diff --git a/CHANGELOG.md b/CHANGELOG.md
index 57762a7bd6d..b2f065d7b36 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -22,6 +22,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Deprecated
 
+- Deprecated method `compute_on_step` ([#792](https://github.com/PyTorchLightning/metrics/pull/792))
+
 
 ### Removed
 
diff --git a/docs/source/pages/implement.rst b/docs/source/pages/implement.rst
index 417b73edac6..35d3829082e 100644
--- a/docs/source/pages/implement.rst
+++ b/docs/source/pages/implement.rst
@@ -64,7 +64,7 @@ The cache is first emptied on the next call to ``update``.
 
 ``forward`` serves the dual purpose of both returning the metric on the current data and updating the internal
 metric state for accumulating over multiple batches. The ``forward()`` method achieves this by combining calls
-to ``update`` and ``compute`` in the following way (assuming metric is initialized with ``compute_on_step=True``):
+to ``update`` and ``compute`` in the following way:
 
 1. Calls ``update()`` to update the global metric state (for accumulation over multiple batches)
 2. Caches the global state.
diff --git a/docs/source/pages/overview.rst b/docs/source/pages/overview.rst
index 60820544d57..5cbe94023b1 100644
--- a/docs/source/pages/overview.rst
+++ b/docs/source/pages/overview.rst
@@ -27,7 +27,7 @@ This metrics API is independent of PyTorch Lightning. Metrics can directly be us
     from torchmetrics.classification import Accuracy
 
     train_accuracy = Accuracy()
-    valid_accuracy = Accuracy(compute_on_step=False)
+    valid_accuracy = Accuracy()
 
     for epoch in range(epochs):
         for x, y in train_data:
@@ -35,16 +35,24 @@ This metrics API is independent of PyTorch Lightning. Metrics can directly be us
 
             # training step accuracy
             batch_acc = train_accuracy(y_hat, y)
+            print(f"Accuracy of batch{i} is {batch_acc}")
 
         for x, y in valid_data:
             y_hat = model(x)
-            valid_accuracy(y_hat, y)
+            valid_accuracy.update(y_hat, y)
 
-    # total accuracy over all training batches
-    total_train_accuracy = train_accuracy.compute()
+        # total accuracy over all training batches
+        total_train_accuracy = train_accuracy.compute()
 
-    # total accuracy over all validation batches
-    total_valid_accuracy = valid_accuracy.compute()
+        # total accuracy over all validation batches
+        total_valid_accuracy = valid_accuracy.compute()
+
+        print(f"Training acc for epoch {epoch}: {total_train_accuracy}")
+        print(f"Validation acc for epoch {epoch}: {total_valid_accuracy}")
+
+        # Reset metric states after each epoch
+        train_accuracy.reset()
+        valid_accuracy.reset()
 
 .. note::
 
diff --git a/tests/bases/test_composition.py b/tests/bases/test_composition.py
index 36aec46b817..36516c9997a 100644
--- a/tests/bases/test_composition.py
+++ b/tests/bases/test_composition.py
@@ -553,26 +553,3 @@ def test_compositional_metrics_update():
 
     assert compos.metric_a._num_updates == 3
     assert compos.metric_b._num_updates == 3
-
-
-@pytest.mark.parametrize("compute_on_step", [True, False])
-@pytest.mark.parametrize("metric_b", [4, DummyMetric(4)])
-def test_compositional_metrics_forward(compute_on_step, metric_b):
-    """test forward method of compositional metrics."""
-    metric_a = DummyMetric(5)
-    metric_a.compute_on_step = compute_on_step
-    compos = metric_a + metric_b
-
-    assert isinstance(compos, CompositionalMetric)
-    for _ in range(3):
-        val = compos()
-        assert val == 9 if compute_on_step else val is None
-
-    assert isinstance(compos.metric_a, DummyMetric)
-    assert compos.metric_a._num_updates == 3
-
-    if isinstance(metric_b, DummyMetric):
-        assert isinstance(compos.metric_b, DummyMetric)
-        assert compos.metric_b._num_updates == 3
-
-    compos.reset()
diff --git a/tests/wrappers/test_multioutput.py b/tests/wrappers/test_multioutput.py
index 35ab90af092..78521c3837f 100644
--- a/tests/wrappers/test_multioutput.py
+++ b/tests/wrappers/test_multioutput.py
@@ -24,7 +24,7 @@ def __init__(
         self,
         base_metric_class,
         num_outputs: int = 1,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Any = None,
         dist_sync_fn: Optional[Callable] = None,
diff --git a/torchmetrics/aggregation.py b/torchmetrics/aggregation.py
index 3b586748471..15bd1268f13 100644
--- a/torchmetrics/aggregation.py
+++ b/torchmetrics/aggregation.py
@@ -34,8 +34,11 @@ class BaseAggregator(Metric):
             - a float: if a float is provided will impude any `nan` values with this value
 
         compute_on_step:
-            Forward only calls ``update()`` and returns None if this is
-            set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -59,7 +62,7 @@ def __init__(
         fn: Union[Callable, str],
         default_value: Union[Tensor, List],
         nan_strategy: Union[str, float] = "error",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
@@ -120,8 +123,11 @@ class MaxMetric(BaseAggregator):
             - a float: if a float is provided will impude any `nan` values with this value
 
         compute_on_step:
-            Forward only calls ``update()`` and returns None if this is
-            set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -147,7 +153,7 @@ class MaxMetric(BaseAggregator):
     def __init__(
         self,
         nan_strategy: Union[str, float] = "warn",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
@@ -185,8 +191,11 @@ class MinMetric(BaseAggregator):
             - a float: if a float is provided will impude any `nan` values with this value
 
         compute_on_step:
-            Forward only calls ``update()`` and returns None if this is
-            set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -212,7 +221,7 @@ class MinMetric(BaseAggregator):
     def __init__(
         self,
         nan_strategy: Union[str, float] = "warn",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
@@ -250,8 +259,11 @@ class SumMetric(BaseAggregator):
             - a float: if a float is provided will impude any `nan` values with this value
 
         compute_on_step:
-            Forward only calls ``update()`` and returns None if this is
-            set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -277,7 +289,7 @@ class SumMetric(BaseAggregator):
     def __init__(
         self,
         nan_strategy: Union[str, float] = "warn",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
@@ -308,8 +320,11 @@ class CatMetric(BaseAggregator):
             - a float: if a float is provided will impude any `nan` values with this value
 
