Add torch.inference mode
@@ -64,6 +64,7 @@ def get_yolo_model(model_path):
return model
@torch.inference_mode()
def get_parsed_content_icon(filtered_boxes, ocr_bbox, image_source, caption_model_processor, prompt=None):
to_pil = ToPILImage()
if ocr_bbox:
The note is not visible to the blocked user.