[cherry-pick] add input type and dtype check for elu_op (#20106) fix...

[cherry-pick] add input type and dtype check for  elu_op (#20106) fix GradientClipByNorm english doc (#20224) (#20457)

* fix GradientClipByNorm english doc (#20224)

* fix GradientClipByNorm english doc
test=develop
test=document_fix

* add input type and dtype check for  elu_op (#20106)

* elu input check
test=develop

* test=develop

paddle/fluid/API.spec

@@ -1110,7 +1110,7 @@ paddle.fluid.clip.ErrorClipByValue ('paddle.fluid.clip.ErrorClipByValue', ('docu
 paddle.fluid.clip.ErrorClipByValue.__init__ (ArgSpec(args=['self', 'max', 'min'], varargs=None, keywords=None, defaults=(None,)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
 paddle.fluid.clip.GradientClipByValue ('paddle.fluid.clip.GradientClipByValue', ('document', 'b7a22f687269cae0c338ef3866322db7'))
 paddle.fluid.clip.GradientClipByValue.__init__ (ArgSpec(args=['self', 'max', 'min'], varargs=None, keywords=None, defaults=(None,)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
-paddle.fluid.clip.GradientClipByNorm ('paddle.fluid.clip.GradientClipByNorm', ('document', 'a5c23d96a3d8c8c1183e9469a5d0d52e'))
+paddle.fluid.clip.GradientClipByNorm ('paddle.fluid.clip.GradientClipByNorm', ('document', '93d62f284d2cdb87f2723fcc63d818f9'))
 paddle.fluid.clip.GradientClipByNorm.__init__ (ArgSpec(args=['self', 'clip_norm'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
 paddle.fluid.clip.GradientClipByGlobalNorm ('paddle.fluid.clip.GradientClipByGlobalNorm', ('document', 'ef50acbe212101121d4b82f693ec1733'))
 paddle.fluid.clip.GradientClipByGlobalNorm.__init__ (ArgSpec(args=['self', 'clip_norm', 'group_name'], varargs=None, keywords=None, defaults=('default_group',)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))

python/paddle/fluid/clip.py

@@ -184,35 +184,77 @@ class GradientClipByValue(BaseGradientClipAttr):
 
 class GradientClipByNorm(BaseGradientClipAttr):
     """ 
-    Clips tensor values to a maximum L2-norm.
+    Convert the input multidimensional Tensor :math:`X` to a multidimensional Tensor whose L2 norm does not exceed the given two-norm maximum ( :math:`clip\_norm` ). 
 
-    This operator limits the L2 norm of the input :math:`X` within :math:`max\_norm`.
-    If the L2 norm of :math:`X` is less than or equal to :math:`max\_norm`, :math:`Out`
-    will be the same as :math:`X`. If the L2 norm of :math:`X` is greater than
-    :math:`max\_norm`, :math:`X` will be linearly scaled to make the L2 norm of
-    :math:`Out` equal to :math:`max\_norm`, as shown in the following formula:
+    The tensor is not passed through this class, but passed through the parametre of ``main_program`` in ``fluid.program_guard``.
+
+    This class limits the L2 norm of the input :math:`X` within :math:`clip\_norm`.
 
     .. math::
+        Out =
+  	\\left \{
+  	\\begin{aligned}
+  	& X & & if (norm(X) \\leq clip\_norm) \\\\
+  	& \\frac{clip\_norm*X}{norm(X)} & & if (norm(X) > clip\_norm) \\\\
+  	\\end{aligned}
+  	\\right.
 
-        Out = \\frac{max\_norm * X}{norm(X)},
 
     where :math:`norm(X)` represents the L2 norm of :math:`X`.
 
+    .. math::
+ 	norm(X) = ( \\sum_{i=1}^{n}|x\_i|^2)^{ \\frac{1}{2}}
+
     Args:
-        clip_norm (float): The maximum norm value
+        clip_norm(float): The maximum norm value
     
