reimplement paddle.nn.functional.sigmoid_focal_loss (#27748)

* reimplement paddle.nn.functional.sigmoid_focal_loss. test=develop

* fix reduction error message. test=develop

* fix exp. test=develop

* reset the shape of logit. test=develop

* delete disable_static in example. test=develop

python/paddle/fluid/tests/unittests/test_sigmoid_focal_loss.py

@@ -0,0 +1,165 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# 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.
+
+import paddle
+import paddle.fluid as fluid
+import numpy as np
+import unittest
+from op_test import OpTest
+from test_sigmoid_focal_loss_op import sigmoid_focal_loss_forward
+
+
+def call_sfl_functional(logit,
+                        label,
+                        normalizer,
+                        alpha=0.25,
+                        gamma=2.0,
+                        reduction='sum'):
+    res = paddle.nn.functional.sigmoid_focal_loss(
+        logit, label, normalizer, alpha=alpha, gamma=gamma, reduction=reduction)
+    return res
+
+
+def test_static(place,
+                logit_np,
+                label_np,
+                normalizer_np,
+                alpha=0.25,
+                gamma=2.0,
+                reduction='sum'):
+    paddle.enable_static()
+    prog = paddle.static.Program()
+    startup_prog = paddle.static.Program()
+    with paddle.static.program_guard(prog, startup_prog):
+        logit = paddle.data(name='logit', shape=logit_np.shape, dtype='float64')
+        label = paddle.data(name='label', shape=label_np.shape, dtype='float64')
+        feed_dict = {"logit": logit_np, "label": label_np}
+
+        normalizer = None
+        if normalizer_np is not None:
+            normalizer = paddle.data(
+                name='normalizer', shape=normalizer_np.shape, dtype='float64')
+            feed_dict["normalizer"] = normalizer_np
+
+        res = call_sfl_functional(logit, label, normalizer, alpha, gamma,
+                                  reduction)
+        exe = paddle.static.Executor(place)
+        static_result = exe.run(prog, feed=feed_dict, fetch_list=[res])
+    return static_result
+
+
+def test_dygraph(place,
+                 logit_np,
+                 label_np,
+                 normalizer_np,
+                 alpha=0.25,
+                 gamma=2.0,
+                 reduction='sum'):
+    paddle.disable_static()
+    logit = paddle.to_tensor(logit_np)
+    label = paddle.to_tensor(label_np)
+    normalizer = None
+    if normalizer_np is not None:
+        normalizer = paddle.to_tensor(normalizer_np)
+    dy_res = call_sfl_functional(logit, label, normalizer, alpha, gamma,
+                                 reduction)
+    dy_result = dy_res.numpy()
+    paddle.enable_static()
+    return dy_result
+
+
+def calc_sigmoid_focal_loss(logit_np,
+                            label_np,
+                            normalizer_np,
+                            alpha=0.25,
+                            gamma=2.0,
+                            reduction='sum'):
+
+    loss = np.maximum(
+        logit_np,
+        0) - logit_np * label_np + np.log(1 + np.exp(-np.abs(logit_np)))
+
+    pred = 1 / (1 + np.exp(-logit_np))
+    p_t = pred * label_np + (1 - pred) * (1 - label_np)
+
+    if alpha is not None:
+        alpha_t = alpha * label_np + (1 - alpha) * (1 - label_np)
+        loss = alpha_t * loss
+
+    if gamma is not None:
+        loss = loss * ((1 - p_t)**gamma)
+
+    if normalizer_np is not None:
+        loss = loss / normalizer_np
+
+    if reduction == 'mean':
+        loss = np.mean(loss)
+    elif reduction == 'sum':
+        loss = np.sum(loss)
+
+    return loss
+
+
+class TestSigmoidFocalLoss(unittest.TestCase):
+    def test_SigmoidFocalLoss(self):
+        logit_np = np.random.uniform(
+            0.1, 0.8, size=(2, 3, 4, 10)).astype(np.float64)
+        label_np = np.random.randint(
+            0, 2, size=(2, 3, 4, 10)).astype(np.float64)
+        normalizer_nps = [
+            np.asarray(
+                [np.sum(label_np > 0)], dtype=label_np.dtype), None
+        ]
+        places = [fluid.CPUPlace()]
+        if fluid.core.is_compiled_with_cuda():
+            places.append(fluid.CUDAPlace(0))
+        reductions = ['sum', 'mean', 'none']
+        alphas = [0.25, 0.5]
+        gammas = [3, 0.]
+        for place in places:
+            for reduction in reductions:
+                for alpha in alphas:
+                    for gamma in gammas:
+                        for normalizer_np in normalizer_nps:
+                            static_result = test_static(place, logit_np,
+                                                        label_np, normalizer_np,
+                                                        alpha, gamma, reduction)
+                            dy_result = test_dygraph(place, logit_np, label_np,
+                                                     normalizer_np, alpha,
+                                                     gamma, reduction)
+                            expected = calc_sigmoid_focal_loss(
+                                logit_np, label_np, normalizer_np, alpha, gamma,
+                                reduction)
+                            self.assertTrue(
+                                np.allclose(static_result, expected))
+                            self.assertTrue(
+                                np.allclose(static_result, dy_result))
+                            self.assertTrue(np.allclose(dy_result, expected))
+
+    def test_SigmoidFocalLoss_error(self):
+        paddle.disable_static()
+        logit = paddle.to_tensor([[0.97], [0.91], [0.03]], dtype='float32')
+        label = paddle.to_tensor([[1.0], [1.0], [0.0]], dtype='float32')
+        self.assertRaises(
+            ValueError,
+            paddle.nn.functional.sigmoid_focal_loss,
+            logit=logit,
+            label=label,
+            normalizer=None,
+            reduction="unsupport reduction")
+        paddle.enable_static()
+
+
+if __name__ == "__main__":
+    unittest.main()

