Conduct modulated pretraining报错

[Swaggeraqua]

你好， 请问我在预训练时，遇到如下错误，我不知道如何解决，同时我也不太理解这段代码的作用。 此外，Extract vision queries起到了什么作用。对应模型的哪一部分。 Conduct modulated pretraining是训练GCP吗 我很好奇，我用自定义的数据集训练，那这个数据集在如上2个步骤都用到了吗，分别起了什么作用？ 如果我的任务是小样本目标检测，那么Modulated Training后还需要微调吗

希望您能帮忙解答，非常感谢

Traceback (most recent call last): File "/home/junyi.zhou/MQ-Det-main/tools/train_net.py", line 482, in main() File "/home/junyi.zhou/MQ-Det-main/tools/train_net.py", line 474, in main model = train(cfg=cfg, File "/home/junyi.zhou/MQ-Det-main/tools/train_net.py", line 226, in train do_train( File "/home/junyi.zhou/MQ-Det-main/maskrcnn_benchmark/engine/trainer.py", line 122, in do_train loss_dict = model(images, targets, captions=captions, positive_map=positive_map, greenlight_map = greenlight_map) File "/home/junyi.zhou/anaconda3/envs/dinov2/lib/python3.9/site-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(*args, **kwargs) File "/home/junyi.zhou/MQ-Det-main/maskrcnn_benchmark/modeling/detector/generalized_vl_rcnn_new.py", line 348, in forward batched_labels_in_caption=[t.get_field('labels_in_caption') for t in targets] File "/home/junyi.zhou/MQ-Det-main/maskrcnn_benchmark/modeling/detector/generalized_vl_rcnn_new.py", line 348, in batched_labels_in_caption=[t.get_field('labels_in_caption') for t in targets] File "/home/junyi.zhou/MQ-Det-main/maskrcnn_benchmark/structures/bounding_box.py", line 61, in get_field return self.extra_fields[field] KeyError: 'labels_in_caption'

[TinyQi]

我也遇到这个情况了，请问你怎么解决的呀？