cookie,omitempty"`
|
||||
UserAgent string `json:"userAgent,omitempty"`
|
||||
Hidden bool `json:"hidden,omitempty"`
|
||||
|
||||
// Intern: Resume nach private/hidden/away darf das normale Download-Limit ignorieren.
|
||||
IgnoreConcurrentLimit bool `json:"-"`
|
||||
}
|
||||
|
||||
type doneListResponse struct {
|
||||
|
||||
@ -456,7 +456,7 @@ func startRecordingInternal(req RecordRequest) (*RecordJob, error) {
|
||||
|
||||
// Limit-Check ATOMAR unter jobsMu
|
||||
s := getSettings()
|
||||
if s.EnableConcurrentDownloadsLimit {
|
||||
if s.EnableConcurrentDownloadsLimit && !req.IgnoreConcurrentLimit {
|
||||
max := s.MaxConcurrentDownloads
|
||||
if max < 1 {
|
||||
max = 1
|
||||
@ -1105,12 +1105,12 @@ func enqueuePostworkOrFail(job *RecordJob, out string, postTarget JobStatus) {
|
||||
|
||||
okQueued := postWorkQ.Enqueue(PostWorkTask{
|
||||
Key: postKey,
|
||||
Path: out,
|
||||
Added: time.Now(),
|
||||
Run: func(ctx context.Context) error {
|
||||
return runQueuedPostwork(ctx, job, out, postTarget, postKey)
|
||||
},
|
||||
})
|
||||
|
||||
if okQueued {
|
||||
publishQueuedPostworkState(job, postKey, postFile, postAssetID)
|
||||
return
|
||||
|
||||
@ -3,6 +3,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"bufio"
|
||||
"crypto/sha1"
|
||||
"encoding/hex"
|
||||
"encoding/json"
|
||||
@ -160,6 +161,139 @@ type TrainingStatsResponse struct {
|
||||
Labels TrainingStatsLabels `json:"labels"`
|
||||
}
|
||||
|
||||
type trainingProgressEvent struct {
|
||||
Type string `json:"type"`
|
||||
Stage string `json:"stage"`
|
||||
Progress float64 `json:"progress"` // 0..1
|
||||
Message string `json:"message,omitempty"`
|
||||
Epoch int `json:"epoch,omitempty"`
|
||||
Epochs int `json:"epochs,omitempty"`
|
||||
}
|
||||
|
||||
func trainingScaleProgress(local float64, start int, end int) int {
|
||||
if math.IsNaN(local) || math.IsInf(local, 0) {
|
||||
local = 0
|
||||
}
|
||||
|
||||
local = clamp01(local)
|
||||
|
||||
if end < start {
|
||||
end = start
|
||||
}
|
||||
|
||||
return start + int(math.Round(local*float64(end-start)))
|
||||
}
|
||||
|
||||
func trainingHandleProgressLine(line string, start int, end int, defaultStep string) bool {
|
||||
line = strings.TrimSpace(line)
|
||||
if line == "" {
|
||||
return false
|
||||
}
|
||||
|
||||
var ev trainingProgressEvent
|
||||
if err := json.Unmarshal([]byte(line), &ev); err != nil {
|
||||
return false
|
||||
}
|
||||
|
||||
if ev.Type != "progress" {
|
||||
return false
|
||||
}
|
||||
|
||||
progress := trainingScaleProgress(ev.Progress, start, end)
|
||||
step := strings.TrimSpace(ev.Message)
|
||||
if step == "" {
|
||||
step = defaultStep
|
||||
}
|
||||
|
||||
trainingSetJobStatus(func(s *TrainingJobStatus) {
|
||||
if progress > s.Progress {
|
||||
s.Progress = progress
|
||||
}
|
||||
s.Step = step
|
||||
})
|
||||
|
||||
return true
|
||||
}
|
||||
|
||||
func trainingRunCommandStreaming(
|
||||
python string,
|
||||
script string,
|
||||
onLine func(line string) bool,
|
||||
args ...string,
|
||||
) (string, error) {
|
||||
cmdArgs := append([]string{script}, args...)
