TST test_mixed_adapter_batches_lora_opt_timing on XPU (huggingface#2021)

tests/test_custom_models.py

@@ -52,7 +52,7 @@
 from peft.utils import ModulesToSaveWrapper, infer_device
 
 from .testing_common import PeftCommonTester
-from .testing_utils import get_state_dict, require_torch_gpu
+from .testing_utils import get_state_dict, require_non_cpu
 
 
 # MLP is a vanilla FF network with only linear layers
@@ -3276,7 +3276,7 @@ def test_mixed_adapter_batches_lora_with_dora_raises(self):
         with pytest.raises(ValueError, match=msg):
             peft_model.forward(**inputs)
 
-    @require_torch_gpu
+    @require_non_cpu
     def test_mixed_adapter_batches_lora_opt_timing(self):
         # Use a more realistic model (opt-125m) and do a simple runtime check to ensure that mixed adapter batches
         # don't add too much overhead. These types of tests are inherently flaky, so we try to add in some robustness.