     Examples:
         .. code-block:: python
 
             import paddle.fluid as fluid
-            w_param_attrs = fluid.ParamAttr(name=None,
-              initializer=fluid.initializer.UniformInitializer(low=-1.0, high=1.0, seed=0),
-              learning_rate=1.0,
-              regularizer=fluid.regularizer.L1Decay(1.0),
-              trainable=True,
-              gradient_clip=fluid.clip.GradientClipByNorm(clip_norm=2.0))
-            x = fluid.layers.data(name='x', shape=[10], dtype='float32')
-            y_predict = fluid.layers.fc(input=x, size=1, param_attr=w_param_attrs)
+            import paddle.fluid.core as core
+            import paddle
+            place = core.CPUPlace()
+            prog = fluid.framework.Program()
+            startup_program = fluid.framework.Program()
+            with fluid.program_guard(
+                        main_program=prog, startup_program=startup_program):
+                image = fluid.data(name='x', shape=[None, 784], dtype='float32', lod_level=0)
+                label = fluid.data(name='y', shape=[None, 1], dtype='int64', lod_level=0)
+                hidden1 = fluid.layers.fc(input=image, size=128, act='relu')
+                hidden2 = fluid.layers.fc(input=hidden1, size=64, act='relu')
+                predict = fluid.layers.fc(input=hidden2, size=10, act='softmax')
+                cost = fluid.layers.cross_entropy(input=predict, label=label)
+                avg_cost = fluid.layers.mean(cost)
+            prog_clip = prog.clone()
+            avg_cost_clip = prog_clip.block(0).var(avg_cost.name)
+            p_g = fluid.backward.append_backward(loss=avg_cost)
+            p_g_clip = fluid.backward.append_backward(loss=avg_cost_clip)
+            with fluid.program_guard(main_program=prog_clip, startup_program=startup_program):
+                fluid.clip.set_gradient_clip(
+                    fluid.clip.GradientClipByNorm(clip_norm=2.0))
+                p_g_clip = fluid.clip.append_gradient_clip_ops(p_g_clip)
+            grad_list = [elem[1] for elem in p_g]
+            grad_clip_list = [elem[1] for elem in p_g_clip]
+            train_reader = paddle.batch( 
+                paddle.reader.shuffle(
+                    paddle.dataset.mnist.train(), buf_size=8192),
+                batch_size=128)
+  
+            exe = fluid.Executor(place)
+            feeder = fluid.DataFeeder(feed_list=[image, label], place=place)
+            exe.run(startup_program)
+  
+            count = 0
+            for data in train_reader():
+                count += 1
+                print("count:%s" % count)
+                if count > 5:
+                   break
+                out = exe.run(prog, feed=feeder.feed(data), fetch_list=grad_list)
+                out_clip = exe.run(prog_clip,
+                                   feed=feeder.feed(data),
+                                   fetch_list=grad_clip_list)
 
     """
 

python/paddle/fluid/layers/nn.py

@@ -11542,6 +11542,17 @@ def elu(x, alpha=1.0, name=None):
                 # [ 1.          15.6       ]]
     """
     helper = LayerHelper('elu', **locals())
+    if not isinstance(x, Variable):
+        raise TypeError(
+            "The type of 'x' in elu must be Variable, but received %s" %
+            (type(x)))
+    if convert_dtype(x.dtype) in ['float16']:
+        warnings.warn(
+            "The data type of 'x' in elu only support float16 in GPU now.")
+    if convert_dtype(x.dtype) not in ['float16', 'float32', 'float64']:
+        raise TypeError(
+            "The data type of 'x' in elu must be float16 (only support on GPU), float32 or float64, but received %s."
+            % (convert_dtype(x.dtype)))
     out = helper.create_variable_for_type_inference(dtype=x.dtype)
     helper.append_op(
         type='elu',

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

@@ -19,6 +19,8 @@ import numpy as np
 import paddle.fluid.core as core
 from op_test import OpTest
 from scipy.special import expit, erf
+import paddle.fluid as fluid
+from paddle.fluid import compiler, Program, program_guard
 
 
 class TestActivation(OpTest):
@@ -519,6 +521,18 @@ class TestELU(TestActivation):
         self.check_grad(['X'], 'Out', max_relative_error=0.02)
 
 
+class TestELUOpError(OpTest):
+    def test_errors(self):
+        with program_guard(Program(), Program()):
+            # The input type of elu_op must be Variable.
+            x1 = fluid.create_lod_tensor(
+                np.array([[-1]]), [[1]], fluid.CPUPlace())
+            self.assertRaises(TypeError, fluid.layers.elu, x1)
+            # The input dtype of elu_op must be float16 float32 or float64.
+            x2 = fluid.layers.data(name='x2', shape=[4], dtype="int32")
+            self.assertRaises(TypeError, fluid.layers.elu, x2)
+
+
 class TestReciprocal(TestActivation):
     def setUp(self):
         self.op_type = "reciprocal"