remove NLL in metric (#18794)

python/mxnet/gluon/metric.py

@@ -1352,9 +1352,12 @@ class :math:`k`.
         Index of invalid label to ignore when
         counting. By default, sets to -1.
         If set to `None`, it will include all entries.
-    axis : int (default -1)
+    axis : int, default -1
         The axis from prediction that was used to
         compute softmax. By default use the last axis.
+    from_logits : boolean, default False
+        Whether `pred` is expected to be a logits tensor.
+        By default, we assume that `pred` encodes a probability distribution.
     name : str
         Name of this metric instance for display.
     output_names : list of str, or None
@@ -1373,12 +1376,13 @@ class :math:`k`.
     >>> print ce.get()
     ('cross-entropy', 0.57159948348999023)
     """
-    def __init__(self, eps=1e-12, ignore_label=None, axis=-1, name='cross-entropy',
-                 output_names=None, label_names=None):
+    def __init__(self, eps=1e-12, ignore_label=None, axis=-1, from_logits=False,
+                 name='cross-entropy', output_names=None, label_names=None):
         super(CrossEntropy, self).__init__(
             name, output_names=output_names, label_names=label_names)
         self.ignore_label = ignore_label
         self.axis = axis
+        self.from_logits = from_logits
         self.eps = eps
 
     def update(self, labels, preds):
@@ -1400,6 +1404,8 @@ def update(self, labels, preds):
             assert label.size == pred.size/pred.shape[-1], \
                 "shape mismatch: %s vs. %s"%(label.shape, pred.shape)
             label = label.reshape((label.size,))
+            if self.from_logits:
+                pred = ndarray.softmax(pred, axis=self.axis)
             pred = ndarray.pick(pred.as_in_context(label.ctx), label.astype(dtype='int32'), axis=self.axis)
             label = label.as_np_ndarray()
             pred = pred.as_np_ndarray()
@@ -1469,11 +1475,11 @@ class Perplexity(CrossEntropy):
     >>> print perp.get()
     ('Perplexity', 1.7710976285155853)
     """
-    def __init__(self, eps=1e-12, ignore_label=None, axis=-1, name='perplexity',
-                 output_names=None, label_names=None):
+    def __init__(self, eps=1e-12, ignore_label=None, axis=-1, from_logits=False,
+                 name='perplexity', output_names=None, label_names=None):
         super(Perplexity, self).__init__(
-            name=name, eps=eps, ignore_label=ignore_label, axis=axis,
-            output_names=output_names, label_names=label_names)
+            eps=eps, ignore_label=ignore_label, axis=axis, from_logits=from_logits,
+            name=name, output_names=output_names, label_names=label_names)
 
     def get(self):
         if self.num_inst == 0:
@@ -1482,77 +1488,6 @@ def get(self):
             return (self.name, math.exp(self.sum_metric/self.num_inst))
 
 
-@register
-@alias('nll_loss')
-@use_np
-class NegativeLogLikelihood(EvalMetric):
-    """Computes the negative log-likelihood loss.
-
-    The negative log-likelihoodd loss over a batch of sample size :math:`N` is given by
-
-    .. math::
-       -\\sum_{n=1}^{N}\\sum_{k=1}^{K}t_{nk}\\log (y_{nk}),
-
-    where :math:`K` is the number of classes, :math:`y_{nk}` is the prediceted probability for
-    :math:`k`-th class for :math:`n`-th sample. :math:`t_{nk}=1` if and only if sample
-    :math:`n` belongs to class :math:`k`.
-
-    Parameters
-    ----------
-    eps : float
-        Negative log-likelihood loss is undefined for predicted value is 0,
-        so predicted values are added with the small constant.
-    name : str
-        Name of this metric instance for display.
-    output_names : list of str, or None
-        Name of predictions that should be used when updating with update_dict.
-        By default include all predictions.
-    label_names : list of str, or None
-        Name of labels that should be used when updating with update_dict.
-        By default include all labels.
-
-    Examples
-    --------
-    >>> predicts = [mx.nd.array([[0.3, 0.7], [0, 1.], [0.4, 0.6]])]
-    >>> labels   = [mx.nd.array([0, 1, 1])]
-    >>> nll_loss = mx.gluon.metric.NegativeLogLikelihood()
-    >>> nll_loss.update(labels, predicts)
-    >>> print nll_loss.get()
-    ('nll-loss', 0.57159948348999023)
-    """
-    def __init__(self, eps=1e-12, name='nll-loss',
-                 output_names=None, label_names=None):
-        super(NegativeLogLikelihood, self).__init__(
-            name, eps=eps,
-            output_names=output_names, label_names=label_names)
-        self.eps = eps
-
-    def update(self, labels, preds):
-        """Updates the internal evaluation result.
-
-        Parameters
-        ----------
-        labels : list of `NDArray`
-            The labels of the data.
-
-        preds : list of `NDArray`
-            Predicted values.
-        """
-        labels, preds = check_label_shapes(labels, preds, True)
-
-        for label, pred in zip(labels, preds):
-            label = label.as_np_ndarray()
-            pred = pred.as_np_ndarray().as_in_ctx(label.ctx)
-
-            label = label.reshape(-1)
-            num_examples = pred.shape[0]
-            assert label.shape[0] == num_examples, (label.shape[0], num_examples)
-            prob = pred[numpy.arange(num_examples, dtype=numpy.int64), numpy.int64(label)]
-            nll = (-numpy.log(prob + self.eps)).sum()
-            self.sum_metric += nll
-            self.num_inst += num_examples
-
-
 @register
 @alias('pearsonr')
 @use_np