python/paddle/nn/functional/loss.py

@@ -32,7 +32,6 @@ from ...fluid.layers import npair_loss  #DEFINE_ALIAS
 from ...fluid.layers import rank_loss  #DEFINE_ALIAS
 from ...fluid.layers import reshape
 from ...fluid.layers import sigmoid_cross_entropy_with_logits  #DEFINE_ALIAS
-from ...fluid.layers import sigmoid_focal_loss  #DEFINE_ALIAS
 from ...fluid.layers import smooth_l1  #DEFINE_ALIAS
 from ...fluid.layers import softmax_with_cross_entropy  #DEFINE_ALIAS
 from ...fluid.layers import square_error_cost  #DEFINE_ALIAS
@@ -1151,3 +1150,165 @@ def cross_entropy(input,
         out = reshape(out, shape=out_shape)
 
     return out
+
+
+def sigmoid_focal_loss(logit,
+                       label,
+                       normalizer=None,
+                       alpha=0.25,
+                       gamma=2.0,
+                       reduction='sum',
+                       name=None):
+    """
+    `Focal Loss <https://arxiv.org/abs/1708.02002>`_ is proposed to address the
+    foreground-background class imbalance for classification tasks. It down-weights
+    easily-classified examples and thus focuses training on hard examples. For example,
+    it is used in one-stage object detection where the foreground-background class
+    imbalance is extremely high.
+
+    This operator measures focal loss function as follows: 
+
+    .. math::
+           Out = -Labels * alpha * {(1 - \\sigma(Logit))}^{gamma}\\log(\\sigma(Logit)) - (1 - Labels) * (1 - alpha) * {\\sigma(Logit)}^{gamma}\\log(1 - \\sigma(Logit))
+
+    We know that :math:`\\sigma(Logit) = \\frac{1}{1 + \\exp(-Logit)}`. 
+
+    Then, if :attr:`normalizer` is not None, this operator divides the
+    normalizer tensor on the loss `Out`:
+
+    .. math::
+           Out = \\frac{Out}{normalizer}
+
+    Finally, this operator applies reduce operation on the loss.
+    If :attr:`reduction` set to ``'none'``, the operator will return the original loss `Out`.
+    If :attr:`reduction` set to ``'mean'``, the reduced mean loss is :math:`Out = MEAN(Out)`.
+    If :attr:`reduction` set to ``'sum'``, the reduced sum loss is :math:`Out = SUM(Out)`.
+
+    Note that the target ``label`` is 0 for the negative class and is 1 for the positive class.
+
+    Args:
+        logit (Tensor): The input logit tensor. The shape is [N, *], where N is batch_size,
+            `*` means any number of additional dimensions. The ``logit`` is usually the
+            output of a convolution layer. Available dtype is float32, float64.
+        label (Tensor): The target label tensor with the same shape as
+            ``logit``. The target label whose value should be numbers between 0 and 1.
+            Available dtype is float32, float64.
+        normalizer (Tensor, optional): The number normalizes the focal loss. It has to be
+            a 1-D Tensor whose shape is `[1, ]`. The data type is float32, float64.
+            For object detection task, it is the the number of positive samples.
+            If set to None, the focal loss will not be normalized. Default is None.
+        alpha(int|float, optional): Hyper-parameter to balance the positive and negative example,
+            it should be between 0 and 1.  Default value is set to 0.25. 
+        gamma(int|float, optional): Hyper-parameter to modulate the easy and hard examples.
+            Default value is set to 2.0.
+        reduction (str, optional): Indicate how to average the loss by batch_size,
+            the candicates are ``'none'`` | ``'mean'`` | ``'sum'``.
+            If :attr:`reduction` is ``'none'``, the unreduced loss is returned;
+            If :attr:`reduction` is ``'mean'``, the reduced mean loss is returned;
+            If :attr:`reduction` is ``'sum'``, the summed loss is returned.
+            Default is ``'sum'``.
+        name (str, optional): Name for the operation (optional, default is None).
+            For more information, please refer to :ref:`api_guide_Name`.
+
+    Returns:
+        Tensor, if :attr:`reduction` is ``'mean'`` or ``'sum'``, the out shape is :math:`[1]`, otherwise the shape is the same as ``logit``. The same dtype as ``logit`` tensor.
