Prune metrics: regression 8/n

CHANGELOG.md

@@ -90,6 +90,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
  [#6584](https://github.com/PyTorchLightning/pytorch-lightning/pull/6584),
 
+ [#6636](https://github.com/PyTorchLightning/pytorch-lightning/pull/6636),
+
 )
 
 

pytorch_lightning/metrics/functional/explained_variance.py

@@ -11,77 +11,21 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-from typing import Sequence, Tuple, Union
+from typing import Sequence, Union
 
 import torch
-from torchmetrics.utilities.checks import _check_same_shape
+from torchmetrics.functional import explained_variance as _explained_variance
 
-
-def _explained_variance_update(preds: torch.Tensor, target: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
- _check_same_shape(preds, target)
- return preds, target
-
-
-def _explained_variance_compute(
- preds: torch.Tensor,
- target: torch.Tensor,
- multioutput: str = 'uniform_average',
-) -> Union[torch.Tensor, Sequence[torch.Tensor]]:
- diff_avg = torch.mean(target - preds, dim=0)
- numerator = torch.mean((target - preds - diff_avg)**2, dim=0)
-
- target_avg = torch.mean(target, dim=0)
- denominator = torch.mean((target - target_avg)**2, dim=0)
-
- # Take care of division by zero
- nonzero_numerator = numerator != 0
- nonzero_denominator = denominator != 0
- valid_score = nonzero_numerator & nonzero_denominator
- output_scores = torch.ones_like(diff_avg)
- output_scores[valid_score] = 1.0 - (numerator[valid_score] / denominator[valid_score])
- output_scores[nonzero_numerator & ~nonzero_denominator] = 0.
-
- # Decide what to do in multioutput case
- # Todo: allow user to pass in tensor with weights
- if multioutput == 'raw_values':
- return output_scores
- if multioutput == 'uniform_average':
- return torch.mean(output_scores)
- if multioutput == 'variance_weighted':
- denom_sum = torch.sum(denominator)
- return torch.sum(denominator / denom_sum * output_scores)
+from pytorch_lightning.utilities.deprecation import deprecated
 
 
+@deprecated(target=_explained_variance, ver_deprecate="1.3.0", ver_remove="1.5.0")
 def explained_variance(
  preds: torch.Tensor,
  target: torch.Tensor,
  multioutput: str = 'uniform_average',
 ) -> Union[torch.Tensor, Sequence[torch.Tensor]]:
  """
- Computes explained variance.
-
- Args:
- preds: estimated labels
- target: ground truth labels
- multioutput: Defines aggregation in the case of multiple output scores. Can be one
- of the following strings (default is `'uniform_average'`.):
-
- * `'raw_values'` returns full set of scores
- * `'uniform_average'` scores are uniformly averaged
- * `'variance_weighted'` scores are weighted by their individual variances
-
- Example:
-
- >>> from pytorch_lightning.metrics.functional import explained_variance
- >>> target = torch.tensor([3, -0.5, 2, 7])
- >>> preds = torch.tensor([2.5, 0.0, 2, 8])
- >>> explained_variance(preds, target)
- tensor(0.9572)
-
- >>> target = torch.tensor([[0.5, 1], [-1, 1], [7, -6]])
- >>> preds = torch.tensor([[0, 2], [-1, 2], [8, -5]])
- >>> explained_variance(preds, target, multioutput='raw_values')
- tensor([0.9677, 1.0000])
+ .. deprecated::
+ Use :func:`torchmetrics.functional.explained_variance`. Will be removed in v1.5.0.
  """
- preds, target = _explained_variance_update(preds, target)
- return _explained_variance_compute(preds, target, multioutput)

pytorch_lightning/metrics/functional/mean_absolute_error.py

@@ -11,40 +11,16 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-from typing import Tuple
 
 import torch
-from torchmetrics.utilities.checks import _check_same_shape
+from torchmetrics.functional import mean_absolute_error as _mean_absolute_error
 
-
-def _mean_absolute_error_update(preds: torch.Tensor, target: torch.Tensor) -> Tuple[torch.Tensor, int]:
- _check_same_shape(preds, target)
- sum_abs_error = torch.sum(torch.abs(preds - target))
- n_obs = target.numel()
- return sum_abs_error, n_obs
-
-
-def _mean_absolute_error_compute(sum_abs_error: torch.Tensor, n_obs: int) -> torch.Tensor:
- return sum_abs_error / n_obs
+from pytorch_lightning.utilities.deprecation import deprecated
 
 
+@deprecated(target=_mean_absolute_error, ver_deprecate="1.3.0", ver_remove="1.5.0")
 def mean_absolute_error(preds: torch.Tensor, target: torch.Tensor) -> torch.Tensor:
  """
- Computes mean absolute error
-
- Args:
- pred: estimated labels
- target: ground truth labels
-
- Return:
- Tensor with MAE
-
- Example:
- >>> from pytorch_lightning.metrics.functional import mean_absolute_error
- >>> x = torch.tensor([0., 1, 2, 3])
- >>> y = torch.tensor([0., 1, 2, 2])
- >>> mean_absolute_error(x, y)
- tensor(0.2500)
+ .. deprecated::
+ Use :func:`torchmetrics.functional.mean_absolute_error`. Will be removed in v1.5.0.
  """
- sum_abs_error, n_obs = _mean_absolute_error_update(preds, target)
- return _mean_absolute_error_compute(sum_abs_error, n_obs)

pytorch_lightning/metrics/functional/mean_relative_error.py

@@ -11,43 +11,16 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-from typing import Tuple
 
 import torch
-from torchmetrics.utilities.checks import _check_same_shape
+from torchmetrics.functional.regression.mean_relative_error import mean_relative_error as _mean_relative_error
 
