feat(training): release V0.8 自调参 Agent

training_server/tuning/tensorboard.py

@@ -0,0 +1,35 @@
+"""TensorBoard scalar ingestion with optional dependency isolation."""
+
+from __future__ import annotations
+
+from pathlib import Path
+
+from .storage import TuningStorage
+
+
+class TensorboardUnavailable(RuntimeError):
+    pass
+
+
+def ingest_scalars(storage: TuningStorage, trial_id: str, log_dir: Path) -> int:
+    """Reload all scalar events and idempotently upsert them into SQLite."""
+    try:
+        from tensorboard.backend.event_processing.event_accumulator import EventAccumulator
+    except ImportError as error:
+        raise TensorboardUnavailable(
+            "缺少 tensorboard，请安装 training_server/requirements.txt"
+        ) from error
+    if not log_dir.is_dir():
+        return 0
+    accumulator = EventAccumulator(str(log_dir), size_guidance={"scalars": 0})
+    try:
+        accumulator.Reload()
+    except (OSError, ValueError):
+        return 0
+    points: list[tuple[str, int, float, float]] = []
+    for tag in accumulator.Tags().get("scalars", []):
+        for event in accumulator.Scalars(tag):
+            points.append((tag, int(event.step), float(event.wall_time), float(event.value)))
+    if points:
+        storage.insert_metrics(trial_id, points)
+    return len(points)