whisper[本地] largeV3 不能正常语音转录，进度会卡住不动 #64
Description
问题描述 | Problem Description
whisper largeV3 转换一个几分钟的视频时没有任何反应(显卡为3070ti laptop)，但看任务管理器显示显存仍然在被很高的占用(但有且只有显存占用，却没有3D占用，正常转录语音时，3D占用基本是拉满的)
等了个把小时依然没有任何反应，看日志也一直处于停在语音转录的那个步骤，才意识到卡了，然后whisper模型切换到medium重新开始任务，就正常能转录了。
还没有试largeV1和largeV2会不会这样，反正目前medium是没问题
日志信息（可选）| Logs (Optional)
===================
开始创建文件任务：C:/Users/GJG/Desktop/Grand Theft Auto VI Trailer 1.mp4
获取视频信息执行命令: ffmpeg -i C:/Users/GJG/Desktop/Grand Theft Auto VI Trailer 1.mp4
视频时长: 90.09秒
执行命令: ffmpeg -ss 00:00:27.027 -i C:/Users/GJG/Desktop/Grand Theft Auto VI Trailer 1.mp4 -vframes 1 -q:v 2 -y D:/installfiles/VideoCaptioner/work-dir/Grand Theft Auto VI Trailer 1/thumbnail.jpg
文件任务创建完成：Task(id=0, queued_at=datetime.datetime(2024, 11, 25, 8, 51, 47, 619746), started_at=datetime.datetime(2024, 11, 25, 8, 51, 47, 619746), completed_at=None, status=<Status.PENDING: '待处理'>, fraction_downloaded=0, work_dir='D:\\installfiles\\VideoCaptioner\\work-dir\\Grand Theft Auto VI Trailer 1', file_path='C:\\Users\\GJG\\Desktop\\Grand Theft Auto VI Trailer 1.mp4', url='', source=<Source.FILE_IMPORT: '文件导入'>, original_language=None, target_language='简体中文', video_info=VideoInfo(file_name='Grand Theft Auto VI Trailer 1', width=1920, height=1080, fps=30.0, duration_seconds=90.09, bitrate_kbps=2059, video_codec='av1', audio_codec='aac', audio_sampling_rate=44100, thumbnail_path='D:\\installfiles\\VideoCaptioner\\work-dir\\Grand Theft Auto VI Trailer 1\\thumbnail.jpg'), audio_format='mp3', audio_save_path='D:\\installfiles\\VideoCaptioner\\work-dir\\Grand Theft Auto VI Trailer 1\\Grand Theft Auto VI Trailer 1.wav', transcribe_model=<TranscribeModelEnum.WHISPER: 'Whisper [本地]'>, transcribe_language='en', use_asr_cache=True, need_word_time_stamp=False, original_subtitle_save_path='D:\\installfiles\\VideoCaptioner\\work-dir\\Grand Theft Auto VI Trailer 1\\subtitle\\【原始字幕】Whisper [本地]large-v3-English.srt', whisper_model='large-v3', whisper_api_key='', whisper_api_base='', whisper_api_model='', whisper_api_prompt='', base_url='https://api.hunyuan.cloud.tencent.com/v1', api_key='sk-F6B0ontcGY5nFlP7VoDnO2BeNV4HTpi7p2xfrwZZ5NNLvgrm', llm_model='hunyuan-lite', need_translate=True, need_optimize=False, result_subtitle_save_path='D:\\installfiles\\VideoCaptioner\\work-dir\\Grand Theft Auto VI Trailer 1\\subtitle\\【翻译字幕】样式字幕.ass', thread_num=10, batch_size=10, subtitle_layout='译文在上', video_save_path='D:\\installfiles\\VideoCaptioner\\work-dir\\Grand Theft Auto VI Trailer 1\\【卡卡】Grand Theft Auto VI Trailer 1.mp4', soft_subtitle=True, subtitle_style_srt='[V4+ Styles]\nFormat: Name,Fontname,Fontsize,PrimaryColour,SecondaryColour,OutlineColour,BackColour,Bold,Italic,Underline,StrikeOut,ScaleX,ScaleY,Spacing,Angle,BorderStyle,Outline,Shadow,Alignment,MarginL,MarginR,MarginV,Encoding\nStyle: Default,微软雅黑,40,&H00fff0fd,&H000000FF,&H00000000,&H00000000,-1,0,0,0,100,100,2.0,0,1,1.8,0,2,10,10,18,1\nStyle: Secondary,微软雅黑,26,&H00ffffff,&H000000FF,&H00000000,&H00000000,-1,0,0,0,100,100,0.2,0,1,2.0,0,2,10,10,18,1')
===========转录任务开始===========
时间：2024-11-25 08:51:47.637740
开始转换音频
转换为音频执行命令: ffmpeg -i C:\Users\GJG\Desktop\Grand Theft Auto VI Trailer 1.mp4 -map 0:a -ac 1 -f mp3 -ar 16000 -af aresample=async=1 -y D:\installfiles\VideoCaptioner\work-dir\Grand Theft Auto VI Trailer 1\Grand Theft Auto VI Trailer 1.wav
开始语音转录
找到模型文件: D:\installfiles\VideoCaptioner\AppData\models\ggml-large-v3.bin
WhisperCPP 执行命令: whisper-cpp -m D:\installfiles\VideoCaptioner\AppData\models\ggml-large-v3.bin C:\Users\GJG\AppData\Local\Temp\bk_asr\20241125085147Grand Theft Auto VI Trailer 1.wav -l en -osrt
音频总时长: 90 秒
=========================以上是历史日志=========================