Update app.py

app.py

@@ -107,11 +107,11 @@ def infer(user_audio) -> tuple[int, np.ndarray]:
         generator = load_csm_1b(model_path, device)
         logging.info("Sesame CSM 1B loaded successfully.")
 
-        whisper_model = whisper.load_model("small.en", device=device, dtype=torch.bfloat16)
+        whisper_model = whisper.load_model("small.en", device=device)
         logging.info("Whisper model loaded successfully.")
 
         tokenizer_gemma = AutoTokenizer.from_pretrained("google/gemma-3-1b-it")
-        model_gemma = AutoModelForCausalLM.from_pretrained("google/gemma-3-1b-it").to(device, dtype=torch.bfloat16)
+        model_gemma = AutoModelForCausalLM.from_pretrained("google/gemma-3-1b-it").to(device)
         logging.info("Gemma 3 1B pt model loaded successfully.")
 
         if not user_audio: