mypy

InnerEye/ML/lightning_metrics.py

@@ -162,9 +162,9 @@ def _get_metrics_at_optimal_cutoff(self) -> Tuple[torch.Tensor, torch.Tensor, to
  return torch.tensor(np.nan), torch.tensor(np.nan), torch.tensor(np.nan), torch.tensor(np.nan)
  fpr, tpr, thresholds = roc(preds, targets)
  optimal_idx = torch.argmax(tpr - fpr) # type: ignore
- optimal_threshold = thresholds[optimal_idx]
+ optimal_threshold = thresholds[optimal_idx] # type: ignore
  acc = accuracy(preds > optimal_threshold, targets)
- false_negative_optimal = 1 - tpr[optimal_idx]
+ false_negative_optimal = 1 - tpr[optimal_idx] # type: ignore
  false_positive_optimal = fpr[optimal_idx]
  return optimal_threshold, false_positive_optimal, false_negative_optimal, acc
 

InnerEye/ML/lightning_models.py

@@ -349,7 +349,7 @@ def training_or_validation_epoch_end(self, is_training: bool) -> None:
  logger.flush()
  super().training_or_validation_epoch_end(is_training)
 
- def transfer_batch_to_device(self, batch: Any, device: Optional[torch.device]) -> Any:
+ def transfer_batch_to_device(self, batch: Any, device: torch.device) -> Any: # type: ignore
  """
  For sequence models, transfer the nested lists of items to the given GPU device.
  For all other models, this relies on the superclass to move the batch of data to the GPU.