device agnostic models testing (#27146)

* device agnostic models testing

* add decorator `require_torch_fp16`

* make style

* apply review suggestion

* Oops, the fp16 decorator was misused

tests/models/vit/test_modeling_vit.py

@@ -22,7 +22,8 @@ from transformers import ViTConfig
 from transformers.testing_utils import (
     require_accelerate,
     require_torch,
-    require_torch_gpu,
+    require_torch_accelerator,
+    require_torch_fp16,
     require_vision,
     slow,
     torch_device,
@@ -316,7 +317,8 @@ class ViTModelIntegrationTest(unittest.TestCase):
 
     @slow
     @require_accelerate
-    @require_torch_gpu
+    @require_torch_accelerator
+    @require_torch_fp16
     def test_inference_fp16(self):
         r"""
         A small test to make sure that inference work in half precision without any problem.