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/xglm/test_modeling_xglm.py

@@ -19,7 +19,13 @@ import math
 import unittest
 
 from transformers import XGLMConfig, is_torch_available
-from transformers.testing_utils import require_torch, require_torch_gpu, slow, torch_device
+from transformers.testing_utils import (
+    require_torch,
+    require_torch_accelerator,
+    require_torch_fp16,
+    slow,
+    torch_device,
+)
 
 from ...generation.test_utils import GenerationTesterMixin
 from ...test_configuration_common import ConfigTester
@@ -492,7 +498,8 @@ class XGLMModelLanguageGenerationTest(unittest.TestCase):
         duration = datetime.datetime.now() - start
         self.assertGreater(duration, datetime.timedelta(seconds=1.25 * MAX_TIME))
 
-    @require_torch_gpu
+    @require_torch_accelerator
+    @require_torch_fp16
     def test_batched_nan_fp16(self):
         model_name = "facebook/xglm-564M"
         tokenizer = XGLMTokenizer.from_pretrained(model_name, use_fast=False, padding_side="left")