Remove sigmoid in BinaryPrecisionRecallCurve

CHANGELOG.md

@@ -54,7 +54,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
--
+- Fixed unintended `sigmoid` normalization in `BinaryPrecisionRecallCurve`, which might cause incorrect ROC computation when updating incrementally with raw logits. ([#3179](https://github.com/Lightning-AI/torchmetrics/pull/3179))
 
 
 ---

src/torchmetrics/classification/precision_recall_curve.py

@@ -163,7 +163,9 @@ def update(self, preds: Tensor, target: Tensor) -> None:
         """Update metric states."""
         if self.validate_args:
             _binary_precision_recall_curve_tensor_validation(preds, target, self.ignore_index)
-        preds, target, _ = _binary_precision_recall_curve_format(preds, target, self.thresholds, self.ignore_index)
+        preds, target, _ = _binary_precision_recall_curve_format(
+            preds, target, self.thresholds, self.ignore_index, None
+        )
         state = _binary_precision_recall_curve_update(preds, target, self.thresholds)
         if isinstance(state, Tensor):
             self.confmat += state

src/torchmetrics/functional/classification/precision_recall_curve.py

@@ -166,6 +166,7 @@ def _binary_precision_recall_curve_format(
     target: Tensor,
     thresholds: Optional[Union[int, list[float], Tensor]] = None,
     ignore_index: Optional[int] = None,
+    normalization: Optional[str] = "sigmoid",
 ) -> tuple[Tensor, Tensor, Optional[Tensor]]:
     """Convert all input to the right format.
 
@@ -182,7 +183,7 @@ def _binary_precision_recall_curve_format(
         preds = preds[idx]
         target = target[idx]
 
-    preds = normalize_logits_if_needed(preds, "sigmoid")
+    preds = normalize_logits_if_needed(preds, normalization)
 
     thresholds = _adjust_threshold_arg(thresholds, preds.device)
     return preds, target, thresholds

src/torchmetrics/utilities/compute.py

@@ -201,7 +201,7 @@ def interp(x: Tensor, xp: Tensor, fp: Tensor) -> Tensor:
     return m[indices] * x + b[indices]
 
 
-def normalize_logits_if_needed(tensor: Tensor, normalization: Literal["sigmoid", "softmax"]) -> Tensor:
+def normalize_logits_if_needed(tensor: Tensor, normalization: Optional[Literal["sigmoid", "softmax"]]) -> Tensor:
     """Normalize logits if needed.
 
     If input tensor is outside the [0,1] we assume that logits are provided and apply the normalization.
@@ -228,6 +228,9 @@ def normalize_logits_if_needed(tensor: Tensor, normalization: Literal["sigmoid",
         tensor([0.0000, 0.5000, 1.0000])
 
     """
+    # if not specified, do nothing.
+    if not normalization:
+        return tensor
     # decrease sigmoid on cpu .
     if tensor.device == torch.device("cpu"):
         if not torch.all((tensor >= 0) * (tensor <= 1)):