|
||||
cmd := exec.Command(python, cmdArgs...)
|
||||
|
||||
cmd.SysProcAttr = &syscall.SysProcAttr{
|
||||
HideWindow: true,
|
||||
CreationFlags: 0x08000000,
|
||||
}
|
||||
|
||||
stdout, err := cmd.StdoutPipe()
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
|
||||
stderr, err := cmd.StderrPipe()
|
||||
if err != nil {
|
||||
return "", err
|
||||
}
|
||||
|
||||
if err := cmd.Start(); err != nil {
|
||||
return "", err
|
||||
}
|
||||
|
||||
var outMu sync.Mutex
|
||||
var lines []string
|
||||
|
||||
readPipe := func(scanner *bufio.Scanner) {
|
||||
scanner.Buffer(make([]byte, 0, 64*1024), 1024*1024)
|
||||
|
||||
for scanner.Scan() {
|
||||
line := strings.TrimSpace(scanner.Text())
|
||||
if line == "" {
|
||||
continue
|
||||
}
|
||||
|
||||
handled := false
|
||||
if onLine != nil {
|
||||
handled = onLine(line)
|
||||
}
|
||||
|
||||
// Progress-Events nicht in den finalen Output übernehmen.
|
||||
if handled {
|
||||
continue
|
||||
}
|
||||
|
||||
outMu.Lock()
|
||||
lines = append(lines, line)
|
||||
outMu.Unlock()
|
||||
}
|
||||
}
|
||||
|
||||
var wg sync.WaitGroup
|
||||
wg.Add(2)
|
||||
|
||||
go func() {
|
||||
defer wg.Done()
|
||||
readPipe(bufio.NewScanner(stdout))
|
||||
}()
|
||||
|
||||
go func() {
|
||||
defer wg.Done()
|
||||
readPipe(bufio.NewScanner(stderr))
|
||||
}()
|
||||
|
||||
err = cmd.Wait()
|
||||
wg.Wait()
|
||||
|
||||
outMu.Lock()
|
||||
out := strings.Join(lines, "\n")
|
||||
outMu.Unlock()
|
||||
|
||||
return strings.TrimSpace(out), err
|
||||
}
|
||||
|
||||
const minTrainingFeedbackCount = 5
|
||||
|
||||
const minDetectorTrainCount = 20
|
||||
@ -589,9 +723,17 @@ func trainingRunJob(root string, count int) {
|
||||
sceneOutput := ""
|
||||
|
||||
sceneScript := trainingScriptPath("train_scene_model.py")
|
||||
sceneOut, sceneErr := trainingRunCommand(
|
||||
sceneOut, sceneErr := trainingRunCommandStreaming(
|
||||
python,
|
||||
sceneScript,
|
||||
func(line string) bool {
|
||||
return trainingHandleProgressLine(
|
||||
line,
|
||||
10,
|
||||
45,
|
||||
"CLIP-Scene-Positionsmodell wird trainiert…",
|
||||
)
|
||||
},
|
||||
"--root", root,
|
||||
)
|
||||
|
||||
@ -650,9 +792,17 @@ func trainingRunJob(root string, count int) {
|
||||
})
|
||||
|
||||
detectorScript := trainingScriptPath("train_detector_model.py")
|
||||
detectorOut, detectorErr := trainingRunCommand(
|
||||
detectorOut, detectorErr := trainingRunCommandStreaming(
|
||||
python,
|
||||
detectorScript,
|
||||
func(line string) bool {
|
||||
return trainingHandleProgressLine(
|
||||
line,
|
||||
60,
|
||||
98,
|
||||
"Object Detector wird trainiert…",
|
||||
)
|
||||
},
|
||||
"--root", root,
|
||||
"--base", "yolo11n.pt",
|
||||
"--epochs", strconv.Itoa(trainingDetectorEpochs()),
|
||||
@ -772,23 +922,22 @@ func trainingStatusHandler(w http.ResponseWriter, r *http.Request) {
|
||||
return
|
||||
}
|
||||
|
||||
if err := trainingEnsureDetectorDirs(root); err != nil {
|
||||
trainingWriteError(w, http.StatusInternalServerError, err.Error())
|
||||
return
|
||||
}
|
||||
job := trainingGetJobStatus()
|
||||
|
||||
// Praktisch für kleine Datensätze:
|
||||
// Wenn genug Train-Daten existieren, aber noch zu wenig Val-Daten,
|
||||
// werden ein paar Train-Samples nach Val kopiert.
