added format

handler.py

@@ -55,7 +55,9 @@ class EndpointHandler:
             enable_text_splitting=True,
         )
         audio_file = io.BytesIO()
-        torchaudio.save(audio_file, torch.tensor(out["wav"]).unsqueeze(0), 24000)
+        torchaudio.save(
+            audio_file, torch.tensor(out["wav"]).unsqueeze(0), 24000, format="wav"
+        )
         inference_time = time.time() - t0
         print(f"I: Time to generate audio: {inference_time} seconds")
         audio_str = base64.b64encode(audio_file.getvalue()).decode("utf-8")