#7741 Cannot copy out of meta tensor; no data!:Traceback (most recent call last): File

2a09**f4 Posted at: 1 hour ago

Cannot copy out of meta tensor; no data!:Traceback (most recent call last):
File "videotrans\process\stt_qwen.py", line 97, in qwen3asr_fun
File "C:\Users\ADMIN\Downloads\New folder (6)\_internal\torch\nn\modules\module.py", line 1751, in _wrapped_call_impl

return self._call_impl(*args, **kwargs)

File "C:\Users\ADMIN\Downloads\New folder (6)\_internal\torch\nn\modules\module.py", line 1762, in _call_impl

return forward_call(*args, **kwargs)

File "accelerate\hooks.py", line 170, in new_forward

args, kwargs = module._hf_hook.pre_forward(module, *args, **kwargs)

File "accelerate\hooks.py", line 369, in pre_forward

return send_to_device(args, self.execution_device), send_to_device(

File "accelerate\utils\operations.py", line 179, in send_to_device

k: t if k in skip_keys else send_to_device(t, device, non_blocking=non_blocking, skip_keys=skip_keys)

File "accelerate\utils\operations.py", line 180, in

for k, t in tensor.items()

File "acce
......
se.py", line 264, in _new_process
videotrans.configure.excepts.VideoTransError: Cannot copy out of meta tensor; no data!:Traceback (most recent call last):
File "videotrans\process\stt_qwen.py", line 97, in qwen3asr_fun
File "C:\Users\ADMIN\Downloads\New folder (6)\_internal\torch\nn\modules\module.py", line 1751, in _wrapped_call_impl

return self._call_impl(*args, **kwargs)

File "C:\Users\ADMIN\Downloads\New folder (6)\_internal\torch\nn\modules\module.py", line 1762, in _call_impl

return forward_call(*args, **kwargs)

File "accelerate\hooks.py", line 170, in new_forward

args, kwargs = module._hf_hook.pre_forward(module, *args, **kwargs)

File "accelerate\hooks.py", line 369, in pre_forward

return send_to_device(args, self.execution_device), send_to_device(

File "accelerate\utils\operations.py", line 179, in send_to_device

k: t if k in skip_keys else send_to_device(t, device, non_blocking=non_blocking, skip_keys=skip_keys)

File "accelerate\utils\operations.py", line 180, in

for k, t in tensor.items()

File "accelerate\utils\operations.py", line 154, in send_to_device

except TypeError:  # .to() doesn't accept non_blocking as kwarg

NotImplementedError: Cannot copy out of meta tensor; no data!

cfg=[TaskCfgVTT]当前工作模式: 翻译视频 单视频模式
原始输入文件名: D:/75顾佳艺7期中级班2020/视频/3..LESSON03场景氛围光影层次_1.mp4,
输出结果保存到文件夹: D:/75顾佳艺7期中级班2020/视频/_video_out/3..LESSON03场景氛围光影层次_1-mp4,
临时文件夹: C:/Users/ADMIN/Downloads/New folder (6)/tmp/1488/7e7ee31d31
已选 清理已存在
已启用CUDA加速
未选 降噪
STT Service:Qwen-ASR(Built-in), 1.7B, 发音语言: Simplified Chinese, 断句方式:默认断句
TranslateSrt:Google Translate (Free),原始语言:Simplified Chinese,目标语言:Vietnamese, 已选 SendSRT
TTS Service:Edge-TTS (Free), 角色:HoaiMy(Female/VN), 配音语言:Vietnamese, 未选 二次语音识别
音量:+0%, 语速:+0%, 已选 音频加速, 已选 视频慢速
字幕: Hard Subtitles
video_duration_ms=9110780

=
system:Windows-10-10.0.26300-SP0
version:v4.14
frozen:True
language:en_US
root_dir:C:/Users/ADMIN/Downloads/New folder (6)

1 Replies

#1 DeepSeek reply at 1 hour ago

The core cause of this error is that the Qwen-ASR model file is incomplete or corrupted (a partial download). The error Cannot copy out of meta tensor; no data! occurs because the model weights were not fully loaded — the tensors exist as empty "meta" placeholders instead of real data.

There are also two secondary issues worth fixing:

  1. Software path contains spaces and brackets — your root directory is C:/Users/ADMIN/Downloads/New folder (6). Spaces, parentheses, and non-English paths frequently break model loading, ffmpeg calls, and internal subprocesses. Move the entire software folder to a simple path such as D:\pyVideoTrans.
  2. 发音语言: Simplified Chinese with Qwen-ASR 1.7B is fine, but if the model cannot load, it will fail regardless.

Solutions (in order)

  1. Move the software to a clean path

    • Extract/move the whole folder to something like D:\pyVideoTrans (no spaces, brackets, or Chinese characters).

  2. Delete the corrupted Qwen-ASR model and re-download

    • Delete the folder: 软件目录/models/models--Qwen--Qwen3-ASR-1.7B
    • Restart the software and run again; it will re-download.
    • If auto-download keeps failing, manually download all files from:
      https://huggingface.co/Qwen/Qwen3-ASR-1.7B/tree/main
      and place them in 软件目录/models/models--Qwen--Qwen3-ASR-1.7B

  3. Verify the model size

    • The 1.7B model is several GB. If the folder is only a few hundred MB, it is incomplete — delete and re-download.

  4. Switch model as a quick test

    • In the STT service dropdown, temporarily select the 0.6B version instead of 1.7B, or switch to faster-whisper(内置) with large-v3, to confirm the software pipeline itself works while the Qwen model re-downloads.

  5. If CUDA-related instability persists

    • Uncheck 启用CUDA加速 and retry once to confirm whether the model loads on CPU. (Qwen-ASR prefers GPU regardless, but this isolates GPU/driver issues.)
Note: your input video path also contains Chinese characters and dots: D:/75顾佳艺7期中级班2020/视频/3..LESSON03场景氛围光影层次_1.mp4. Rename the video to a simple English name (e.g. lesson03.mp4) to avoid additional ffmpeg errors.

请查阅相关文档：

Thinking...
Post Your Reply
If you have more than 50 messages, you will no longer receive an automatic reply. Please ask a new question!
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.

Dondate via ko-fi.com

Related resource