[fp16] suppot fp16 in argmin (#50858)

python/paddle/fluid/tests/unittests/test_arg_min_max_v2_op.py

@@ -379,5 +379,18 @@ class TestArgMaxOpFp16(unittest.TestCase):
                 out = exe.run(feed={'x': x_np}, fetch_list=[out])
 
 
+class TestArgMinOpFp16(unittest.TestCase):
+    def test_fp16(self):
+        x_np = np.random.random((10, 16)).astype('float16')
+        with paddle.static.program_guard(paddle.static.Program()):
+            x = paddle.static.data(shape=[10, 16], name='x', dtype='float16')
+            out = paddle.argmin(x)
+            if core.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': x_np}, fetch_list=[out])
+
+
 if __name__ == '__main__':
     unittest.main()

python/paddle/tensor/search.py

@@ -224,7 +224,7 @@ def argmin(x, axis=None, keepdim=False, dtype="int64", name=None):
     element along the provided axis.
 
     Args:
-        x(Tensor): An input N-D Tensor with type float32, float64, int16,
+        x(Tensor): An input N-D Tensor with type float16, float32, float64, int16,
             int32, int64, uint8.
         axis(int, optional): Axis to compute indices along. The effective range
             is [-R, R), where R is x.ndim. when axis < 0, it works the same way
@@ -282,7 +282,15 @@ def argmin(x, axis=None, keepdim=False, dtype="int64", name=None):
         check_variable_and_dtype(
             x,
             'x',
-            ['float32', 'float64', 'int16', 'int32', 'int64', 'uint8'],
+            [
+                'float16',
+                'float32',
+                'float64',
+                'int16',
+                'int32',
+                'int64',
+                'uint8',
+            ],
             'paddle.argmin',
         )
         check_dtype(var_dtype, 'dtype', ['int32', 'int64'], 'argmin')