Metric docs fix

CHANGELOG.md

@@ -21,9 +21,11 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 ### Added
 
-- Added metric Base classes ([#1326](https://github.com/PyTorchLightning/pytorch-lightning/pull/1326), [#1877](https://github.com/PyTorchLightning/pytorch-lightning/pull/1877))
-- Added Sklearn metrics classes ([#1327](https://github.com/PyTorchLightning/pytorch-lightning/pull/1327))
-- Added Native torch metrics ([#1488](https://github.com/PyTorchLightning/pytorch-lightning/pull/1488))
+- Added metrics 
+  * Base classes ([#1326](https://github.com/PyTorchLightning/pytorch-lightning/pull/1326), [#1877](https://github.com/PyTorchLightning/pytorch-lightning/pull/1877))
+  * Sklearn metrics classes ([#1327](https://github.com/PyTorchLightning/pytorch-lightning/pull/1327))
+  * Native torch metrics ([#1488](https://github.com/PyTorchLightning/pytorch-lightning/pull/1488))
+  * docs for all Metrics ([#2184](https://github.com/PyTorchLightning/pytorch-lightning/pull/2184), [#2209](https://github.com/PyTorchLightning/pytorch-lightning/pull/2209))
 - Added type hints in `Trainer.fit()` and `Trainer.test()` to reflect that also a list of dataloaders can be passed in ([#1723](https://github.com/PyTorchLightning/pytorch-lightning/pull/1723))
 - Allow dataloaders without sampler field present ([#1907](https://github.com/PyTorchLightning/pytorch-lightning/pull/1907))
 - Added option `save_last` to save the model at the end of every epoch in `ModelCheckpoint` [(#1908)](https://github.com/PyTorchLightning/pytorch-lightning/pull/1908)

docs/source/metrics.rst

@@ -12,7 +12,8 @@ Metrics are used to monitor model performance.
 In this package we provide two major pieces of functionality.
 
 1. A Metric class you can use to implement metrics with built-in distributed (ddp) support which are device agnostic.
-2. A collection of popular metrics already implemented for you.
+2. A collection of ready to use pupular metrics. There are two types of metrics: Class metrics and Functional metrics.
+3. A interface to call `sklearns metrics <https://scikit-learn.org/stable/modules/classes.html#module-sklearn.metrics>`_
 
 Example::
 
@@ -28,12 +29,17 @@ Out::
 
     tensor(0.7500)
 
+.. warning::
+    The metrics package is still in development! If we're missing a metric or you find a mistake, please send a PR!
+    to a few metrics. Please feel free to create an issue/PR if you have a proposed 
+    metric or have found a bug.
+
 --------------
 
 Implement a metric
 ------------------
-You can implement metrics as either a PyTorch metric or a Numpy metric. Numpy metrics
-will slow down training, use PyTorch metrics when possible.
+You can implement metrics as either a PyTorch metric or a Numpy metric (It is recommend to use PyTorch metrics when possible,
+since Numpy metrics slow down training).
 
 Use :class:`TensorMetric` to implement native PyTorch metrics. This class
 handles automated DDP syncing and converts all inputs and outputs to tensors.
@@ -76,7 +82,7 @@ Here's an example showing how to implement a NumpyMetric
 
 Class Metrics
 -------------
-The following are metrics which can be instantiated as part of a module definition (even with just
+Class metrics can be instantiated as part of a module definition (even with just
 plain PyTorch).
 
 .. testcode::
@@ -316,3 +322,98 @@ to_onehot (F)
 
 .. autofunction:: pytorch_lightning.metrics.functional.to_onehot
     :noindex:
+
+----------------
+
+Sklearn interface
+-----------------
+
+Lightning supports `sklearns metrics module <https://scikit-learn.org/stable/modules/classes.html#module-sklearn.metrics>`_ 
+as a backend for calculating metrics. Sklearns metrics are well tested and robust, 
+but requires conversion between pytorch and numpy thus may slow down your computations.
+
+To use the sklearn backend of metrics simply import as
+
+.. code-block:: python
+
+    import pytorch_lightning.metrics.sklearns import plm
+    metric = plm.Accuracy(normalize=True)
+    val = metric(pred, target)
+
+Each converted sklearn metric comes has the same interface as its 
+originally counterpart (e.g. accuracy takes the additional `normalize` keyword). 
+Like the native Lightning metrics these converted sklearn metrics also come 
+with built-in distributed (ddp) support.
+
+SklearnMetric (sk)
+^^^^^^^^^^^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.SklearnMetric
+    :noindex:
+
+Accuracy (sk)
+^^^^^^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.Accuracy
+    :noindex:
+
+AUC (sk)
+^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.AUC
+    :noindex:
+
+AveragePrecision (sk)
+^^^^^^^^^^^^^^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.AveragePrecision
+    :noindex:
+
+
+ConfusionMatrix (sk)
+^^^^^^^^^^^^^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.ConfusionMatrix
+    :noindex:
+
+F1 (sk)
+^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.F1
+    :noindex:
+
+FBeta (sk)
+^^^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.FBeta
+    :noindex:
+
+Precision (sk)
+^^^^^^^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.Precision
+    :noindex:
+
+Recall (sk)
+^^^^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.Recall
+    :noindex:
+
+PrecisionRecallCurve (sk)
+^^^^^^^^^^^^^^^^^^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.PrecisionRecallCurve
+    :noindex:
+
+ROC (sk)
+^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.ROC
+    :noindex:
+
+AUROC (sk)
+^^^^^^^^^^
+
+.. autofunction:: pytorch_lightning.metrics.sklearns.AUROC
+    :noindex:

pytorch_lightning/metrics/__init__.py

@@ -1,15 +1,41 @@
 from pytorch_lightning.metrics.converters import numpy_metric, tensor_metric
 from pytorch_lightning.metrics.metric import Metric, TensorMetric, NumpyMetric
-from pytorch_lightning.metrics.sklearn import (
-    SklearnMetric,
+from pytorch_lightning.metrics.classification import (
     Accuracy,
     AveragePrecision,
-    AUC,
     ConfusionMatrix,
     F1,
     FBeta,
-    Precision,
     Recall,
-    PrecisionRecallCurve,
     ROC,
-    AUROC)
+    AUROC,
+    DiceCoefficient,
+    MulticlassPrecisionRecall,
+    MulticlassROC,
+    Precision,
+    PrecisionRecall,
+)
+from pytorch_lightning.metrics.sklearns import (
+    AUC,
+    PrecisionRecallCurve,
+    SklearnMetric,
+)
+
+__all__ = [
+    'AUC',
+    'AUROC',
+    'Accuracy',
+    'AveragePrecision',
+    'ConfusionMatrix',
+    'DiceCoefficient',
+    'F1',
+    'FBeta',
+    'MulticlassPrecisionRecall',
+    'MulticlassROC',
+    'Precision',
+    'PrecisionRecall',
+    'PrecisionRecallCurve',
+    'ROC',
+    'Recall',
+    'SklearnMetric',
+]