[tests] enable automatic speech recognition pipeline tests on XPU (#2…

…9308)

* use require_torch_gpu

* enable on XPU

tests/pipelines/test_pipelines_automatic_speech_recognition.py

@@ -1221,15 +1221,15 @@ def test_whisper_longform(self):
 
         processor = AutoProcessor.from_pretrained("openai/whisper-tiny.en")
         model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-tiny.en")
-        model = model.to("cuda")
+        model = model.to(torch_device)
 
         pipe = pipeline(
             "automatic-speech-recognition",
             model=model,
             tokenizer=processor.tokenizer,
             feature_extractor=processor.feature_extractor,
             max_new_tokens=128,
-            device="cuda:0",
+            device=torch_device,
         )
 
         ds = load_dataset("distil-whisper/meanwhile", "default")["test"]
@@ -1246,7 +1246,7 @@ def test_seamless_v2(self):
         pipe = pipeline(
             "automatic-speech-recognition",
             model="facebook/seamless-m4t-v2-large",
-            device="cuda:0",
+            device=torch_device,
         )
 
         dataset = load_dataset("hf-internal-testing/librispeech_asr_dummy", "clean", split="validation")