BinaryAUROC raises IndexError with max_fpr > 0 and only true edges

[klieret]

🐛 Bug

To Reproduce

from torchmetrics.classification import BinaryAUROC
from torch import Tensor as T

m = BinaryAUROC(max_fpr=0.01)
# index error raised here
m(T([0, 0.5, 0.9]), T([1, 1, 1]))

If, in contrast, BinaryAUROC is initialized without max_fpr, you get the behavior from the next section:

Expected behavior

/Users/fuchur/micromamba/envs/gnn/lib/python3.11/site-packages/torchmetrics/utilities/prints.py:36: UserWarning: No negative samples in targets, false positive value should be meaningless. Returning zero tensor in false positive score
  warnings.warn(*args, **kwargs)
0

Actual behavior

IndexError with the following stacktrace:

File ~/micromamba/envs/gnn/lib/python3.11/site-packages/torch/nn/modules/module.py:1501, in Module._call_impl(self, *args, **kwargs)
   1496 # If we don't have any hooks, we want to skip the rest of the logic in
   1497 # this function, and just call forward.
   1498 if not (self._backward_hooks or self._backward_pre_hooks or self._forward_hooks or self._forward_pre_hooks
   1499         or _global_backward_pre_hooks or _global_backward_hooks
   1500         or _global_forward_hooks or _global_forward_pre_hooks):
-> 1501     return forward_call(*args, **kwargs)
   1502 # Do not call functions when jit is used
   1503 full_backward_hooks, non_full_backward_hooks = [], []

File ~/micromamba/envs/gnn/lib/python3.11/site-packages/torchmetrics/metric.py:236, in Metric.forward(self, *args, **kwargs)
    234     self._forward_cache = self._forward_full_state_update(*args, **kwargs)
    235 else:
--> 236     self._forward_cache = self._forward_reduce_state_update(*args, **kwargs)
    238 return self._forward_cache

File ~/micromamba/envs/gnn/lib/python3.11/site-packages/torchmetrics/metric.py:303, in Metric._forward_reduce_state_update(self, *args, **kwargs)
    301 # calculate batch state and compute batch value
    302 self.update(*args, **kwargs)
--> 303 batch_val = self.compute()
    305 # reduce batch and global state
    306 self._update_count = _update_count + 1

File ~/micromamba/envs/gnn/lib/python3.11/site-packages/torchmetrics/metric.py:532, in Metric._wrap_compute.<locals>.wrapped_func(*args, **kwargs)
    524 # compute relies on the sync context manager to gather the states across processes and apply reduction
    525 # if synchronization happened, the current rank accumulated states will be restored to keep
    526 # accumulation going if ``should_unsync=True``,
    527 with self.sync_context(
    528     dist_sync_fn=self.dist_sync_fn,
    529     should_sync=self._to_sync,
    530     should_unsync=self._should_unsync,
    531 ):
--> 532     value = compute(*args, **kwargs)
    533     self._computed = _squeeze_if_scalar(value)
    535 return self._computed

File ~/micromamba/envs/gnn/lib/python3.11/site-packages/torchmetrics/classification/auroc.py:114, in BinaryAUROC.compute(self)
    112 else:
    113     state = self.confmat
--> 114 return _binary_auroc_compute(state, self.thresholds, self.max_fpr)

File ~/micromamba/envs/gnn/lib/python3.11/site-packages/torchmetrics/functional/classification/auroc.py:97, in _binary_auroc_compute(state, thresholds, max_fpr, pos_label)
     95 # Add a single point at max_fpr and interpolate its tpr value
     96 stop = torch.bucketize(max_area, fpr, out_int32=True, right=True)
---> 97 weight = (max_area - fpr[stop - 1]) / (fpr[stop] - fpr[stop - 1])
     98 interp_tpr: Tensor = torch.lerp(tpr[stop - 1], tpr[stop], weight)
     99 tpr = torch.cat([tpr[:stop], interp_tpr.view(1)])

IndexError: index 4 is out of bounds for dimension 0 with size 4

Environment

• TorchMetrics version (and how you installed TM, e.g. conda, pip, build from source): pip-installed 1.0.0
• Python & PyTorch Version (e.g., 1.0): 3.11.3, 2.0.1
• Any other relevant information such as OS (e.g., Linux): OSX

Additional context

I'd be happy to work on a fix.

[github-actions]

Hi! thanks for your contribution!, great first issue!