+
+    Examples:
+
+        .. code-block:: python
+
+            import paddle
+
+            logit = paddle.to_tensor([[0.97, 0.91, 0.03], [0.55, 0.43, 0.71]], dtype='float32')
+            label = paddle.to_tensor([[1.0, 0.0, 0.0], [0.0, 1.0, 0.0]], dtype='float32')
+            one = paddle.to_tensor([1.], dtype='float32')
+            fg_label = paddle.greater_equal(label, one)
+            fg_num = paddle.reduce_sum(paddle.cast(fg_label, dtype='float32'))
+            output = paddle.nn.functional.sigmoid_focal_loss(logit, label, normalizer=fg_num)
+            print(output.numpy())  # [0.65782464]
+
+    """
+    if reduction not in ['sum', 'mean', 'none']:
+        raise ValueError(
+            "The value of 'reduction' in sigmoid_focal_loss "
+            "should be 'sum', 'mean' or 'none', but received %s, which is not allowed."
+            % reduction)
+
+    if normalizer is not None:
+        fluid.data_feeder.check_variable_and_dtype(normalizer, 'normalizer',
+                                                   ['float32', 'float64'],
+                                                   'sigmoid_focal_loss')
+        normalizer_shape = list(normalizer.shape)
+        normalizer_dims = len(normalizer_shape)
+        if normalizer_dims > 1:
+            raise ValueError(
+                "Expected one dimension of normalizer in sigmoid_focal_loss but got {}.".
+                format(normalizer_dims))
+
+    if in_dygraph_mode():
+        one = _varbase_creator(dtype=logit.dtype)
+        core.ops.fill_constant(one, 'value',
+                               float(1.0), 'force_cpu', False, 'dtype',
+                               one.dtype, 'str_value', '1.0', 'shape',
+                               logit.shape)
+        loss = core.ops.sigmoid_cross_entropy_with_logits(logit, label)
+        pred = core.ops.sigmoid(logit)
+        p_t = core.ops.elementwise_add(
+            core.ops.elementwise_mul(pred, label),
+            core.ops.elementwise_mul(
+                core.ops.elementwise_sub(one, pred),
+                core.ops.elementwise_sub(one, label)))
+
+        alpha = fluid.dygraph.base.to_variable([alpha], dtype=loss.dtype)
+        alpha_t = core.ops.elementwise_add(
+            core.ops.elementwise_mul(alpha, label),
+            core.ops.elementwise_mul(
+                core.ops.elementwise_sub(one, alpha),
+                core.ops.elementwise_sub(one, label)))
+        loss = core.ops.elementwise_mul(alpha_t, loss)
+
+        gamma = fluid.dygraph.base.to_variable([gamma], dtype=loss.dtype)
+        gamma_t = core.ops.elementwise_pow(
+            core.ops.elementwise_sub(one, p_t), gamma)
+        loss = core.ops.elementwise_mul(gamma_t, loss)
+
+        if normalizer is not None:
+            loss = core.ops.elementwise_div(loss, normalizer)
+
+        if reduction == "sum":
+            return core.ops.reduce_sum(loss, 'reduce_all', True)
+        elif reduction == "mean":
+            return core.ops.mean(loss)
+
+        return loss
+
+    fluid.data_feeder.check_variable_and_dtype(
+        logit, 'logit', ['float32', 'float64'], 'sigmoid_focal_loss')
+    fluid.data_feeder.check_variable_and_dtype(
+        label, 'label', ['float32', 'float64'], 'sigmoid_focal_loss')
+
+    bce_name = None
+    if reduction == 'none' and normalizer is None:
+        bce_name = name
+    loss = paddle.nn.functional.binary_cross_entropy_with_logits(
+        logit, label, reduction='none', name=bce_name)
+
+    pred = fluid.layers.sigmoid(logit)
+    p_t = pred * label + (1 - pred) * (1 - label)
+
+    alpha_t = alpha * label + (1 - alpha) * (1 - label)
+    loss = paddle.multiply(alpha_t, loss)
+
+    gamma_t = paddle.pow((1 - p_t), gamma)
+    loss = paddle.multiply(gamma_t, loss)
+
+    if normalizer is not None:
+        normalizer_name = name if reduction == 'none' else None
+        loss = paddle.divide(loss, normalizer, name=normalizer_name)
+
+    if reduction == 'mean':
+        loss = paddle.mean(loss, name=name)
+    elif reduction == 'sum':
+        loss = paddle.sum(loss, name=name)
+
+    return loss