fixes

src/torchmetrics/aggregation.py

@@ -77,9 +77,9 @@ def _cast_and_nan_check_input(
     ) -> Tuple[Tensor, Tensor]:
         """Convert input ``x`` to a tensor and check for Nans."""
         if not isinstance(x, Tensor):
-            x = torch.as_tensor(x, dtype=torch.float32, device=self.device)
+            x = torch.as_tensor(x, dtype=self.dtype, device=self.device)
         if weight is not None and not isinstance(weight, Tensor):
-            weight = torch.as_tensor(weight, dtype=torch.float32, device=self.device)
+            weight = torch.as_tensor(weight, dtype=self.dtype, device=self.device)
 
         nans = torch.isnan(x)
         if weight is not None:
@@ -101,7 +101,7 @@ def _cast_and_nan_check_input(
                 x[nans | nans_weight] = self.nan_strategy
                 weight[nans | nans_weight] = self.nan_strategy
 
-        return x.float(), weight.float()
+        return x.to(self.dtype), weight.to(self.dtype)
 
     def update(self, value: Union[float, Tensor]) -> None:
         """Overwrite in child class."""
@@ -557,9 +557,9 @@ def update(self, value: Union[float, Tensor], weight: Union[float, Tensor] = 1.0
         """
         # broadcast weight to value shape
         if not isinstance(value, Tensor):
-            value = torch.as_tensor(value, dtype=torch.float32, device=self.device)
+            value = torch.as_tensor(value, dtype=self.dtype, device=self.device)
         if weight is not None and not isinstance(weight, Tensor):
-            weight = torch.as_tensor(weight, dtype=torch.float32, device=self.device)
+            weight = torch.as_tensor(weight, dtype=self.dtype, device=self.device)
         weight = torch.broadcast_to(weight, value.shape)
         value, weight = self._cast_and_nan_check_input(value, weight)
 

tests/unittests/bases/test_aggregation.py

@@ -214,10 +214,11 @@ def test_with_default_dtype(metric_class, compare_function):
     """Test that the metric works with a default dtype of float64."""
     torch.set_default_dtype(torch.float64)
     metric = metric_class()
+    assert metric.dtype == torch.float64
     values = torch.randn(10000)
     metric.update(values)
     result = metric.compute()
     assert result.dtype == torch.float64
     assert result.dtype == values.dtype
-    assert torch.allclose(result, compare_function(values), atol=1e-12)
+    assert result == compare_function(values)
     torch.set_default_dtype(torch.float32)

tests/unittests/bases/test_metric.py

@@ -573,3 +573,19 @@ def test_update_properties(metric, method):
     m.reset()
     assert not m.update_called
     assert m.update_count == 0
+
+
+def test_dtype_property():
+    """Test that dtype property works as expected."""
+    metric = DummyMetricSum()
+    assert metric.dtype == torch.float32
+    metric.set_dtype(torch.float64)
+    assert metric.dtype == torch.float64
+
+    torch.set_default_dtype(torch.float64)
+    metric = DummyMetricSum()
+    assert metric.dtype == torch.float64
+    torch.set_default_dtype(torch.float32)
+    assert metric.dtype == torch.float64  # should not change after initialization
+    metric.set_dtype(torch.float32)
+    assert metric.dtype == torch.float32