|
||||
if err := trainingEnsureDetectorValidationSample(root); err != nil {
|
||||
fmt.Println("⚠️ detector val sample ensure failed:", err)
|
||||
if !job.Running {
|
||||
if err := trainingEnsureDetectorDirs(root); err != nil {
|
||||
trainingWriteError(w, http.StatusInternalServerError, err.Error())
|
||||
return
|
||||
}
|
||||
|
||||
if err := trainingEnsureDetectorValidationSample(root); err != nil {
|
||||
fmt.Println("⚠️ detector val sample ensure failed:", err)
|
||||
}
|
||||
}
|
||||
|
||||
feedbackPath := filepath.Join(root, "feedback.jsonl")
|
||||
feedbackCount, _ := trainingCountAnnotations(feedbackPath)
|
||||
|
||||
job := trainingGetJobStatus()
|
||||
|
||||
detectorDatasetYAML := filepath.Join(root, "detector", "dataset", "dataset.yaml")
|
||||
detectorTrainImages := filepath.Join(root, "detector", "dataset", "images", "train")
|
||||
detectorTrainLabels := filepath.Join(root, "detector", "dataset", "labels", "train")
|
||||
|
||||
@ -4075,7 +4075,7 @@ export default function App() {
|
||||
) : null}
|
||||
|
||||
{selectedTab === 'training' ? (
|
||||
<TrainingTab />
|
||||
<TrainingTab onTrainingRunningChange={setTrainingTabRunning} />
|
||||
) : null}
|
||||
|
||||
{selectedTab === 'categories' ? <CategoriesTab /> : null}
|
||||
|
||||
@ -141,6 +141,131 @@ type TrainingStats = {
|
||||
}
|
||||
}
|
||||
|
||||
type TrainingNoticeKind = 'success' | 'error' | 'info' | 'warning'
|
||||
|
||||
type TrainingNotice = {
|
||||
kind: TrainingNoticeKind
|
||||
title: string
|
||||
message: string
|
||||
detail?: string
|
||||
}
|
||||
|
||||
function trainingNoticeClass(kind: TrainingNoticeKind) {
|
||||
switch (kind) {
|
||||
case 'success':
|
||||
return {
|
||||
wrap: 'border-emerald-200 bg-emerald-50 text-emerald-900 dark:border-emerald-400/30 dark:bg-emerald-500/10 dark:text-emerald-100',
|
||||
icon: 'bg-emerald-500 text-white',
|
||||
detail: 'text-emerald-800/80 dark:text-emerald-100/70',
|
||||
}
|
||||
case 'error':
|
||||
return {
|
||||
wrap: 'border-red-200 bg-red-50 text-red-900 dark:border-red-400/30 dark:bg-red-500/10 dark:text-red-100',
|
||||
icon: 'bg-red-500 text-white',
|
||||
detail: 'text-red-800/80 dark:text-red-100/70',
|
||||
}
|
||||
case 'warning':
|
||||
return {
|
||||
wrap: 'border-amber-200 bg-amber-50 text-amber-900 dark:border-amber-400/30 dark:bg-amber-500/10 dark:text-amber-100',
|
||||
icon: 'bg-amber-500 text-white',
|
||||
detail: 'text-amber-800/80 dark:text-amber-100/70',
|
||||
}
|
||||
default:
|
||||
return {
|
||||
wrap: 'border-indigo-200 bg-indigo-50 text-indigo-900 dark:border-indigo-400/30 dark:bg-indigo-500/10 dark:text-indigo-100',
|
||||
icon: 'bg-indigo-500 text-white',
|
||||
detail: 'text-indigo-800/80 dark:text-indigo-100/70',
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
function TrainingNoticeCard(props: {
|
||||
notice: TrainingNotice
|
||||
onClose?: () => void
|
||||
}) {
|
||||
const cls = trainingNoticeClass(props.notice.kind)
|
||||
|
||||
const icon =
|
||||
props.notice.kind === 'success'
|
||||
? '✓'
|
||||
: props.notice.kind === 'error'
|
||||
? '!'
