后者才是测试数据的评估效果吗？前者是训练数据的？

如果是的话 为什么训练数据集和测试数据集差这么多呢？

resume from ./output/ch_db_mv3/best_accuracy
[2022/03/30 11:22:48] root INFO: metric in ckpt ***************
[2022/03/30 11:22:48] root INFO: hmean:0.9468599033816426
[2022/03/30 11:22:48] root INFO: precision:0.9245283018867925
[2022/03/30 11:22:48] root INFO: recall:0.9702970297029703
[2022/03/30 11:22:48] root INFO: Student2_precision:0.9038461538461539
[2022/03/30 11:22:48] root INFO: Student2_recall:0.9306930693069307
[2022/03/30 11:22:48] root INFO: Student2_hmean:0.9170731707317074
[2022/03/30 11:22:48] root INFO: Teacher_precision:0.6451612903225806
[2022/03/30 11:22:48] root INFO: Teacher_recall:0.7920792079207921
[2022/03/30 11:22:48] root INFO: Teacher_hmean:0.711111111111111
[2022/03/30 11:22:48] root INFO: fps:14.090875190233184
[2022/03/30 11:22:48] root INFO: best_epoch:220
[2022/03/30 11:22:48] root INFO: start_epoch:221
eval model:: 100%|██████████████████████████████| 51/51 [00:04<00:00, 10.71it/s]
[2022/03/30 11:22:53] root INFO: metric eval ***************
[2022/03/30 11:22:53] root INFO: precision:0.8372093023255814
[2022/03/30 11:22:53] root INFO: recall:0.3564356435643564
[2022/03/30 11:22:53] root INFO: hmean:0.5
[2022/03/30 11:22:53] root INFO: Student2_precision:0.813953488372093
[2022/03/30 11:22:53] root INFO: Student2_recall:0.3465346534653465
[2022/03/30 11:22:53] root INFO: Student2_hmean:0.48611111111111105
[2022/03/30 11:22:53] root INFO: Teacher_precision:0.6333333333333333
[2022/03/30 11:22:53] root INFO: Teacher_recall:0.18811881188118812
[2022/03/30 11:22:53] root INFO: Teacher_hmean:0.2900763358778626
[2022/03/30 11:22:53] root INFO: fps:15.879676743045582