refactor: Allow synthesize() to use gpu if available

app.py

@@ -165,27 +165,25 @@ async def translate_text(text: str,
         "tgtLang": tgtLang
     }
     return result_dict
-
 @app.post("/synthesize/")
 async def synthesize(text: str):
     model = VitsModel.from_pretrained("facebook/mms-tts-tgl")
-    tokenizer = AutoTokenizer.from_pretrained("facebook/mms-tts-tgl")
-    
+    tokenizer = VitsTokenizer.from_pretrained("facebook/mms-tts-tgl")
+
+    device = "cuda" if torch.cuda.is_available() else "cpu"
+    model.to(device)
+
     inputs = tokenizer(text, return_tensors="pt")
-    
+    input_ids = inputs["input_ids"].to(device)
+
+
     with torch.no_grad():
-        output = model(**inputs).waveform
-    
-    data_np = output.numpy()
-    data_np_squeezed = np.squeeze(data_np)
-    
+        outputs = model(input_ids)
+
+    speech = outputs["waveform"]
     temp_file = create_temp_filename()
 
-    scipy.io.wavfile.write(
-        temp_file, 
-        rate=model.config.sampling_rate,
-        data=data_np_squeezed
-        )    
+    torchaudio.save(temp_file, speech.cpu(), 16000) 
     logging.info(f"Synthesizing completed for text: {text}")
 
     return FileResponse(