|
||||
: props.notice.kind === 'warning'
|
||||
? '⚠'
|
||||
: 'i'
|
||||
|
||||
return (
|
||||
<div
|
||||
className={[
|
||||
'rounded-2xl border px-4 py-3 shadow-sm',
|
||||
cls.wrap,
|
||||
].join(' ')}
|
||||
role={props.notice.kind === 'error' ? 'alert' : 'status'}
|
||||
aria-live={props.notice.kind === 'error' ? 'assertive' : 'polite'}
|
||||
>
|
||||
<div className="flex items-start gap-3">
|
||||
<div
|
||||
className={[
|
||||
'mt-0.5 flex h-6 w-6 shrink-0 items-center justify-center rounded-full text-xs font-black',
|
||||
cls.icon,
|
||||
].join(' ')}
|
||||
aria-hidden="true"
|
||||
>
|
||||
{icon}
|
||||
</div>
|
||||
|
||||
<div className="min-w-0 flex-1">
|
||||
<div className="text-sm font-semibold">
|
||||
{props.notice.title}
|
||||
</div>
|
||||
|
||||
<div className="mt-0.5 break-words text-sm leading-relaxed">
|
||||
{props.notice.message}
|
||||
</div>
|
||||
|
||||
{props.notice.detail ? (
|
||||
<details className="mt-2">
|
||||
<summary className="cursor-pointer select-none text-xs font-medium opacity-80 hover:opacity-100">
|
||||
Details anzeigen
|
||||
</summary>
|
||||
|
||||
<pre
|
||||
className={[
|
||||
'mt-2 max-h-40 overflow-auto whitespace-pre-wrap rounded-lg bg-black/5 p-2 text-[11px] leading-relaxed dark:bg-black/20',
|
||||
cls.detail,
|
||||
].join(' ')}
|
||||
>
|
||||
{props.notice.detail}
|
||||
</pre>
|
||||
</details>
|
||||
) : null}
|
||||
</div>
|
||||
|
||||
{props.onClose ? (
|
||||
<button
|
||||
type="button"
|
||||
onClick={props.onClose}
|
||||
className="shrink-0 rounded-lg px-2 py-1 text-xs font-semibold opacity-70 transition hover:bg-black/5 hover:opacity-100 dark:hover:bg-white/10"
|
||||
aria-label="Meldung schließen"
|
||||
title="Meldung schließen"
|
||||
>
|
||||
✕
|
||||
</button>
|
||||
) : null}
|
||||
</div>
|
||||
</div>
|
||||
)
|
||||
}
|
||||
|
||||
function backendText(data: any, fallback: string) {
|
||||
return String(
|
||||
data?.message ||
|
||||
data?.error ||
|
||||
data?.detail ||
|
||||
fallback
|
||||
).trim()
|
||||
}
|
||||
|
||||
function countPercent(count: number, total: number) {
|
||||
if (!Number.isFinite(count) || !Number.isFinite(total) || total <= 0) return '0%'
|
||||
return `${Math.round((count / total) * 100)}%`
|
||||
@ -1647,7 +1772,7 @@ function TrainingStatsModal(props: {
|
||||
<div className="flex items-center justify-between gap-3">
|
||||
<div>
|
||||
<div className="text-[11px] font-medium uppercase tracking-wide text-gray-500 dark:text-gray-400">
|
||||
Gesamt-Confidence
|
||||
Daten-Confidence
|
||||
</div>
|
||||
|
||||
<div className="mt-2 text-xl font-bold text-gray-900 dark:text-white">
|
||||
@ -1673,7 +1798,7 @@ function TrainingStatsModal(props: {
|
||||
</div>
|
||||
|
||||
<div className="mt-2 text-xs leading-relaxed text-gray-500 dark:text-gray-400">
|
||||
Grober Wert aus Feedback-Menge, Boxen, Label-Abdeckung und Korrekturanteil.
