Make num_classes optional, in case of micro averaging

src/torchmetrics/classification/stat_scores.py

@@ -307,7 +307,7 @@ class MulticlassStatScores(_AbstractStatScores):
 
     def __init__(
         self,
-        num_classes: int,
+        num_classes: Optional[int] = None,
         top_k: int = 1,
         average: Optional[Literal["micro", "macro", "weighted", "none"]] = "macro",
         multidim_average: Literal["global", "samplewise"] = "global",

src/torchmetrics/functional/classification/accuracy.py

@@ -166,7 +166,7 @@ def binary_accuracy(
 def multiclass_accuracy(
     preds: Tensor,
     target: Tensor,
-    num_classes: int,
+    num_classes: Optional[int] = None,
     average: Optional[Literal["micro", "macro", "weighted", "none"]] = "macro",
     top_k: int = 1,
     multidim_average: Literal["global", "samplewise"] = "global",

src/torchmetrics/functional/classification/stat_scores.py

@@ -311,15 +311,15 @@ def _multiclass_stat_scores_tensor_validation(
             "Either `preds` and `target` both should have the (same) shape (N, ...), or `target` should be (N, ...)"
             " and `preds` should be (N, C, ...)."
         )
-
-    check_value = num_classes if ignore_index is None else num_classes + 1
-    for t, name in ((target, "target"),) + ((preds, "preds"),) if not preds.is_floating_point() else ():  # noqa: RUF005
-        num_unique_values = len(torch.unique(t, dim=None))
-        if num_unique_values > check_value:
-            raise RuntimeError(
-                f"Detected more unique values in `{name}` than expected. Expected only {check_value} but found"
-                f" {num_unique_values} in `{name}`. Found values: {torch.unique(t, dim=None)}."
-            )
+    if num_classes is not None:
+        check_value = num_classes if ignore_index is None else num_classes + 1
+        for t, name in ((target, "target"),) + ((preds, "preds"),) if not preds.is_floating_point() else ():  # noqa: RUF005
+            num_unique_values = len(torch.unique(t, dim=None))
+            if num_unique_values > check_value:
+                raise RuntimeError(
+                    f"Detected more unique values in `{name}` than expected. Expected only {check_value} but found"
+                    f" {num_unique_values} in `{name}`. Found values: {torch.unique(t, dim=None)}."
+                )
 
 
 def _multiclass_stat_scores_format(

tests/unittests/classification/test_accuracy.py

@@ -625,6 +625,12 @@ def test_corner_cases():
     res = metric(preds, target)
     assert res == 1.0
 
+    metric_micro1 = MulticlassAccuracy(num_classes=None, average="micro", ignore_index=0)
+    metric_micro2 = MulticlassAccuracy(num_classes=3, average="micro", ignore_index=0)
+    res1 = metric_micro1(preds, target)
+    res2 = metric_micro2(preds, target)
+    assert res1 == res2
+
 
 @pytest.mark.parametrize(
     ("metric", "kwargs"),