-
-def _mean_relative_error_update(preds: torch.Tensor, target: torch.Tensor) -> Tuple[torch.Tensor, int]:
- _check_same_shape(preds, target)
- target_nz = target.clone()
- target_nz[target == 0] = 1
- sum_rltv_error = torch.sum(torch.abs((preds - target) / target_nz))
- n_obs = target.numel()
- return sum_rltv_error, n_obs
-
-
-def _mean_relative_error_compute(sum_rltv_error: torch.Tensor, n_obs: int) -> torch.Tensor:
- return sum_rltv_error / n_obs
+from pytorch_lightning.utilities.deprecation import deprecated
 
 
+@deprecated(target=_mean_relative_error, ver_deprecate="1.3.0", ver_remove="1.5.0")
 def mean_relative_error(preds: torch.Tensor, target: torch.Tensor) -> torch.Tensor:
  """
- Computes mean relative error
-
- Args:
- pred: estimated labels
- target: ground truth labels
-
- Return:
- Tensor with mean relative error
-
- Example:
-
- >>> x = torch.tensor([0., 1, 2, 3])
- >>> y = torch.tensor([0., 1, 2, 2])
- >>> mean_relative_error(x, y)
- tensor(0.1250)
-
+ .. deprecated::
+ Use :func:`torchmetrics.functional.regression.mean_relative_error`. Will be removed in v1.5.0.
  """
- sum_rltv_error, n_obs = _mean_relative_error_update(preds, target)
- return _mean_relative_error_compute(sum_rltv_error, n_obs)

pytorch_lightning/metrics/functional/mean_squared_error.py

@@ -11,40 +11,16 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-from typing import Tuple
 
 import torch
-from torchmetrics.utilities.checks import _check_same_shape
+from torchmetrics.functional import mean_squared_error as _mean_squared_error
 
-
-def _mean_squared_error_update(preds: torch.Tensor, target: torch.Tensor) -> Tuple[torch.Tensor, int]:
- _check_same_shape(preds, target)
- sum_squared_error = torch.sum(torch.pow(preds - target, 2))
- n_obs = target.numel()
- return sum_squared_error, n_obs
-
-
-def _mean_squared_error_compute(sum_squared_error: torch.Tensor, n_obs: int) -> torch.Tensor:
- return sum_squared_error / n_obs
+from pytorch_lightning.utilities.deprecation import deprecated
 
 
+@deprecated(target=_mean_squared_error, ver_deprecate="1.3.0", ver_remove="1.5.0")
 def mean_squared_error(preds: torch.Tensor, target: torch.Tensor) -> torch.Tensor:
  """
- Computes mean squared error
-
- Args:
- preds: estimated labels
- target: ground truth labels
-
- Return:
- Tensor with MSE
-
- Example:
- >>> from pytorch_lightning.metrics.functional import mean_squared_error
- >>> x = torch.tensor([0., 1, 2, 3])
- >>> y = torch.tensor([0., 1, 2, 2])
- >>> mean_squared_error(x, y)
- tensor(0.2500)
+ .. deprecated::
+ Use :func:`torchmetrics.functional.mean_squared_error`. Will be removed in v1.5.0.
  """
- sum_squared_error, n_obs = _mean_squared_error_update(preds, target)
- return _mean_squared_error_compute(sum_squared_error, n_obs)

pytorch_lightning/metrics/functional/mean_squared_log_error.py

@@ -11,40 +11,16 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-from typing import Tuple
 
 import torch
-from torchmetrics.utilities.checks import _check_same_shape
+from torchmetrics.functional import mean_squared_log_error as _mean_squared_log_error
 
-
-def _mean_squared_log_error_update(preds: torch.Tensor, target: torch.Tensor) -> Tuple[torch.Tensor, int]:
- _check_same_shape(preds, target)
- sum_squared_log_error = torch.sum(torch.pow(torch.log1p(preds) - torch.log1p(target), 2))
- n_obs = target.numel()
- return sum_squared_log_error, n_obs
-
-
-def _mean_squared_log_error_compute(sum_squared_log_error: torch.Tensor, n_obs: int) -> torch.Tensor:
- return sum_squared_log_error / n_obs
+from pytorch_lightning.utilities.deprecation import deprecated
 
 
+@deprecated(target=_mean_squared_log_error, ver_deprecate="1.3.0", ver_remove="1.5.0")
 def mean_squared_log_error(preds: torch.Tensor, target: torch.Tensor) -> torch.Tensor:
  """
- Computes mean squared log error
-
- Args:
- preds: estimated labels
- target: ground truth labels
-
- Return:
- Tensor with RMSLE
-
- Example:
- >>> from pytorch_lightning.metrics.functional import mean_squared_log_error
- >>> x = torch.tensor([0., 1, 2, 3])
- >>> y = torch.tensor([0., 1, 2, 2])
- >>> mean_squared_log_error(x, y)
- tensor(0.0207)
+ .. deprecated::
+ Use :func:`torchmetrics.functional.mean_squared_log_error`. Will be removed in v1.5.0.
  """
- sum_squared_log_error, n_obs = _mean_squared_log_error_update(preds, target)
- return _mean_squared_log_error_compute(sum_squared_log_error, n_obs)