|
||||
Daten-Confidence aus Feedback-Menge, Boxen, Label-Abdeckung und Korrekturanteil. Kein direkter Modell-Qualitätswert.
|
||||
</div>
|
||||
</div>
|
||||
</div>
|
||||
@ -1736,7 +1861,9 @@ function TrainingStatsModal(props: {
|
||||
)
|
||||
}
|
||||
|
||||
export default function TrainingTab() {
|
||||
export default function TrainingTab(props: {
|
||||
onTrainingRunningChange?: (running: boolean) => void
|
||||
}) {
|
||||
const [labels, setLabels] = useState<TrainingLabels>(emptyLabels)
|
||||
const [sample, setSample] = useState<TrainingSample | null>(null)
|
||||
const [correction, setCorrection] = useState<CorrectionState>(() => predictionToCorrection(null))
|
||||
@ -1882,10 +2009,14 @@ export default function TrainingTab() {
|
||||
const loadNext = useCallback(async (opts?: {
|
||||
forceNew?: boolean
|
||||
refreshPrediction?: boolean
|
||||
preserveNotice?: boolean
|
||||
}) => {
|
||||
setLoading(true)
|
||||
setError(null)
|
||||
setMessage(null)
|
||||
|
||||
if (!opts?.preserveNotice) {
|
||||
setError(null)
|
||||
setMessage(null)
|
||||
}
|
||||
|
||||
try {
|
||||
const params = new URLSearchParams()
|
||||
@ -2023,6 +2154,12 @@ export default function TrainingTab() {
|
||||
void loadTrainingStats()
|
||||
}, [statsModalOpen, loadTrainingStats])
|
||||
|
||||
const onTrainingRunningChange = props.onTrainingRunningChange
|
||||
|
||||
useEffect(() => {
|
||||
onTrainingRunningChange?.(trainingRunning)
|
||||
}, [trainingRunning, onTrainingRunningChange])
|
||||
|
||||
useEffect(() => {
|
||||
if (!boxLabel) return
|
||||
|
||||
@ -2052,7 +2189,7 @@ export default function TrainingTab() {
|
||||
const wasRunning = wasTrainingRunningRef.current
|
||||
|
||||
if (wasRunning && !trainingRunning && trainingStatus?.training?.finishedAt) {
|
||||
void loadNext({ refreshPrediction: true })
|
||||
void loadNext({ refreshPrediction: true, preserveNotice: true })
|
||||
}
|
||||
|
||||
wasTrainingRunningRef.current = trainingRunning
|
||||
@ -2125,37 +2262,16 @@ export default function TrainingTab() {
|
||||
return
|
||||
}
|
||||
|
||||
setTrainingProgress((prev) => (prev > 0 ? prev : 8))
|
||||
setTrainingStep((prev) => prev || 'Training wird vorbereitet…')
|
||||
const serverProgress = Number(trainingStatus?.training?.progress ?? 0)
|
||||
const serverStep = String(trainingStatus?.training?.step ?? '')
|
||||
|
||||
const startedAt = Date.now()
|
||||
|
||||
const timer = window.setInterval(() => {
|
||||
