Duplicate sentences with large-v3

Hi, I see duplicate sentences when using large-v3, the issue can be reproduced no matter I set condition_on_previous_text to True or False, any suggestion on how to fix it? Here's the codes ``` def transcribe_audio(audio_file, beam_size=5, vad_filter=True): """Transcribe audio using Faster-Whisper""" print(f"Transcribing: {audio_file}") start_time = time.time() # 新增：记录开始时间 segments, info = whisper_model.transcribe( audio_file, beam_size=5, language=LANGUAGE, vad_filter=True, word_timestamps=True, condition_on_previous_text=False, vad_parameters=dict(min_silence_duration_ms=200), ) segments = list(segments) transcribe_time = time.time() - start_time # 新增：计算耗时 print(f"Language: {info.language} (probability: {info.language_probability:.2f})") print(f"Segments: {len(segments)}") print(f"Transcription time: {transcribe_time:.2f}s") # 新增：打印耗时 return segments, info, transcribe_time # 修改：返回值增加 transcribe_time ``` The segments output ``` TRANSCRIPTION WITH SPEAKER DIARIZATION ================================================================================ [18.62s -> 22.83s] SPEAKER_2 莫子缓缓地说道 [23.52s -> 25.68s] SPEAKER_4 恒济原将继续 [26.90s -> 29.32s] SPEAKER_4 宇宙是一台机器 [29.32s -> 33.40s] SPEAKER_4 我造出了这台机器 [33.98s -> 36.84s] SPEAKER_4 恒纪元将继续 [38.22s -> 40.82s] SPEAKER_4 宇宙是一台机器 [40.82s -> 45.02s] SPEAKER_4 我造出了这台机器 ``` It is not caused by speaker diarization as I print the segments before invoking speaker diarization