         compute_on_step:
-            Forward only calls ``update()`` and returns None if this is
-            set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -335,7 +350,7 @@ class CatMetric(BaseAggregator):
     def __init__(
         self,
         nan_strategy: Union[str, float] = "warn",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
@@ -371,8 +386,11 @@ class MeanMetric(BaseAggregator):
             - a float: if a float is provided will impude any `nan` values with this value
 
         compute_on_step:
-            Forward only calls ``update()`` and returns None if this is
-            set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -398,7 +416,7 @@ class MeanMetric(BaseAggregator):
     def __init__(
         self,
         nan_strategy: Union[str, float] = "warn",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/audio/pesq.py b/torchmetrics/audio/pesq.py
index 6f977798469..5a980a76141 100644
--- a/torchmetrics/audio/pesq.py
+++ b/torchmetrics/audio/pesq.py
@@ -44,7 +44,11 @@ class PerceptualEvaluationSpeechQuality(Metric):
         keep_same_device:
             whether to move the pesq value to the device of preds
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
@@ -89,7 +93,7 @@ def __init__(
         self,
         fs: int,
         mode: str,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable[[Tensor], Tensor]] = None,
diff --git a/torchmetrics/audio/pit.py b/torchmetrics/audio/pit.py
index e69a078ccd7..69f247c38dc 100644
--- a/torchmetrics/audio/pit.py
+++ b/torchmetrics/audio/pit.py
@@ -39,6 +39,10 @@ class PermutationInvariantTraining(Metric):
             or the larger the better.
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -78,7 +82,7 @@ def __init__(
         self,
         metric_func: Callable,
         eval_func: str = "max",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable[[Tensor], Tensor]] = None,
diff --git a/torchmetrics/audio/sdr.py b/torchmetrics/audio/sdr.py
index feb47a49cb4..ddfe41e2b54 100644
--- a/torchmetrics/audio/sdr.py
+++ b/torchmetrics/audio/sdr.py
@@ -49,6 +49,10 @@ class SignalDistortionRatio(Metric):
             signals may sometimes be zero
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -110,7 +114,7 @@ def __init__(
         filter_length: int = 512,
         zero_mean: bool = False,
         load_diag: Optional[float] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable[[Tensor], Tensor]] = None,
@@ -168,6 +172,10 @@ class ScaleInvariantSignalDistortionRatio(Metric):
             if to zero mean target and preds or not
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -206,7 +214,7 @@ class ScaleInvariantSignalDistortionRatio(Metric):
     def __init__(
         self,
         zero_mean: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable[[Tensor], Tensor]] = None,
diff --git a/torchmetrics/audio/snr.py b/torchmetrics/audio/snr.py
index d31613e4348..ffdc39d1fed 100644
--- a/torchmetrics/audio/snr.py
+++ b/torchmetrics/audio/snr.py
@@ -39,6 +39,10 @@ class SignalNoiseRatio(Metric):
             if to zero mean target and preds or not
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -77,7 +81,7 @@ class SignalNoiseRatio(Metric):
     def __init__(
         self,
         zero_mean: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable[[Tensor], Tensor]] = None,
@@ -121,6 +125,10 @@ class ScaleInvariantSignalNoiseRatio(Metric):
     Args:
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -159,7 +167,7 @@ class ScaleInvariantSignalNoiseRatio(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable[[Tensor], Tensor]] = None,
diff --git a/torchmetrics/audio/stoi.py b/torchmetrics/audio/stoi.py
index 94a92bbeaa0..a7327554c8e 100644
--- a/torchmetrics/audio/stoi.py
+++ b/torchmetrics/audio/stoi.py
@@ -48,6 +48,10 @@ class ShortTimeObjectiveIntelligibility(Metric):
             whether to use the extended STOI described in [4]
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -96,7 +100,7 @@ def __init__(
         self,
         fs: int,
         extended: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable[[Tensor], Tensor]] = None,
diff --git a/torchmetrics/classification/accuracy.py b/torchmetrics/classification/accuracy.py
index b6dec6f8036..8d64862bafe 100644
--- a/torchmetrics/classification/accuracy.py
+++ b/torchmetrics/classification/accuracy.py
@@ -127,7 +127,11 @@ class Accuracy(StatScores):
               still applies in both cases, if set.
 
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
@@ -179,7 +183,7 @@ def __init__(
         top_k: Optional[int] = None,
         multiclass: Optional[bool] = None,
         subset_accuracy: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/auc.py b/torchmetrics/classification/auc.py
index a10fddbe98d..d6b52945721 100644
--- a/torchmetrics/classification/auc.py
+++ b/torchmetrics/classification/auc.py
@@ -33,7 +33,11 @@ class AUC(Metric):
             setting this argument to ``True`` will use a stable sorting algorithm to
             sort the input in descending order
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -50,7 +54,7 @@ class AUC(Metric):
     def __init__(
         self,
         reorder: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/auroc.py b/torchmetrics/classification/auroc.py
index 29fd15a44b8..9b8c13f3b98 100644
--- a/torchmetrics/classification/auroc.py
+++ b/torchmetrics/classification/auroc.py
@@ -47,32 +47,35 @@ class AUROC(Metric):
         with an warning.
 