const elapsed = Date.now() - startedAt
|
||||
|
||||
setTrainingProgress((prev) => {
|
||||
const serverProgress = trainingStatus?.training?.progress
|
||||
if (typeof serverProgress === 'number' && serverProgress > prev) {
|
||||
return serverProgress
|
||||
}
|
||||
|
||||
if (elapsed > 90_000) {
|
||||
setTrainingStep('Detector wird trainiert…')
|
||||
return Math.min(prev + 0.4, 92)
|
||||
}
|
||||
|
||||
if (elapsed > 25_000) {
|
||||
setTrainingStep('Detector wird trainiert…')
|
||||
return Math.min(prev + 0.8, 80)
|
||||
}
|
||||
|
||||
setTrainingStep('Trainingsdaten werden verarbeitet…')
|
||||
return Math.min(prev + 1.2, 55)
|
||||
})
|
||||
}, 700)
|
||||
|
||||
return () => window.clearInterval(timer)
|
||||
}, [trainingRunning, trainingStatus?.training?.progress])
|
||||
setTrainingProgress(Number.isFinite(serverProgress) ? clampPercent(serverProgress) : 0)
|
||||
setTrainingStep(serverStep || 'Training läuft…')
|
||||
}, [
|
||||
trainingRunning,
|
||||
trainingStatus?.training?.progress,
|
||||
trainingStatus?.training?.step,
|
||||
])
|
||||
|
||||
const saveFeedback = useCallback(
|
||||
async (accepted: boolean) => {
|
||||
@ -2194,11 +2310,17 @@ export default function TrainingTab() {
|
||||
const data = await res.json().catch(() => null)
|
||||
|
||||
if (!res.ok) {
|
||||
throw new Error(data?.error || `HTTP ${res.status}`)
|
||||
throw new Error(backendText(data, `HTTP ${res.status}`))
|
||||
}
|
||||
|
||||
setMessage(
|
||||
accepted
|
||||
? 'Feedback gespeichert: Prediction wurde als korrekt übernommen.'
|
||||
: 'Korrektur gespeichert.'
|
||||
)
|
||||
|
||||
await loadTrainingStatus()
|
||||
await loadNext()
|
||||
await loadNext({ preserveNotice: true })
|
||||
} catch (e) {
|
||||
setError(e instanceof Error ? e.message : String(e))
|
||||
} finally {
|
||||
@ -2223,9 +2345,11 @@ export default function TrainingTab() {
|
||||
const data = await res.json().catch(() => null)
|
||||
|
||||
if (!res.ok) {
|
||||
throw new Error(data?.error || `HTTP ${res.status}`)
|
||||
throw new Error(backendText(data, `HTTP ${res.status}`))
|
||||
}
|
||||
|
||||
setMessage(backendText(data, 'Training wurde gestartet.'))
|
||||
|
||||
await loadTrainingStatus()
|
||||
|
||||
// WICHTIG:
|
||||
@ -2269,10 +2393,10 @@ export default function TrainingTab() {
|
||||
canTrain: false,
|
||||
})
|
||||
|
||||
setMessage(data?.message || 'Alle Trainingsdaten wurden gelöscht.')
|
||||
setMessage(backendText(data, 'Alle Trainingsdaten wurden gelöscht.'))
