Move qwen_omni_utils import after loading processor to enable remote code

transcript.py

 import argparse
 import torch
 from transformers import Qwen2_5OmniForConditionalGeneration, Qwen2_5OmniProcessor
-from qwen_omni_utils import process_mm_info
 from resemblyzer import VoiceEncoder
 from sklearn.cluster import AgglomerativeClustering
 import webrtcvad
@@ -78,6 +77,9 @@ def main():
     model = Qwen2_5OmniForConditionalGeneration.from_pretrained("Qwen/Qwen2.5-Omni-7B", torch_dtype="auto", device_map="auto", trust_remote_code=True)
     processor = Qwen2_5OmniProcessor.from_pretrained("Qwen/Qwen2.5-Omni-7B", trust_remote_code=True)
 
+    # Import utils after loading model
+    from qwen_omni_utils import process_mm_info
+
     # Load audio
     audio, sr = librosa.load(audio_file, sr=16000)