feat: add relu composite (#50819)

* feat: add relu composite rule

* feat: add relu composite rule, maximum op

* feat: add relu composite rule, maximum op

* feat: add relu composite rule, polish comments

* feat: add relu composite rule, polish comments

* feat: add relu composite rule, add python api of relu

* feat: add relu composite rule, commit hook

* fix: maximum type error & ban cinn test

* fix: maximum input sequence bugs

* resolve conflicts

* fix: code style bugs

* add: relu fp16 test

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

@@ -1711,8 +1711,11 @@ class TestRound_ZeroDim(TestRound):
 class TestRelu(TestActivation):
     def setUp(self):
         self.op_type = "relu"
+        self.python_api = paddle.nn.functional.relu
+        self.prim_op_type = "comp"
         self.init_dtype()
         self.init_shape()
+        self.skip_cinn()
 
         np.random.seed(1024)
         if self.dtype == np.uint16:
@@ -1733,13 +1736,22 @@ class TestRelu(TestActivation):
     def test_check_grad(self):
         if self.dtype == np.float16:
             return
-        self.check_grad(['X'], 'Out')
+        self.check_grad(['X'], 'Out', check_prim=True)
+
+    def test_check_output(self):
+        self.check_output(check_prim=True)
+
+    def skip_cinn(self):
+        self.enable_cinn = False
 
 
 class TestRelu_ZeroDim(TestRelu):
     def init_shape(self):
         self.shape = []
 
+    def skip_cinn(self):
+        self.enable_cinn = False
+
 
 class TestReluAPI(unittest.TestCase):
     # test paddle.nn.ReLU, paddle.nn.functional.relu
@@ -3826,7 +3838,7 @@ create_test_act_fp16_class(TestAcosh, grad_atol=0.85)
 create_test_act_fp16_class(TestAsinh, grad_atol=0.85)
 create_test_act_fp16_class(TestAtanh, grad_atol=0.85)
 create_test_act_fp16_class(TestRound, grad_check=False)
-create_test_act_fp16_class(TestRelu)
+create_test_act_fp16_class(TestRelu, check_prim=True)
 create_test_act_fp16_class(TestGelu)
 create_test_act_fp16_class(TestBRelu)
 create_test_act_fp16_class(TestRelu6)

python/paddle/incubate/autograd/composite_rules.py

@@ -330,3 +330,10 @@ def fill_any_like(x, fill_value, dtype, place=None):
     dtype = dtypes.dtype(dtype)
     val = full(x.shape, fill_value, dtype)
     return val
+
+
+@REGISTER_COMPOSITE('relu')
+def relu_composite(x):
+    """define composite rule of op relu."""
+    # relu(x) = max(x, 0)
+    return maximum(x, zeros_like(x))

python/paddle/incubate/autograd/primitives.py

@@ -60,6 +60,7 @@ from paddle.tensor import tanh  # noqa: F401
 from paddle.tensor import uniform  # noqa: F401
 from paddle.tensor import zeros  # noqa: F401
 from paddle.tensor.creation import assign  # noqa: F401
+from paddle.tensor.creation import zeros_like  # noqa: F401
 from paddle.tensor.manipulation import cast  # noqa: F401
 from paddle.tensor.math import maximum  # noqa: F401
 from paddle.tensor.math import minimum  # noqa: F401
@@ -89,9 +90,9 @@ math_op = [
     'logcumsumexp',
     'logit',
     'max',
+    'maximum',
     'min',
     'minimum',
-    'maximum'
 ]
 
 trigonometric_op = [
@@ -126,5 +127,6 @@ others = [
     'concat',
     'uniform',
     'greater_equal',
+    'zeros_like',
 ]
 """