⚡ Bolt: [performance improvement] Disable synchronous stdout logging in YOLO inference

.jules/bolt.md

@@ -0,0 +1,3 @@
+## 2024-04-12 - Disable synchronous stdout logging in Ultralytics YOLO inference
+**Learning:** Ultralytics YOLO models write detailed inference stats to stdout by default on every prediction. Synchronous logging to stdout creates blocking I/O overhead that accumulates, particularly when processing sequences of inputs (like document pages) inside loops.
+**Action:** Always set `verbose=False` in `model.predict()` (and similar model inference calls) within loop bodies or production code paths to prevent unnecessary I/O blocking overhead, improving throughput.

commonforms/inference.py

@@ -124,6 +124,7 @@ def extract_widgets(
                     conf=confidence,
                     augment=False,
                     imgsz=ONNX_IMAGE_SIZE,
+                    verbose=False, # Optimization: avoid synchronous stdout blocking overhead
                 )
                 for p in pages
             ]
@@ -135,6 +136,7 @@ def extract_widgets(
                 augment=True,
                 imgsz=image_size,
                 device=self.device,
+                verbose=False, # Optimization: avoid synchronous stdout blocking overhead
             )
 
         widgets = {}