tests/python/unittest/test_metric.py

@@ -39,19 +39,25 @@ def test_metrics():
     check_metric('perplexity', axis=-1)
     check_metric('pearsonr')
     check_metric('pcc')
-    check_metric('nll_loss')
+    check_metric('ce')
     check_metric('loss')
     composite = mx.gluon.metric.create(['acc', 'f1'])
     check_metric(composite)
 
-def test_nll_loss():
-    metric = mx.gluon.metric.create('nll_loss')
+def test_ce():
+    metric = mx.gluon.metric.create('ce')
     pred = mx.nd.array([[0.2, 0.3, 0.5], [0.6, 0.1, 0.3]])
     label = mx.nd.array([2, 1])
     metric.update([label], [pred])
     _, loss = metric.get()
     expected_loss = -(np.log(pred[0][2].asscalar()) + np.log(pred[1][1].asscalar())) / 2
     assert loss == expected_loss
+    metric = mx.gluon.metric.create('ce', from_logits=True)
+    pred = mx.nd.log(pred)
+    metric.update([label], [pred])
+    _, loss = metric.get()
+    np.testing.assert_almost_equal(loss, expected_loss)
+
 
 def test_acc():
     pred = mx.nd.array([[0.3, 0.7], [0, 1.], [0.4, 0.6]])
@@ -159,15 +165,15 @@ def test_multiclass_f1():
     macroF1.update([label11, label12], [pred11, pred12])
     assert microF1.num_inst == 6
     assert macroF1.num_inst == 6
-    
+
     # from sklearn.metrics import f1_score
     # overall_pred = [0, 1, 2, 0, 1, 2]
     # overall_label = [0, 2, 1, 0, 0, 1]
     fmacro = 0.26666666666666666 #f1_score(overall_label, overall_pred, average="macro")
     fmicro = 0.3333333333333333 #f1_score(overall_label, overall_pred, average="micro")
     np.testing.assert_almost_equal(microF1.get()[1], fmicro)
     np.testing.assert_almost_equal(macroF1.get()[1], fmacro)
-    
+
 @xfail_when_nonstandard_decimal_separator
 def test_multilabel_f1():
     microF1 = mx.gluon.metric.create("f1", class_type="multilabel", average="micro")
@@ -183,7 +189,7 @@ def test_multilabel_f1():
     macroF1.update([label], [pred])
     microF1.update([label], [pred])
     assert macroF1.get()[1] == 0.5 # one class is 1.0, the other is 0. (divided by 0)
-    np.testing.assert_almost_equal(microF1.get()[1], 2.0 / 3)  
+    np.testing.assert_almost_equal(microF1.get()[1], 2.0 / 3)
     macroF1.reset()
     microF1.reset()
 
@@ -209,7 +215,7 @@ def test_mcc():
     microMCC = mx.gluon.metric.create("mcc")
 
     assert np.isnan(microMCC.get()[1])
-    
+
     # check divide by zero
     pred = mx.nd.array([[0.9, 0.1],
                         [0.8, 0.2]])