torch.multiprocessing.spawn.ProcessRaisedException about upt HOT 2 CLOSED

Hi @yaoyaosanqi,

The model returns a dict containing the losses during training, while during test it returns a list of detections. The error loss_dict['interaction_loss'].isnan(): TypeError: list indices must be integers or slices, not str suggests a list has been returned during training.

I need to know what changes you made to the code to find the problem. Run git status to find scripts that have been modified and look into the changes.

Fred.

from upt.

Following your suggestion, I checked the code and found that I commented out the training part when I was debugging before.

The issue was caused by my negligence. Already solved, thank you very much!

#if self.training:##$$targets:4list,只经过数据预处理,一个list一张图片 # interaction_loss = self.compute_interaction_loss(boxes, bh, bo, logits, prior, targets) # loss_dict = dict( # interaction_loss=interaction_loss # ) # return loss_dict

from upt.