fix fp16 for tile op (#50913)

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

@@ -366,6 +366,20 @@ class TestTileAPI_ZeroDim(unittest.TestCase):
         paddle.enable_static()
 
 
+class Testfp16TileOp(unittest.TestCase):
+    def testfp16(self):
+        input_x = (np.random.random([1, 2, 3])).astype('float16')
+        with paddle.static.program_guard(paddle.static.Program()):
+            x = paddle.static.data(name="x", shape=[1, 2, 3], dtype='float16')
+            repeat_times = [2, 2]
+            out = paddle.tile(x, repeat_times=repeat_times)
+            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__":
     paddle.enable_static()
     unittest.main()

python/paddle/tensor/manipulation.py

@@ -3084,7 +3084,7 @@ def tile(x, repeat_times, name=None):
     Both the number of dimensions of ``x`` and the number of elements in ``repeat_times`` should be less than or equal to 6.
 
     Args:
-        x (Tensor): The input tensor, its data type should be bool, float32, float64, int32 or int64.
+        x (Tensor): The input tensor, its data type should be bool, float16, float32, float64, int32 or int64.
         repeat_times (list|tuple|Tensor): The number of repeating times. If repeat_times is a list or tuple, all its elements
             should be integers or 1-D Tensors with the data type int32. If repeat_times is a Tensor, it should be an 1-D Tensor with the data type int32.
         name (str, optional): Name for the operation (optional, default is None). For more information, please refer to :ref:`api_guide_Name`.
@@ -3145,7 +3145,10 @@ def tile(x, repeat_times, name=None):
                     ), 'Elements in repeat_times must be 1-D Tensors or integers.'
 
         check_variable_and_dtype(
-            x, 'x', ['bool', 'float32', 'float64', 'int32', 'int64'], 'tile'
+            x,
+            'x',
+            ['bool', 'float16', 'float32', 'float64', 'int32', 'int64'],
+            'tile',
         )
         if convert_dtype(x.dtype) == 'bool' and not x.stop_gradient:
             raise ValueError(