fix yet another test

InnerEye/ML/lightning_metrics.py

@@ -8,7 +8,7 @@
 import numpy as np
 import torch
 import torch.nn.functional as F
-from pytorch_lightning import metrics
+import torchmetrics as metrics
 from pytorch_lightning.metrics import Metric
 from pytorch_lightning.metrics.functional import accuracy, auc, auroc, precision_recall_curve, roc
 from torch.nn import ModuleList
@@ -68,7 +68,7 @@ def has_predictions(self) -> bool:
  Returns True if the present object stores at least 1 prediction (self.update has been called at least once),
  or False if no predictions are stored.
  """
- return len(self.y_pred) > 0 # type: ignore
+ return self.n_obs > 0 # type: ignore
 
 
 class Accuracy05(metrics.Accuracy):