Training error on single GPU

[happyencoding]

Hi, first of all this is excellent work! I am trying to train some data on the segmentation portion.
When I training using the command“ python main.py --backbone mit_b3 -c exp_name --lamda 1e-6 --gpu 0 ” on single gpu on the
nyudv2 dataset. But have an error always.The error in the screenshot below.

[yikaiw]

Seems not an error related to the method. Could you check this part in CPU and output its value, by print((sum([torch.abs(m).sum() for m in mask])?

[happyencoding]

Thank you for your attention,I will check it.