diff --git a/python/paddle/fluid/tests/unittests/test_conj_op.py b/python/paddle/fluid/tests/unittests/test_conj_op.py
index 44698f5679b2de1975ea2f596bbee891243a8fd8..b3c24ba5f0017a658320cf3490ea83aafe86b3f3 100644
--- a/python/paddle/fluid/tests/unittests/test_conj_op.py
+++ b/python/paddle/fluid/tests/unittests/test_conj_op.py
@@ -133,5 +133,20 @@ class TestComplexConjOp(unittest.TestCase):
                     np.testing.assert_array_equal(result, target)
 
 
+class Testfp16ConjOp(unittest.TestCase):
+    def testfp16(self):
+        input_x = (
+            np.random.random((12, 14)) + 1j * np.random.random((12, 14))
+        ).astype('float16')
+        with static.program_guard(static.Program()):
+            x = static.data(name="x", shape=[12, 14], dtype='float16')
+            out = paddle.conj(x)
+            if paddle.is_compiled_with_cuda():
+                place = paddle.CUDAPlace(0)
+                exe = paddle.static.Executor(place)
+                exe.run(paddle.static.default_startup_program())
+                out = exe.run(feed={'x': input_x}, fetch_list=[out])
+
+
 if __name__ == "__main__":
     unittest.main()
diff --git a/python/paddle/tensor/math.py b/python/paddle/tensor/math.py
index 591b885439517ed2eaf9edc2b6c82710dcca2a9a..6391bb093195742d7930a506c90bf3ee307c3da7 100644
--- a/python/paddle/tensor/math.py
+++ b/python/paddle/tensor/math.py
@@ -3861,7 +3861,7 @@ def conj(x, name=None):
 
     Args:
         x (Tensor): The input Tensor which hold the complex numbers.
-            Optional data types are: complex64, complex128, float32, float64, int32 or int64.
+            Optional data types are:float16, complex64, complex128, float32, float64, int32 or int64.
         name (str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
 
     Returns:
@@ -3889,7 +3889,15 @@ def conj(x, name=None):
         check_variable_and_dtype(
             x,
             "x",
-            ['complex64', 'complex128', 'float32', 'float64', 'int32', 'int64'],
+            [
+                'complex64',
+                'complex128',
+                'float16',
+                'float32',
+                'float64',
+                'int32',
+                'int64',
+            ],
             'conj',
         )