RuntimeError

Question

RuntimeError

liuzt1024 opened this issue 3 years ago · comments

_liuzt commented 3 years ago

RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:2 and cuda:0!

jianpursuit · Answer 1 · Thu Oct 28 2021 16:08:58 GMT+0800 (China Standard Time)

We didn't encounter the similar problem, can you provide more information about this RuntimeError?

_liuzt · Answer 2 · Fri Oct 29 2021 14:09:15 GMT+0800 (China Standard Time)

after this function：tood_head.py
losses_cls, losses_bbox,
cls_avg_factors, bbox_avg_factors = multi_apply(self.loss_single)
print(f'bbox_avg_factors : {bbox_avg_factors}')
get result:
bbox_avg_factors : [tensor(49.6258, device='cuda:3'), tensor(14.4483, device='cuda:3'), tensor(0, device='cuda:0'), tensor(0, device='cuda:0'), tensor(0, device='cuda:0')]

_liuzt · Answer 3 · Fri Oct 29 2021 14:15:33 GMT+0800 (China Standard Time)

I use gpu3 for training
cls_avg_factors : [tensor(0., device='cuda:3'), tensor(0., device='cuda:3'), tensor(9., device='cuda:3'), tensor(9., device='cuda:3'), tensor(0., device='cuda:3')]
bbox_avg_factors : [tensor(0, device='cuda:0'), tensor(0, device='cuda:0'), tensor(6.4019, device='cuda:3'), tensor(6.3117, device='cuda:3'), tensor(0, device='cuda:0')]