     Args:
-       num_classes: integer with number of classes for multi-label and multiclass problems.
-           Should be set to ``None`` for binary problems
-       pos_label: integer determining the positive class. Default is ``None``
-           which for binary problem is translate to 1. For multiclass problems
-           this argument should not be set as we iteratively change it in the
-           range [0,num_classes-1]
-       average:
-           - ``'micro'`` computes metric globally. Only works for multilabel problems
-           - ``'macro'`` computes metric for each class and uniformly averages them
-           - ``'weighted'`` computes metric for each class and does a weighted-average,
-             where each class is weighted by their support (accounts for class imbalance)
-           - ``None`` computes and returns the metric per class
-       max_fpr:
-           If not ``None``, calculates standardized partial AUC over the
-           range [0, max_fpr]. Should be a float between 0 and 1.
-       compute_on_step:
-           Forward only calls ``update()`` and return None if this is set to False.
-       dist_sync_on_step:
-           Synchronize metric state across processes at each ``forward()``
-           before returning the value at the step.
-       process_group:
-           Specify the process group on which synchronization is called.
-
-       dist_sync_fn:
-           Callback that performs the allgather operation on the metric state. When ``None``, DDP
-           will be used to perform the allgather
+        num_classes: integer with number of classes for multi-label and multiclass problems.
+            Should be set to ``None`` for binary problems
+        pos_label: integer determining the positive class. Default is ``None``
+            which for binary problem is translate to 1. For multiclass problems
+            this argument should not be set as we iteratively change it in the
+            range [0,num_classes-1]
+        average:
+            - ``'micro'`` computes metric globally. Only works for multilabel problems
+            - ``'macro'`` computes metric for each class and uniformly averages them
+            - ``'weighted'`` computes metric for each class and does a weighted-average,
+              where each class is weighted by their support (accounts for class imbalance)
+            - ``None`` computes and returns the metric per class
+        max_fpr:
+            If not ``None``, calculates standardized partial AUC over the
+            range [0, max_fpr]. Should be a float between 0 and 1.
+        compute_on_step:
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
+        dist_sync_on_step:
+            Synchronize metric state across processes at each ``forward()``
+            before returning the value at the step.
+        process_group:
+            Specify the process group on which synchronization is called.
+        dist_sync_fn:
+            Callback that performs the allgather operation on the metric state. When ``None``, DDP
+            will be used to perform the allgather
 
     Raises:
         ValueError:
@@ -116,7 +119,7 @@ def __init__(
         pos_label: Optional[int] = None,
         average: Optional[str] = "macro",
         max_fpr: Optional[float] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/avg_precision.py b/torchmetrics/classification/avg_precision.py
index 8a8aa0b04d4..dbfcabcbf13 100644
--- a/torchmetrics/classification/avg_precision.py
+++ b/torchmetrics/classification/avg_precision.py
@@ -58,7 +58,11 @@ class AveragePrecision(Metric):
               the metric for every class.
 
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -93,7 +97,7 @@ def __init__(
         num_classes: Optional[int] = None,
         pos_label: Optional[int] = None,
         average: Optional[str] = "macro",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
diff --git a/torchmetrics/classification/binned_precision_recall.py b/torchmetrics/classification/binned_precision_recall.py
index 0abe42b6d6d..0e63710f1a2 100644
--- a/torchmetrics/classification/binned_precision_recall.py
+++ b/torchmetrics/classification/binned_precision_recall.py
@@ -62,7 +62,11 @@ class BinnedPrecisionRecallCurve(Metric):
             It is used for computation will lead to more detailed curve and accurate estimates,
             but will be slower and consume more memory.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -122,7 +126,7 @@ def __init__(
         self,
         num_classes: int,
         thresholds: Union[int, Tensor, List[float], None] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
@@ -210,7 +214,11 @@ class BinnedAveragePrecision(BinnedPrecisionRecallCurve):
             It is used for computation will lead to more detailed curve and accurate estimates,
             but will be slower and consume more memory
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         process_group:
             Specify the process group on which synchronization is called.
 
@@ -262,7 +270,11 @@ class BinnedRecallAtFixedPrecision(BinnedPrecisionRecallCurve):
             It is used for computation will lead to more detailed curve and accurate estimates,
             but will be slower and consume more memory
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         process_group:
             Specify the process group on which synchronization is called.
 
@@ -295,7 +307,7 @@ def __init__(
         num_classes: int,
         min_precision: float,
         thresholds: Union[int, Tensor, List[float], None] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
diff --git a/torchmetrics/classification/calibration_error.py b/torchmetrics/classification/calibration_error.py
index 3a8d0782efc..f387aebdc9f 100644
--- a/torchmetrics/classification/calibration_error.py
+++ b/torchmetrics/classification/calibration_error.py
@@ -54,7 +54,12 @@ class CalibrationError(Metric):
         norm: Norm used to compare empirical and expected probability bins.
             Defaults to "l1", or Expected Calibration Error.
         debias: Applies debiasing term, only implemented for l2 norm. Defaults to True.
-        compute_on_step:  Forward only calls ``update()`` and return None if this is set to False.
+        compute_on_step:
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step: Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
         process_group: Specify the process group on which synchronization is called.
diff --git a/torchmetrics/classification/cohen_kappa.py b/torchmetrics/classification/cohen_kappa.py
index 1583405f352..592ebd1c644 100644
--- a/torchmetrics/classification/cohen_kappa.py
+++ b/torchmetrics/classification/cohen_kappa.py
@@ -59,7 +59,11 @@ class labels.
             of binary or multi-label inputs. Default value of 0.5 corresponds to input being probabilities.
 
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
 
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
@@ -86,7 +90,7 @@ def __init__(
         num_classes: int,
         weights: Optional[str] = None,
         threshold: float = 0.5,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
diff --git a/torchmetrics/classification/confusion_matrix.py b/torchmetrics/classification/confusion_matrix.py
index e1df6b4ba4f..0a2a0731438 100644
--- a/torchmetrics/classification/confusion_matrix.py
+++ b/torchmetrics/classification/confusion_matrix.py
@@ -56,7 +56,11 @@ class ConfusionMatrix(Metric):
         multilabel:
             determines if data is multilabel or not.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -100,7 +104,7 @@ def __init__(
         normalize: Optional[str] = None,
         threshold: float = 0.5,
         multilabel: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
diff --git a/torchmetrics/classification/f_beta.py b/torchmetrics/classification/f_beta.py
index 24ad3bc47ca..50e3da60bf5 100644
--- a/torchmetrics/classification/f_beta.py
+++ b/torchmetrics/classification/f_beta.py
@@ -107,7 +107,11 @@ class FBetaScore(StatScores):
             for a more detailed explanation and examples.
 
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
@@ -142,7 +146,7 @@ def __init__(
         ignore_index: Optional[int] = None,
         top_k: Optional[int] = None,
         multiclass: Optional[bool] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
@@ -248,7 +252,11 @@ class F1Score(FBetaScore):
             for a more detailed explanation and examples.
 
