torchcodec0.9.1+torch2.91

[huangxuegang1129-oss]

🐛 Describe the bug

RuntimeError: Could not load libtorchcodec. Likely causes:
1. FFmpeg is not properly installed in your environment. We support
versions 4, 5, 6, 7, and 8. On Windows, ensure you've installed
the "full-shared" version which ships DLLs.
2. The PyTorch version (2.9.1+cu130) is not compatible with
this version of TorchCodec. Refer to the version compatibility
table:
https://github.com/pytorch/torchcodec?tab=readme-ov-file#installing-torchcodec.
3. Another runtime dependency; see exceptions below.
The following exceptions were raised as we tried to load libtorchcodec:

[start of libtorchcodec loading traceback]
FFmpeg version 8: Could not load this library: G:\Indextts2.venv\Lib\site-packages\torchcodec\libtorchcodec_core8.dll
FFmpeg version 7: Could not load this library: G:\Indextts2.venv\Lib\site-packages\torchcodec\libtorchcodec_core7.dll
FFmpeg version 6: Could not load this library: G:\Indextts2.venv\Lib\site-packages\torchcodec\libtorchcodec_core6.dll
FFmpeg version 5: Could not load this library: G:\Indextts2.venv\Lib\site-packages\torchcodec\libtorchcodec_core5.dll
FFmpeg version 4: Could not load this library: G:\Indextts2.venv\Lib\site-packages\torchcodec\libtorchcodec_core4.dll
[end of libtorchcodec loading traceback].

Versions

torchcodec0.9.1+torch2.91

[WangHelin1997]

I fixed it by: conda install ffmpeg -y && conda update ffmpeg -y

[cometbroom]

Those dlls are loaded with ctypes.CDLL, issue on windows is that loading the dll could also fail because of a dependent missing and windows API doesn't report it. https://docs.python.org/3.13/library/ctypes.html#ctypes.CDLL

You can use dumpbin /DEPENDENTS {dll_name} in a visual studio initialized CMD (vcvarsall.bat) to get a list of dependents for the DLL and see which ones you're missing.

in ...\Lib\site-packages\torchcodec directory
dumpbin /DEPENDENTS libtorchcodec_core7.dll :

\Lib\site-packages\torchcodec>dumpbin /DEPENDENTS libtorchcodec_core7.dll
Microsoft (R) COFF/PE Dumper Version 14.50.35721.0
Copyright (C) Microsoft Corporation.  All rights reserved.


Dump of file libtorchcodec_core7.dll

File Type: DLL

  Image has the following dependencies:

    c10.dll
    c10_cuda.dll
    nppicc64_13.dll
    avfilter-10.dll
    avformat-61.dll
    avcodec-61.dll
    swresample-5.dll
    swscale-8.dll
    avutil-59.dll
    torch_cuda.dll
    torch_cpu.dll
    KERNEL32.dll
    MSVCP140.dll
    VCRUNTIME140.dll
    VCRUNTIME140_1.dll
    api-ms-win-crt-runtime-l1-1-0.dll
    api-ms-win-crt-heap-l1-1-0.dll
    api-ms-win-crt-math-l1-1-0.dll
    api-ms-win-crt-environment-l1-1-0.dll
    api-ms-win-crt-stdio-l1-1-0.dll
    api-ms-win-crt-convert-l1-1-0.dll

You can ignore c10, c10_cuda, torch_cuda, torch_cpu. Those are loaded differently I believe.

For my case nppicc64_13.dll was missing. That dll is part of nppicc.lib which is provided by CUDA Toolkit. It can be found in your CUDA directory under lib\x64. If you installed CUDA with default settings, this short-path should get you there C:\PROGRA~1\NVIDIA~2\CUDA\v13.0\lib\x64

While I had this already added to PATH, apparently it wasn't enough so I had to manually add the dll directory to DLL searchpath in python

import os

os.add_dll_directory(r"C:\PROGRA~1\NVIDIA~2\CUDA\v13.0\lib\x64")

After doing that. torchcodec was successfully imported and tested it on decoding audio

[BInyLU]

这些 DLL 文件是通过 ctypes.CDLL 加载的，问题在于 Windows 系统上，由于缺少依赖项，DLL 加载也可能失败，而 Windows API 不会报告此问题。https ://docs.python.org/3.13/library/ctypes.html#ctypes.CDLL

您可以使用dumpbin /DEPENDENTS {dll_name}Visual Studio 初始化的 CMD (vcvarsall.bat) 获取 DLL 的依赖项列表，并查看缺少哪些依赖项。

在 ...\Lib\site-packages\torchcodec 目录中 dumpbin /DEPENDENTS libtorchcodec_core7.dll：

\Lib\site-packages\torchcodec>dumpbin /DEPENDENTS libtorchcodec_core7.dll
Microsoft (R) COFF/PE Dumper Version 14.50.35721.0
Copyright (C) Microsoft Corporation.  All rights reserved.


Dump of file libtorchcodec_core7.dll

File Type: DLL

  Image has the following dependencies:

    c10.dll
    c10_cuda.dll
    nppicc64_13.dll
    avfilter-10.dll
    avformat-61.dll
    avcodec-61.dll
    swresample-5.dll
    swscale-8.dll
    avutil-59.dll
    torch_cuda.dll
    torch_cpu.dll
    KERNEL32.dll
    MSVCP140.dll
    VCRUNTIME140.dll
    VCRUNTIME140_1.dll
    api-ms-win-crt-runtime-l1-1-0.dll
    api-ms-win-crt-heap-l1-1-0.dll
    api-ms-win-crt-math-l1-1-0.dll
    api-ms-win-crt-environment-l1-1-0.dll
    api-ms-win-crt-stdio-l1-1-0.dll
    api-ms-win-crt-convert-l1-1-0.dll

您可以忽略 c10、c10_cuda、torch_cuda 和 torch_cpu。它们的加载方式应该有所不同。

我的情况nppicc64_13.dll是缺少这个 DLL 文件。该 DLL 文件是 CUDA 工具包的一部分nppicc.lib。它可以在你的 CUDA 目录下找到lib\x64。如果你使用默认设置安装了 CUDA，那么这个简短的路径应该可以带你到那里。C:\PROGRA~1\NVIDIA~2\CUDA\v13.0\lib\x64

虽然我已经将它添加到 PATH 环境变量中，但显然这还不够，所以我必须手动将 dll 目录添加到 Python 的 DLL 搜索路径中。

import os

os.add_dll_directory(r"C:\PROGRA1\NVIDIA2\CUDA\v13.0\lib\x64")
完成上述操作后，torchcodec 已成功导入，并测试了其音频解码功能。

not work