在A910上跑迁移后的模型出现以下问题：

number of params: 41302368 

loading annotations into memory... 

Done (t=49.54s) 

creating index... 

index created! 

loading annotations into memory... 

Done (t=1.83s) 

creating index... 

index created! 

Start training 

torch.Size([200, 92]) 

tensor([32,  1, 85,  1,  1,  1,  1, 51, 48, 50, 50, 79,  1, 51, 50], 

       device='npu:0') 

Segmentation fault (core dumped) 

[ma-user detr-main]$/home/ma-user/miniconda3/envs/py38_pytorch/lib/python3.8/multiprocessing/resource_tracker.py:216: UserWarning: resource_tracker: There appear to be 91 leaked semaphore objects to clean up at shutdown 

  warnings.warn('resource_tracker: There appear to be %d '

有没有大佬指导一下怎么解决