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
@@ -281,7 +289,7 @@ def __init__(
         ignore_index: Optional[int] = None,
         top_k: Optional[int] = None,
         multiclass: Optional[bool] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/hamming.py b/torchmetrics/classification/hamming.py
index ea9e34c8ba4..9cbe3256780 100644
--- a/torchmetrics/classification/hamming.py
+++ b/torchmetrics/classification/hamming.py
@@ -43,7 +43,11 @@ class HammingDistance(Metric):
             Threshold for transforming probability or logit predictions to binary (0,1) predictions, in the case
             of binary or multi-label inputs. Default value of 0.5 corresponds to input being probabilities.
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -75,7 +79,7 @@ class HammingDistance(Metric):
     def __init__(
         self,
         threshold: float = 0.5,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/hinge.py b/torchmetrics/classification/hinge.py
index ac69491d97e..2948bac0dec 100644
--- a/torchmetrics/classification/hinge.py
+++ b/torchmetrics/classification/hinge.py
@@ -93,7 +93,7 @@ def __init__(
         self,
         squared: bool = False,
         multiclass_mode: Optional[Union[str, MulticlassMode]] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/jaccard.py b/torchmetrics/classification/jaccard.py
index 35e7fd4380e..a638fbff1d2 100644
--- a/torchmetrics/classification/jaccard.py
+++ b/torchmetrics/classification/jaccard.py
@@ -60,7 +60,11 @@ class JaccardIndex(ConfusionMatrix):
             - ``'none'``: no reduction will be applied
 
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -87,7 +91,7 @@ def __init__(
         absent_score: float = 0.0,
         threshold: float = 0.5,
         reduction: str = "elementwise_mean",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
diff --git a/torchmetrics/classification/kl_divergence.py b/torchmetrics/classification/kl_divergence.py
index 1a6f6af1407..a929229bbf5 100644
--- a/torchmetrics/classification/kl_divergence.py
+++ b/torchmetrics/classification/kl_divergence.py
@@ -71,7 +71,7 @@ def __init__(
         self,
         log_prob: bool = False,
         reduction: Optional[str] = "mean",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/matthews_corrcoef.py b/torchmetrics/classification/matthews_corrcoef.py
index c643927f314..2b168a8b328 100644
--- a/torchmetrics/classification/matthews_corrcoef.py
+++ b/torchmetrics/classification/matthews_corrcoef.py
@@ -54,7 +54,11 @@ class MatthewsCorrCoef(Metric):
         threshold:
             Threshold value for binary or multi-label probabilites.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -81,7 +85,7 @@ def __init__(
         self,
         num_classes: int,
         threshold: float = 0.5,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/precision_recall.py b/torchmetrics/classification/precision_recall.py
index 627fde14a4b..27b2ff33652 100644
--- a/torchmetrics/classification/precision_recall.py
+++ b/torchmetrics/classification/precision_recall.py
@@ -93,7 +93,11 @@ class Precision(StatScores):
             for a more detailed explanation and examples.
 
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
@@ -132,7 +136,7 @@ def __init__(
         ignore_index: Optional[int] = None,
         top_k: Optional[int] = None,
         multiclass: Optional[bool] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
@@ -244,7 +248,11 @@ class Recall(StatScores):
             for a more detailed explanation and examples.
 
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
@@ -283,7 +291,7 @@ def __init__(
         ignore_index: Optional[int] = None,
         top_k: Optional[int] = None,
         multiclass: Optional[bool] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/precision_recall_curve.py b/torchmetrics/classification/precision_recall_curve.py
index 54ca2c08696..c38bfb7ebab 100644
--- a/torchmetrics/classification/precision_recall_curve.py
+++ b/torchmetrics/classification/precision_recall_curve.py
@@ -44,7 +44,11 @@ class PrecisionRecallCurve(Metric):
             this argument should not be set as we iteratively change it in the
             range [0,num_classes-1]
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -89,7 +93,7 @@ def __init__(
         self,
         num_classes: Optional[int] = None,
         pos_label: Optional[int] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
diff --git a/torchmetrics/classification/roc.py b/torchmetrics/classification/roc.py
index c52597f0359..de82c825e03 100644
--- a/torchmetrics/classification/roc.py
+++ b/torchmetrics/classification/roc.py
@@ -45,7 +45,11 @@ class ROC(Metric):
             this argument should not be set as we iteratively change it in the
             range [0,num_classes-1]
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -116,7 +120,7 @@ def __init__(
         self,
         num_classes: Optional[int] = None,
         pos_label: Optional[int] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/specificity.py b/torchmetrics/classification/specificity.py
index c21ccb8211a..37a49190b4d 100644
--- a/torchmetrics/classification/specificity.py
+++ b/torchmetrics/classification/specificity.py
@@ -94,7 +94,11 @@ class Specificity(StatScores):
             for a more detailed explanation and examples.
 
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
@@ -133,7 +137,7 @@ def __init__(
         ignore_index: Optional[int] = None,
         top_k: Optional[int] = None,
         multiclass: Optional[bool] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/classification/stat_scores.py b/torchmetrics/classification/stat_scores.py
index 89337403401..63d885fa10e 100644
--- a/torchmetrics/classification/stat_scores.py
+++ b/torchmetrics/classification/stat_scores.py
@@ -91,7 +91,11 @@ class StatScores(Metric):
             for a more detailed explanation and examples.
 
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
@@ -143,7 +147,7 @@ def __init__(
         ignore_index: Optional[int] = None,
         mdmc_reduce: Optional[str] = None,
         multiclass: Optional[bool] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/detection/map.py b/torchmetrics/detection/map.py
index e26fca8ae79..d1de3ddc16a 100644
--- a/torchmetrics/detection/map.py
+++ b/torchmetrics/detection/map.py
@@ -169,7 +169,11 @@ class MeanAveragePrecision(Metric):
         class_metrics:
             Option to enable per-class metrics for mAP and mAR_100. Has a performance impact.
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
@@ -235,7 +239,7 @@ def __init__(
         rec_thresholds: Optional[List[float]] = None,
         max_detection_thresholds: Optional[List[int]] = None,
         class_metrics: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/image/fid.py b/torchmetrics/image/fid.py
index 5a023bd9425..b3f5b1613c5 100644
--- a/torchmetrics/image/fid.py
+++ b/torchmetrics/image/fid.py
@@ -162,7 +162,11 @@ class FrechetInceptionDistance(Metric):
               an ``[N,d]`` matrix where ``N`` is the batch size and ``d`` is the feature size.
 
