Test2

app.py

@@ -11,7 +11,7 @@ text = "some example text in the English language"
 def greet(text):
     inputs = tokenizer(text, return_tensors="pt")
     with torch.no_grad():
-        output = model(**inputs).waveform
+        output = model(**inputs, decoder_input_ids=inputs["input_ids"]).waveform
         out = output[0]
         wavfile.write("tmp.wav", rate=16000, data=out)
         return open("tmp.wav", "rb").read()