Update app.py

app.py

@@ -119,10 +119,10 @@ def predict(_query, _chatbot, _task_history):
     _task_history.append((_query, full_response))
     print(f"Qwen-7B-Chat: {_parse_text(full_response)}")
 
-
+@spaces.gpu
 def read_text(text):
     print("___Tekst do przeczytania!")
-    inputs = tokenizer_tss(text, return_tensors="pt").to("cuda")
+    inputs = tokenizer_tss(text, return_tensors="pt")
     with torch.no_grad():
         output = tts_model(**inputs).waveform.squeeze().cpu().numpy()
     sf.write('temp_file.wav', output, tts_model.config.sampling_rate)