|
||||
|
||||
await loadTrainingStatus()
|
||||
await loadNext({ forceNew: true })
|
||||
await loadNext({ forceNew: true, preserveNotice: true })
|
||||
} catch (e) {
|
||||
setError(e instanceof Error ? e.message : String(e))
|
||||
} finally {
|
||||
@ -2527,6 +2651,38 @@ export default function TrainingTab() {
|
||||
|
||||
const showImageBoxes = !loading && !trainingRunning
|
||||
|
||||
const activeNotice = useMemo<TrainingNotice | null>(() => {
|
||||
if (error) {
|
||||
return {
|
||||
kind: 'error',
|
||||
title: 'Aktion fehlgeschlagen',
|
||||
message: error,
|
||||
}
|
||||
}
|
||||
|
||||
if (message) {
|
||||
const looksPartial =
|
||||
message.toLowerCase().includes('übersprungen') ||
|
||||
message.toLowerCase().includes('fehlgeschlagen')
|
||||
|
||||
return {
|
||||
kind: looksPartial ? 'warning' : 'success',
|
||||
title: looksPartial ? 'Training teilweise abgeschlossen' : 'Erfolg',
|
||||
message,
|
||||
}
|
||||
}
|
||||
|
||||
if (trainingRunning) {
|
||||
return {
|
||||
kind: 'info',
|
||||
title: 'Training läuft',
|
||||
message: shownTrainingStep || 'Training läuft im Hintergrund.',
|
||||
}
|
||||
}
|
||||
|
||||
return null
|
||||
}, [error, message, trainingRunning, shownTrainingStep])
|
||||
|
||||
const detectorBoxesPanel = (
|
||||
<div className="rounded-lg bg-gray-50 p-2 ring-1 ring-black/5 dark:bg-white/5 dark:ring-white/10">
|
||||
<div className="flex items-center justify-between gap-2">
|
||||
@ -2668,6 +2824,22 @@ export default function TrainingTab() {
|
||||
|
||||
return (
|
||||
<>
|
||||
{activeNotice ? (
|
||||
<div className="mb-3">
|
||||
<TrainingNoticeCard
|
||||
notice={activeNotice}
|
||||
onClose={
|
||||
trainingRunning
|
||||
? undefined
|
||||
: () => {
|
||||
setError(null)
|
||||
setMessage(null)
|
||||
}
|
||||
}
|
||||
/>
|
||||
</div>
|
||||
) : null}
|
||||
|
||||
<div className="grid grid-cols-1 items-stretch gap-3 lg:grid-cols-[300px_minmax(0,1fr)_300px] xl:grid-cols-[320px_minmax(0,1fr)_320px]">
|
||||
{/* Sidebar links */}
|
||||
<aside className="max-h-[calc(100dvh-190px)] overflow-y-auto rounded-xl border border-gray-200 bg-white p-3 shadow-sm dark:border-white/10 dark:bg-gray-900/60">
|
||||
@ -2686,8 +2858,8 @@ export default function TrainingTab() {
|
||||
'focus:outline-none focus:ring-2 focus:ring-indigo-500/40',
|
||||
'dark:bg-white/10 dark:text-gray-200 dark:ring-white/10 dark:hover:bg-indigo-500/20 dark:hover:text-indigo-100 dark:hover:ring-indigo-300/30',
|
||||
].join(' ')}
|
||||
title="Training-Statistiken anzeigen"
|
||||
aria-label="Training-Statistiken anzeigen"
|
||||
title="Training-Datenstatistiken anzeigen"
|
||||
aria-label="Training-Datenstatistiken anzeigen"
|
||||
>
|
||||
{feedbackCount}
|
||||
</button>
|
||||
@ -2779,18 +2951,6 @@ export default function TrainingTab() {
|
||||
)}
|
||||
</div>
|
||||
</div>
|
||||
|
||||
{message ? (
|
||||
<div className="mt-3 rounded-lg bg-emerald-50 px-3 py-2 text-xs text-emerald-700 dark:bg-emerald-500/10 dark:text-emerald-200">
|
||||
{message}
|
||||
</div>
|
||||
) : null}
|
||||
|
||||
{error ? (
|
||||
<div className="mt-3 rounded-lg bg-red-50 px-3 py-2 text-xs text-red-700 dark:bg-red-500/10 dark:text-red-200">
|
||||
{error}
|
||||
</div>
|
||||
) : null}
|
||||
</aside>
|
||||
|
||||
{/* Mitte */}
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user