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
diff --git a/torchmetrics/image/inception.py b/torchmetrics/image/inception.py
index a9a2fac1dd1..2601b982419 100644
--- a/torchmetrics/image/inception.py
+++ b/torchmetrics/image/inception.py
@@ -62,7 +62,11 @@ class InceptionScore(Metric):
         splits: integer determining how many splits the inception score calculation should be split among
 
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
diff --git a/torchmetrics/image/kid.py b/torchmetrics/image/kid.py
index 4726298ac11..4ccf0201d9c 100644
--- a/torchmetrics/image/kid.py
+++ b/torchmetrics/image/kid.py
@@ -113,7 +113,11 @@ class KernelInceptionDistance(Metric):
         coef:
             Bias term in the polynomial kernel.
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
diff --git a/torchmetrics/image/lpip.py b/torchmetrics/image/lpip.py
index f492961057c..6e1cfa179e2 100644
--- a/torchmetrics/image/lpip.py
+++ b/torchmetrics/image/lpip.py
@@ -59,7 +59,11 @@ class LearnedPerceptualImagePatchSimilarity(Metric):
         net_type: str indicating backbone network type to use. Choose between `'alex'`, `'vgg'` or `'squeeze'`
         reduction: str indicating how to reduce over the batch dimension. Choose between `'sum'` or `'mean'`.
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step
@@ -101,7 +105,7 @@ def __init__(
         self,
         net_type: str = "alex",
         reduction: str = "mean",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable[[Tensor], List[Tensor]] = None,
diff --git a/torchmetrics/image/psnr.py b/torchmetrics/image/psnr.py
index 5f92154ebe4..0108cc5e3f6 100644
--- a/torchmetrics/image/psnr.py
+++ b/torchmetrics/image/psnr.py
@@ -44,7 +44,11 @@ class PeakSignalNoiseRatio(Metric):
             Dimensions to reduce PSNR scores over, provided as either an integer or a list of integers. Default is
             None meaning scores will be reduced across all dimensions and all batches.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -77,7 +81,7 @@ def __init__(
         base: float = 10.0,
         reduction: str = "elementwise_mean",
         dim: Optional[Union[int, Tuple[int, ...]]] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
diff --git a/torchmetrics/image/ssim.py b/torchmetrics/image/ssim.py
index 1f5aab9973e..b94cbf4ebd3 100644
--- a/torchmetrics/image/ssim.py
+++ b/torchmetrics/image/ssim.py
@@ -63,7 +63,7 @@ def __init__(
         data_range: Optional[float] = None,
         k1: float = 0.01,
         k2: float = 0.03,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
@@ -161,7 +161,7 @@ def __init__(
         k2: float = 0.03,
         betas: Tuple[float, ...] = (0.0448, 0.2856, 0.3001, 0.2363, 0.1333),
         normalize: Optional[Literal["relu", "simple"]] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
diff --git a/torchmetrics/metric.py b/torchmetrics/metric.py
index f5620ff5294..d45a9bcaf84 100644
--- a/torchmetrics/metric.py
+++ b/torchmetrics/metric.py
@@ -13,6 +13,7 @@
 # limitations under the License.
 import functools
 import inspect
+import warnings
 from abc import ABC, abstractmethod
 from contextlib import contextmanager
 from copy import deepcopy
@@ -62,6 +63,10 @@ class Metric(Module, ABC):
     Args:
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -80,7 +85,7 @@ class Metric(Module, ABC):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
@@ -93,8 +98,10 @@ def __init__(
 
         self._device = torch.device("cpu")
 
+        if compute_on_step is not None:
+            warnings.warn("Argument `decimal_places` is deprecated in v0.8 and will be removed in v0.9")
+
         self.dist_sync_on_step = dist_sync_on_step
-        self.compute_on_step = compute_on_step
         self.process_group = process_group
         self.dist_sync_fn = dist_sync_fn
         self._to_sync = True
@@ -201,29 +208,28 @@ def forward(self, *args: Any, **kwargs: Any) -> Any:
         with torch.no_grad():
             self.update(*args, **kwargs)
 
-        if self.compute_on_step:
-            self._to_sync = self.dist_sync_on_step
-            # skip restore cache operation from compute as cache is stored below.
-            self._should_unsync = False
+        self._to_sync = self.dist_sync_on_step
+        # skip restore cache operation from compute as cache is stored below.
+        self._should_unsync = False
 
-            # save context before switch
-            cache = {attr: getattr(self, attr) for attr in self._defaults}
+        # save context before switch
+        cache = {attr: getattr(self, attr) for attr in self._defaults}
 
-            # call reset, update, compute, on single batch
-            self.reset()
-            self.update(*args, **kwargs)
-            self._forward_cache = self.compute()
+        # call reset, update, compute, on single batch
+        self.reset()
+        self.update(*args, **kwargs)
+        self._forward_cache = self.compute()
 
-            # restore context
-            for attr, val in cache.items():
-                setattr(self, attr, val)
-            self._is_synced = False
+        # restore context
+        for attr, val in cache.items():
+            setattr(self, attr, val)
+        self._is_synced = False
 
-            self._should_unsync = True
-            self._to_sync = True
-            self._computed = None
+        self._should_unsync = True
+        self._to_sync = True
+        self._computed = None
 
-            return self._forward_cache
+        return self._forward_cache
 
     def _sync_dist(self, dist_sync_fn: Callable = gather_all_tensors, process_group: Optional[Any] = None) -> None:
         input_dict = {attr: getattr(self, attr) for attr in self._reductions}
diff --git a/torchmetrics/regression/cosine_similarity.py b/torchmetrics/regression/cosine_similarity.py
index 42c1e4d9095..8ff2cbbe9f1 100644
--- a/torchmetrics/regression/cosine_similarity.py
+++ b/torchmetrics/regression/cosine_similarity.py
@@ -41,7 +41,11 @@ class CosineSimilarity(Metric):
         reduction: how to reduce over the batch dimension using 'sum', 'mean' or 'none'
                     (taking the individual scores)
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -69,7 +73,7 @@ class CosineSimilarity(Metric):
     def __init__(
         self,
         reduction: str = "sum",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/regression/explained_variance.py b/torchmetrics/regression/explained_variance.py
index f4711e67579..b55dc505bf5 100644
--- a/torchmetrics/regression/explained_variance.py
+++ b/torchmetrics/regression/explained_variance.py
@@ -51,7 +51,11 @@ class ExplainedVariance(Metric):
             * `'variance_weighted'` scores are weighted by their individual variances
 
