#4161 TaskCfgVTT(is_cuda=True, uuid='e6e9c2f4e7', cache_folder='E:/软件/win-pyvideotrans-v3.99-418/tmp/16876/e6e9c2f4e7', target

171.92* Posted at: 2 hours ago 👁14

语音识别阶段出错 [faster-whisper(本地)] 出错了,可能内存或显存不足 Model:large-v3-turbo GPU0
Traceback (most recent call last):
File "videotrans\configure\_base.py", line 281, in _new_process
File "videotrans\process\signelobj.py", line 80, in submit_task_gpu
File "concurrent\futures\process.py", line 720, in submit
concurrent.futures.process.BrokenProcessPool: A child process terminated abruptly, the process pool is not usable anymore

Traceback (most recent call last):
File "videotrans\configure\_base.py", line 281, in _new_process
File "videotrans\process\signelobj.py", line 80, in submit_task_gpu
File "concurrent\futures\process.py", line 720, in submit
concurrent.futures.process.BrokenProcessPool: A child process terminated abruptly, the process pool is not usable anymore

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "videotrans\task\job.py", line 105, in run
File "videotrans\task\trans_create.py", line 361, in recogn
File "videotrans\recognition\__init__.py", line 293, in run
File "videotrans\recognition\_base.py", line 143, in run
File "videotrans\recognition\_overall.py", line 33, in _exec
File "videotrans\recognition\_overall.py", line 105, in _faster
File "videotrans\configure\_base.py", line 303, in _new_process
RuntimeError: 出错了,可能内存或显存不足 Model:large-v3-turbo GPU0
Traceback (most recent call last):
File "videotrans\configure\_base.py", line 281, in _new_process
File "videotrans\process\signelobj.py", line 80, in submit_task_gpu
File "concurrent\futures\process.py", line 720, in submit
concurrent.futures.process.BrokenProcessPool: A child process terminated abruptly, the process pool is not usable anymore
TaskCfgVTT(is_cuda=True, uuid='e6e9c2f4e7', cache_folder='E:/软件/win-pyvideotrans-v3.99-418/tmp/16876/e6e9c2f4e7', target_dir='F:/下载/哔哩哔哩视频/新建文件夹/01.金鹿号-mp4', source_language='英语', source_language_code='en', source_sub='F:/下载/哔哩哔哩视频/新建文件夹/01.金鹿号-mp4/en.srt', source_wav='E:/软件/win-pyvideotrans-v3.99-418/tmp/16876/e6e9c2f4e7/en.wav', source_wav_output='F:/下载/哔哩哔哩视频/新建文件夹/01.金鹿号-mp4/en.m4a', target_language='简体中文', target_language_code='zh-cn', target_sub='F:/下载/哔哩哔哩视频/新建文件夹/01.金鹿号-mp4/zh-cn.srt', target_wav='E:/软件/win-pyvideotrans-v3.99-418/tmp/16876/e6e9c2f4e7/target.wav', target_wav_output='F:/下载/哔哩哔哩视频/新建文件夹/01.金鹿号-mp4/zh-cn.m4a', name='F:/下载/哔哩哔哩视频/01.金鹿号.mp4', noextname='01.金鹿号', basename='01.金鹿号.mp4', ext='mp4', dirname='F:/下载/哔哩哔哩视频', shound_del_name=None, translate_type=1, tts_type=0, volume='+0%', pitch='+0Hz', voice_rate='+0%', voice_role='YunJhe(Male/TW)', voice_autorate=True, video_autorate=False, remove_silent_mid=False, align_sub_audio=True, detect_language='en', recogn_type=0, model_name='large-v3-turbo', shibie_audio=None, remove_noise=False, enable_diariz=True, nums_diariz=0, rephrase=0, fix_punc=False, subtitle_language=None, app_mode='biaozhun', subtitles='', targetdir_mp4='F:/下载/哔哩哔哩视频/新建文件夹/01.金鹿号-mp4/01.金鹿号.mp4', novoice_mp4='E:/软件/win-pyvideotrans-v3.99-418/tmp/16876/e6e9c2f4e7/novoice.mp4', is_separate=False, embed_bgm=False, instrument=None, vocal=None, back_audio='', clear_cache=False, background_music=None, subtitle_type=0, only_out_mp4=False, recogn2pass=False, output_srt=0, copysrt_rawvideo=False)


=
system:Windows-10-10.0.26100-SP0
version:v3.99
frozen:True
language:zh
root_dir:E:/软件/win-pyvideotrans-v3.99-418

1 Replies

这是由于显存或内存不足导致的。错误信息明确指出在使用 faster-whisper(本地)large-v3-turbo 模型时,GPU 进程池异常终止,原因是显存或内存不足。

请按以下步骤尝试解决:

  1. 降低模型大小:在主界面的语音识别渠道中,将模型从 large-v3-turbo 更换为更小的模型,例如 mediumsmallbaselarge-v3-turbo 模型对显存和内存要求较高。
  2. 调整高级设置:进入 菜单 -> 工具/选项 -> 高级选项,在“语音识别参数”区域进行如下调整,以降低显存占用:

    • CUDA数据类型:将 float32 改为 float16int8
    • 识别准确度beam_size:将 5 改为 1
    • 识别准确度best_of:将 5 改为 1
    • 启用上下文感知:将 true 改为 false
  3. 开启预分割音频:在“语音识别参数”区域,选中 Whisper预分割音频?。这会将长音频提前切分,有助于降低处理长视频时的峰值显存占用。
  4. 检查二次识别:如果你在主界面勾选了“二次识别”,请尝试取消勾选。同时翻译多个视频时,二次识别可能与首次转录同时运行,导致显存需求翻倍。
  5. 调整其他参数:在“语音识别参数”区域,确保 最长语音持续秒数 不要过大(例如设为 6-10 秒),并选中 合并过短字幕到邻近
  6. 检查CUDA环境:确认你的 NVIDIA 显卡驱动已更新,并正确安装了 CUDA 12.8 和 cuDNN 9.x。在 CMD 中输入 nvcc -Vnvidia-smi 进行验证。

如果以上调整后问题依旧,可能是你的硬件(显存或内存)确实无法满足 large-v3-turbo 模型的最低要求,请务必更换为更小的模型。

请查阅相关文档:

Post Your Reply
Open source and free maintenance is not easy. If this project is helpful to you, please consider making a small donation to help the project continue to maintain and update.

Related resource