Created by: Cristhine

训练网络的过程中,可以正常训练部分batch 结果如下： Loaded 59468 images after run Iter 0, lr 0.000000, smoke_acc 0.0488, phone_acc 0.0413, safe_acc 0.0919, hand_acc 0.0653, forward_acc 0.0428, average loss 1047.4664, smooth_loss 10474.6641, cur_loss 10474.6641, yolo_loss 1047.2461, rcnn_cls_loss 0.1949, rcnn_box_loss 0.0251, smoke_loss 0.0000, phone_loss 0.0000, safe_loss 0.0003, hand_loss 0.0000, forward_loss 0.0000, average time: 4.63962554932e-05 after run after run after run after run after run after run after run after run after run after run Iter 10, lr 0.000020, smoke_acc 0.1884, phone_acc 0.2613, safe_acc 0.6603, hand_acc 0.5488, forward_acc 0.3850, average loss 3413.7736, smooth_loss 4055.6727, cur_loss 32.6192, yolo_loss 3411.5710, rcnn_cls_loss 1.9221, rcnn_box_loss 0.2540, smoke_loss 0.0058, phone_loss 0.0058, safe_loss 0.0052, hand_loss 0.0087, forward_loss 0.0010, average time: 5.36966350079 after run after run after run after run after run 然后在某一个batch就不往下训练了，没有报错信息，一直停在这里，定位在reader.py函数里是，一直在读取数据；