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -88,7 +92,7 @@ class ExplainedVariance(Metric):
     def __init__(
         self,
         multioutput: str = "uniform_average",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/regression/log_mse.py b/torchmetrics/regression/log_mse.py
index 70f057b6db8..d59a563cea2 100644
--- a/torchmetrics/regression/log_mse.py
+++ b/torchmetrics/regression/log_mse.py
@@ -30,7 +30,11 @@ class MeanSquaredLogError(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -56,7 +60,7 @@ class MeanSquaredLogError(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/regression/mae.py b/torchmetrics/regression/mae.py
index b4eeecbe62a..f4d623f5f50 100644
--- a/torchmetrics/regression/mae.py
+++ b/torchmetrics/regression/mae.py
@@ -30,7 +30,11 @@ class MeanAbsoluteError(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -52,7 +56,7 @@ class MeanAbsoluteError(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/regression/mape.py b/torchmetrics/regression/mape.py
index d71cc8f6719..f858d4b80b1 100644
--- a/torchmetrics/regression/mape.py
+++ b/torchmetrics/regression/mape.py
@@ -33,7 +33,11 @@ class MeanAbsolutePercentageError(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -64,7 +68,7 @@ class MeanAbsolutePercentageError(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/regression/mse.py b/torchmetrics/regression/mse.py
index 3f3ab129f5a..866aac39650 100644
--- a/torchmetrics/regression/mse.py
+++ b/torchmetrics/regression/mse.py
@@ -30,7 +30,11 @@ class MeanSquaredError(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -55,7 +59,7 @@ class MeanSquaredError(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/regression/pearson.py b/torchmetrics/regression/pearson.py
index 8224a2e112c..492f112a7e4 100644
--- a/torchmetrics/regression/pearson.py
+++ b/torchmetrics/regression/pearson.py
@@ -69,7 +69,11 @@ class PearsonCorrCoef(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -98,7 +102,7 @@ class PearsonCorrCoef(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
     ) -> None:
diff --git a/torchmetrics/regression/r2.py b/torchmetrics/regression/r2.py
index 07b3815a516..646c0f84bfe 100644
--- a/torchmetrics/regression/r2.py
+++ b/torchmetrics/regression/r2.py
@@ -59,7 +59,11 @@ class R2Score(Metric):
             * ``'variance_weighted'`` scores are weighted by their individual variances
 
