Fix model loading to use AutoModelForCausalLM for Qwen2.5-Omni-7B

transcript.py

 import argparse
 import torch
-from transformers import AutoProcessor, AutoModel, BitsAndBytesConfig
+from transformers import AutoProcessor, AutoModelForCausalLM, BitsAndBytesConfig
 from resemblyzer import VoiceEncoder
 from sklearn.cluster import AgglomerativeClustering
 import webrtcvad
@@ -76,7 +76,7 @@ def main():
     # Load Qwen2.5-Omni-7B model with 4-bit quantization
     quantization_config = BitsAndBytesConfig(load_in_4bit=True)
     processor = AutoProcessor.from_pretrained("Qwen/Qwen2.5-Omni-7B")
-    model = AutoModel.from_pretrained("Qwen/Qwen2.5-Omni-7B", quantization_config=quantization_config, device_map="auto")
+    model = AutoModelForCausalLM.from_pretrained("Qwen/Qwen2.5-Omni-7B", quantization_config=quantization_config, device_map="auto")
 
     # Load audio
     audio, sr = librosa.load(audio_file, sr=16000)