Some Problems when Trying to Train on our own datasets

[ZhangJiayuan-BUAA]

Hi, thank you for your great work. When i was trying to use your model to train my dataset, an error occurred. Just after printing start training, the field class_error in variable data_loader is null.
main_track.py train_stats = train_one_epoch(model, criterion, data_loader_train, optimizer, device, epoch, args.clip_max_norm)