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -99,7 +103,7 @@ def __init__(
         num_outputs: int = 1,
         adjusted: int = 0,
         multioutput: str = "uniform_average",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/regression/spearman.py b/torchmetrics/regression/spearman.py
index 5e1ce05529c..c43f1adc3cb 100644
--- a/torchmetrics/regression/spearman.py
+++ b/torchmetrics/regression/spearman.py
@@ -34,7 +34,11 @@ class SpearmanCorrCoef(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -60,7 +64,7 @@ class SpearmanCorrCoef(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable] = None,
diff --git a/torchmetrics/regression/symmetric_mape.py b/torchmetrics/regression/symmetric_mape.py
index f05822a6115..e926bc4d6e2 100644
--- a/torchmetrics/regression/symmetric_mape.py
+++ b/torchmetrics/regression/symmetric_mape.py
@@ -33,7 +33,11 @@ class SymmetricMeanAbsolutePercentageError(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()`` before returning the value at the step.
         process_group:
@@ -61,7 +65,7 @@ class SymmetricMeanAbsolutePercentageError(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/regression/tweedie_deviance.py b/torchmetrics/regression/tweedie_deviance.py
index bf1d37c5f90..ec62915dc62 100644
--- a/torchmetrics/regression/tweedie_deviance.py
+++ b/torchmetrics/regression/tweedie_deviance.py
@@ -54,7 +54,11 @@ class TweedieDevianceScore(Metric):
             - power = 3 : Inverse Gaussian distribution. (Requires: targets > 0 and preds > 0.)
             - otherwise : Positive stable distribution. (Requires: targets > 0 and preds > 0.)
         compute_on_step:
-            Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -82,7 +86,7 @@ class TweedieDevianceScore(Metric):
     def __init__(
         self,
         power: float = 0.0,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/retrieval/average_precision.py b/torchmetrics/retrieval/average_precision.py
index 177b1d6952d..66e02c5b48b 100644
--- a/torchmetrics/retrieval/average_precision.py
+++ b/torchmetrics/retrieval/average_precision.py
@@ -45,7 +45,11 @@ class RetrievalMAP(RetrievalMetric):
         ignore_index:
             Ignore predictions where the target is equal to this number.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
diff --git a/torchmetrics/retrieval/base.py b/torchmetrics/retrieval/base.py
index b4a3759bea6..68c6c08ad7f 100644
--- a/torchmetrics/retrieval/base.py
+++ b/torchmetrics/retrieval/base.py
@@ -54,7 +54,11 @@ class RetrievalMetric(Metric, ABC):
         ignore_index:
             Ignore predictions where the target is equal to this number.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -80,7 +84,7 @@ def __init__(
         self,
         empty_target_action: str = "neg",
         ignore_index: Optional[int] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/retrieval/fall_out.py b/torchmetrics/retrieval/fall_out.py
index 2980694465c..9faa6778220 100644
--- a/torchmetrics/retrieval/fall_out.py
+++ b/torchmetrics/retrieval/fall_out.py
@@ -50,7 +50,11 @@ class RetrievalFallOut(RetrievalMetric):
             Ignore predictions where the target is equal to this number.
         k: consider only the top k elements for each query (default: `None`, which considers them all)
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -85,7 +89,7 @@ def __init__(
         empty_target_action: str = "pos",
         ignore_index: Optional[int] = None,
         k: Optional[int] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/retrieval/hit_rate.py b/torchmetrics/retrieval/hit_rate.py
index b70c6ebab07..3a8c5616a92 100644
--- a/torchmetrics/retrieval/hit_rate.py
+++ b/torchmetrics/retrieval/hit_rate.py
@@ -48,7 +48,11 @@ class RetrievalHitRate(RetrievalMetric):
             Ignore predictions where the target is equal to this number.
         k: consider only the top k elements for each query (default: `None`, which considers them all)
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -83,7 +87,7 @@ def __init__(
         empty_target_action: str = "neg",
         ignore_index: Optional[int] = None,
         k: Optional[int] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/retrieval/ndcg.py b/torchmetrics/retrieval/ndcg.py
index 17e23dccd10..3465ecd18db 100644
--- a/torchmetrics/retrieval/ndcg.py
+++ b/torchmetrics/retrieval/ndcg.py
@@ -48,7 +48,11 @@ class RetrievalNormalizedDCG(RetrievalMetric):
             Ignore predictions where the target is equal to this number.
         k: consider only the top k elements for each query (default: `None`, which considers them all)
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -83,7 +87,7 @@ def __init__(
         empty_target_action: str = "neg",
         ignore_index: Optional[int] = None,
         k: Optional[int] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/retrieval/precision.py b/torchmetrics/retrieval/precision.py
index 8f19a74e0bd..8584d672f8b 100644
--- a/torchmetrics/retrieval/precision.py
+++ b/torchmetrics/retrieval/precision.py
@@ -48,7 +48,11 @@ class RetrievalPrecision(RetrievalMetric):
             Ignore predictions where the target is equal to this number.
         k: consider only the top k elements for each query (default: `None`, which considers them all)
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -83,7 +87,7 @@ def __init__(
         empty_target_action: str = "neg",
         ignore_index: Optional[int] = None,
         k: Optional[int] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/retrieval/r_precision.py b/torchmetrics/retrieval/r_precision.py
index 12b530ce0ea..ba350cf2115 100644
--- a/torchmetrics/retrieval/r_precision.py
+++ b/torchmetrics/retrieval/r_precision.py
@@ -45,7 +45,11 @@ class RetrievalRPrecision(RetrievalMetric):
         ignore_index:
             Ignore predictions where the target is equal to this number.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
diff --git a/torchmetrics/retrieval/recall.py b/torchmetrics/retrieval/recall.py
index 2e8a1504c61..1674adc6e52 100644
--- a/torchmetrics/retrieval/recall.py
+++ b/torchmetrics/retrieval/recall.py
@@ -48,7 +48,11 @@ class RetrievalRecall(RetrievalMetric):
             Ignore predictions where the target is equal to this number.
         k: consider only the top k elements for each query (default: `None`, which considers them all)
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -83,7 +87,7 @@ def __init__(
         empty_target_action: str = "neg",
         ignore_index: Optional[int] = None,
         k: Optional[int] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/retrieval/reciprocal_rank.py b/torchmetrics/retrieval/reciprocal_rank.py
index cdcf25a6213..79a6a4a8e75 100644
--- a/torchmetrics/retrieval/reciprocal_rank.py
+++ b/torchmetrics/retrieval/reciprocal_rank.py
@@ -45,7 +45,11 @@ class RetrievalMRR(RetrievalMetric):
         ignore_index:
             Ignore predictions where the target is equal to this number.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
diff --git a/torchmetrics/text/bert.py b/torchmetrics/text/bert.py
index 99cc0f1563d..b53b88b5b0d 100644
--- a/torchmetrics/text/bert.py
+++ b/torchmetrics/text/bert.py
@@ -97,7 +97,11 @@ class BERTScore(Metric):
         baseline_url:
             A url path to the user's own  csv/tsv file with the baseline scale.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -148,7 +152,7 @@ def __init__(
         rescale_with_baseline: bool = False,
         baseline_path: Optional[str] = None,
         baseline_url: Optional[str] = None,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/text/bleu.py b/torchmetrics/text/bleu.py
index 31b0c6709dc..dea595f4b44 100644
--- a/torchmetrics/text/bleu.py
+++ b/torchmetrics/text/bleu.py
@@ -33,9 +33,13 @@ class BLEUScore(Metric):
         n_gram:
             Gram value ranged from 1 to 4 (Default 4)
         smooth:
-            Whether or not to apply smoothing – see [2]
+            Whether or not to apply smoothing, see [2]
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -72,7 +76,7 @@ def __init__(
         self,
         n_gram: int = 4,
         smooth: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable] = None,
diff --git a/torchmetrics/text/cer.py b/torchmetrics/text/cer.py
index 078bf58b6fc..a60ae6a8eaf 100644
--- a/torchmetrics/text/cer.py
+++ b/torchmetrics/text/cer.py
@@ -43,7 +43,11 @@ class CharErrorRate(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -70,7 +74,7 @@ class CharErrorRate(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/text/chrf.py b/torchmetrics/text/chrf.py
index 487c26b627f..11cfcce3834 100644
--- a/torchmetrics/text/chrf.py
+++ b/torchmetrics/text/chrf.py
@@ -65,6 +65,10 @@ class CHRFScore(Metric):
             An indication whether a sentence-level chrF/chrF++ score to be returned.
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -107,7 +111,7 @@ def __init__(
         lowercase: bool = False,
         whitespace: bool = False,
         return_sentence_level_score: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable] = None,
diff --git a/torchmetrics/text/eed.py b/torchmetrics/text/eed.py
index 99c7b90e72a..e7abd71cb13 100644
--- a/torchmetrics/text/eed.py
+++ b/torchmetrics/text/eed.py
@@ -39,7 +39,11 @@ class ExtendedEditDistance(Metric):
         insertion:
             penalty for insertion or substitution of character
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -77,7 +81,7 @@ def __init__(
         rho: float = 0.3,
         deletion: float = 0.2,
         insertion: float = 1.0,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/text/mer.py b/torchmetrics/text/mer.py
index 6d52a842692..96986d2432e 100644
--- a/torchmetrics/text/mer.py
+++ b/torchmetrics/text/mer.py
@@ -41,7 +41,11 @@ class MatchErrorRate(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -68,7 +72,7 @@ class MatchErrorRate(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/text/rouge.py b/torchmetrics/text/rouge.py
index cf86729b2ed..f8628f602d3 100644
--- a/torchmetrics/text/rouge.py
+++ b/torchmetrics/text/rouge.py
@@ -44,6 +44,10 @@ class ROUGEScore(Metric):
             Keys that are allowed are ``rougeL``, ``rougeLsum``, and ``rouge1`` through ``rouge9``.
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -91,7 +95,7 @@ def __init__(
         use_stemmer: bool = False,
         accumulate: Literal["avg", "best"] = "best",
         rouge_keys: Union[str, Tuple[str, ...]] = ("rouge1", "rouge2", "rougeL", "rougeLsum"),  # type: ignore
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable] = None,
diff --git a/torchmetrics/text/sacre_bleu.py b/torchmetrics/text/sacre_bleu.py
index 196b173d9ff..5e28a689ecc 100644
--- a/torchmetrics/text/sacre_bleu.py
+++ b/torchmetrics/text/sacre_bleu.py
@@ -47,6 +47,10 @@ class SacreBLEUScore(BLEUScore):
             If ``True``, BLEU score over lowercased text is calculated.
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -87,7 +91,7 @@ def __init__(
         smooth: bool = False,
         tokenize: Literal["none", "13a", "zh", "intl", "char"] = "13a",
         lowercase: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable] = None,
diff --git a/torchmetrics/text/squad.py b/torchmetrics/text/squad.py
index 8f0fa4deb69..22c4cf9a018 100644
--- a/torchmetrics/text/squad.py
+++ b/torchmetrics/text/squad.py
@@ -33,6 +33,10 @@ class SQuAD(Metric):
     Args:
         compute_on_step:
             Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -64,7 +68,7 @@ class SQuAD(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Optional[Callable] = None,
diff --git a/torchmetrics/text/ter.py b/torchmetrics/text/ter.py
index df471297018..6f75c28c91a 100644
--- a/torchmetrics/text/ter.py
+++ b/torchmetrics/text/ter.py
@@ -39,7 +39,11 @@ class TranslationEditRate(Metric):
         return_sentence_level_score:
             An indication whether a sentence-level TER to be returned.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -74,7 +78,7 @@ def __init__(
         lowercase: bool = True,
         asian_support: bool = False,
         return_sentence_level_score: bool = False,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/text/wer.py b/torchmetrics/text/wer.py
index 630359f8114..ed6ced11f32 100644
--- a/torchmetrics/text/wer.py
+++ b/torchmetrics/text/wer.py
@@ -41,7 +41,11 @@ class WordErrorRate(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -68,7 +72,7 @@ class WordErrorRate(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/text/wil.py b/torchmetrics/text/wil.py
index d3940d5db8e..e8ca6686c6a 100644
--- a/torchmetrics/text/wil.py
+++ b/torchmetrics/text/wil.py
@@ -41,7 +41,11 @@ class WordInfoLost(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -68,7 +72,7 @@ class WordInfoLost(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/text/wip.py b/torchmetrics/text/wip.py
index 1a0351999ac..fd1f735c0a3 100644
--- a/torchmetrics/text/wip.py
+++ b/torchmetrics/text/wip.py
@@ -41,7 +41,11 @@ class WordInfoPreserved(Metric):
 
     Args:
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -68,7 +72,7 @@ class WordInfoPreserved(Metric):
 
     def __init__(
         self,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/wrappers/bootstrapping.py b/torchmetrics/wrappers/bootstrapping.py
index a06ad03a08e..5172e72aa91 100644
--- a/torchmetrics/wrappers/bootstrapping.py
+++ b/torchmetrics/wrappers/bootstrapping.py
@@ -56,7 +56,7 @@ def __init__(
         quantile: Optional[Union[float, Tensor]] = None,
         raw: bool = False,
         sampling_strategy: str = "poisson",
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
@@ -88,7 +88,11 @@ class basically keeps multiple copies of the same base metric in memory and when
                 when the number of samples is large. If ``'multinomial'`` is chosen, we will apply true bootstrapping
                 at the batch level to approximate bootstrapping over the hole dataset.
             compute_on_step:
-                Forward only calls ``update()`` and return ``None`` if this is set to ``False``.
+                Forward only calls ``update()`` and returns None if this is set to False.
+
+                .. deprecated:: v0.8
+                    Argument has no use anymore and will be removed v0.9.
+
             dist_sync_on_step:
                 Synchronize metric state across processes at each ``forward()``
                 before returning the value at the step
diff --git a/torchmetrics/wrappers/minmax.py b/torchmetrics/wrappers/minmax.py
index 8386516f7a3..69c90551c96 100644
--- a/torchmetrics/wrappers/minmax.py
+++ b/torchmetrics/wrappers/minmax.py
@@ -28,7 +28,11 @@ class MinMaxMetric(Metric):
         base_metric:
             The metric of which you want to keep track of its maximum and minimum values.
         compute_on_step:
-            Forward only calls ``update()`` and return None if this is set to False.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Synchronize metric state across processes at each ``forward()``
             before returning the value at the step.
@@ -66,7 +70,7 @@ class MinMaxMetric(Metric):
     def __init__(
         self,
         base_metric: Metric,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,
diff --git a/torchmetrics/wrappers/multioutput.py b/torchmetrics/wrappers/multioutput.py
index e8317b5a903..99a5a948862 100644
--- a/torchmetrics/wrappers/multioutput.py
+++ b/torchmetrics/wrappers/multioutput.py
@@ -56,7 +56,11 @@ class MultioutputWrapper(Metric):
             This is sometimes unnecessary but harmless for metrics such as `R2Score` but useful
             for certain classification metrics that can't handle additional 1-item dimensions.
         compute_on_step:
-            Whether to recompute the metric value on each update step.
+            Forward only calls ``update()`` and returns None if this is set to False.
+
+            .. deprecated:: v0.8
+                Argument has no use anymore and will be removed v0.9.
+
         dist_sync_on_step:
             Required for distributed training support.
         process_group:
@@ -97,7 +101,7 @@ def __init__(
         output_dim: int = -1,
         remove_nans: bool = True,
         squeeze_outputs: bool = True,
-        compute_on_step: bool = True,
+        compute_on_step: Optional[bool] = None,
         dist_sync_on_step: bool = False,
         process_group: Optional[Any] = None,
         dist_sync_fn: Callable = None,