APLoss RuntimeError

[s-rog]

  File "~/micromamba/envs/jpt/lib/python3.10/site-packages/libauc/losses/auc.py", line 313, in forward
    assert sum(pos_mask) > 0, "Input data has no positive sample! Please use 'libauc.sampler.DualSampler' for data resampling!"
RuntimeError: Boolean value of Tensor with more than one value is ambiguous

I'm trying to use this for a binary problem but have not been able to get it to work (torch 2.3).
I looked into a code but couldn't find any glaring errors as sum(Tensor) should reduce pos_mask down to a single element...

Edit: Some of my input shapes were wrong