AssertionError: assert targets is not None

不知道为什么运行一个epoch之后计算mAP时会报出AssertionError: assert targets is not None的错误，但是训练过程并没有出现这个错误，只是完成一个epoch后计算AP会出现这个错误，有朋友知道吗？

iou_loss: 1.5, l1_loss: 0.5, conf_loss: 0.9, cls_loss: 0.5, lr: 2.754e-05, size: 544, ETA: 1:01:29
2021-07-25 12:51:48 | INFO | yolox.core.trainer:245 - epoch: 1/10, iter: 240/245, mem: 2040Mb, iter_time: 1.437s, data_time: 1.005s, total_loss: 3.4, iou_loss: 1.4, l1_loss: 0.6, conf_loss: 0.8, cls_loss: 0.6, lr: 2.999e-05, size: 448, ETA: 1:00:52
2021-07-25 12:51:56 | INFO | yolox.core.trainer:307 - Save weights to ./YOLOX_outputs\yolox_voc_s
0%| | 0/163 [00:02<?, ?it/s]
2021-07-25 12:51:59 | INFO | yolox.core.trainer:184 - Training of experiment is done and the best AP is 0.00
2021-07-25 12:51:59 | ERROR | yolox.core.launch:68 - An error has been caught in function 'launch', process 'MainProcess' (8292), thread 'MainThread' (20376):

File "D:\YOLOX-main\yolox\models\yolox.py", line 33, in forward
assert targets is not None
└ None

Start by reproducing the one-epoch training and AP evaluation flow, then inspect yolox/models/yolox.py at line 33 and the trainer path that invokes it during evaluation. Compare how targets are passed during training versus AP calculation. Done means evaluation completes without the assertion and reports AP.