针对视觉模型（YOLO、ViTs、RT-DETR、DINOv3）的一站式训练：预训练、微调、蒸馏。
All-in-one training for vision models (YOLO, ViTs, RT-DETR, DINOv3): pretraining, fine-tuning, distillation.
