#8668 [WinError 126] The specified module could not be found. Error loading

197.59**0 Posted at: 2 hours ago

[WinError 126] The specified module could not be found. Error loading "E:\برنامج دبلجة\win-pyvideotrans-v4.17-1028\_internal\torch\lib\cublas64_12.dll" or one of its dependencies.,Traceback (most recent call last):
File "videotrans\util\checkgpu.py", line 15, in run
File "videotrans\util\gpus.py", line 24, in getset_gpu
File "E:\برنامج دبلجة\win-pyvideotrans-v4.17-1028\_internal\torch\__init__.py", line 270, in

_load_dll_libraries()

File "E:\برنامج دبلجة\win-pyvideotrans-v4.17-1028\_internal\torch\__init__.py", line 266, in _load_dll_libraries

raise err

OSError: [WinError 126] The specified module could not be found. Error loading "E:\برنامج دبلجة\win-pyvideotrans-v4.17-1028\_internal\torch\lib\cublas64_12.dll" or one of its dependencies.

=
system:Windows-10-10.0.19042-SP0
version:v4.17
frozen:True
language:en_US
root_dir:E:/برنامج دبلجة/win-pyvideotrans-v4.17-1028

1 Replies

#1 DeepSeek reply at 2 hours ago

The core cause: the software is in a path containing non-English characters and spaces (E:\برنامج دبلجة\...), which prevents torch from loading cublas64_12.dll; in addition, the local CUDA environment (driver / CUDA 12.8 / cuDNN 9) is incomplete or not on PATH.

Solution:

Move and rename the software first (the most critical step)
- Extract/move the whole folder to a short path with only English letters and digits, for example D:\pyVideoTrans.
- Never keep it in C:\Program Files, C:\Windows, or any path containing Arabic/Chinese characters, spaces, or emoji.
- Never double-click sp.exe inside the archive; extract it fully first.
Configure the CUDA environment (only if you have an NVIDIA GPU and want CUDA acceleration)
- Update the NVIDIA graphics driver to the latest version.
- Install CUDA 12.8 and cuDNN 9.11; append the cuDNN bin and lib directories to the system Path environment variable, then restart the computer.
- In the cublas64_12.dll error above, WinError 126 usually means the driver is too old or cuDNN is missing/not on PATH.
If you have no NVIDIA GPU, or do not need CUDA acceleration
- Do not check CUDA Acceleration on the main interface; the software will run on CPU. This error comes from the startup GPU detection thread, and after moving the software to a pure-English path it should no longer fail to load.
After the move, verify
- If the version number in the top-left corner still shows v4.17 and the error persists, first confirm the new path contains no non-ASCII characters, then retry.

Please refer to the related documentation: