No function after installation

[marceldenzer]

I try to use it on a macbook pro m1 pro. After the installation, none of the tools work - i get some errors in terminal:

Failed to load Demucs tab. Please check your configuration.
Error: dlopen(/Users/marceldenzer/Documents/tts-generation-webui-main/installer_files/env/lib/python3.10/site-packages/torchaudio/lib/libtorchaudio.so, 0x0006): Symbol not found: __ZN2at8internal15invoke_parallelExxxRKNSt3__18functionIFvxxEEE
Referenced from: /Users/marceldenzer/Documents/tts-generation-webui-main/installer_files/env/lib/python3.10/site-packages/torchaudio/lib/libtorchaudio.so
Expected in: <421678CD-1041-32CD-92EF-29D22242240C> /Users/marceldenzer/Documents/tts-generation-webui-main/installer_files/env/lib/libtorch_cpu.dylib

I already uninstalled all dependencies so the install script can reinstall them, without any conclusion.

[rsxdalv]

Thanks for checking in. That is a torch installation issue (related to #196 )
Generally it all goes away if the torchaudio and torch mesh together. Could you please provide me with the installer_scripts/output.log file?
Generally the fix has been to reinstall torch/torchaudio:

1. Open installer_scripts/conda_env_bash.sh
2. In this terminal do: conda install -y -k torch=2.3.1 torchvision torchaudio cpuonly -c pytorch
3. Save the output for reference here.

Also, for additional diagnostics, could you please run installer_scripts/diagnostics.sh and provide me with the output.

I am simultaneously trying to fix it for you and determine what caused torch and torchaudio to become separated.

[rsxdalv]

Finally, as for:

Failed to load Demucs tab. Please check your configuration.

Even after fixing this error, some of the tools, such as AudioCraft Plus Extension might not work with MacOS as of yet, because the original developers have focused on NVidia/CUDA tech that is available on Linux/Windows.

[marceldenzer]

I get the following error wenn i try to open the conda_env_bash.sh: conda_env_bash.sh: line 3: ./installer_files/mamba/micromamba: No such file or directory

output.log

diagnostic.txt

[StaticJunkk]

Loading Demucs tab...
Failed to load Demucs tab. Please check your configuration.
Error: dlopen(/tts-generation-webui/installer_files/env/lib/python3.10/site-packages/torchaudio/lib/libtorchaudio.so, 0x0006): Symbol not found: __ZN2at8internal15invoke_parallelExxxRKNSt3__18functionIFvxxEEE
Referenced from: /tts-generation-webui/installer_files/env/lib/python3.10/site-packages/torchaudio/lib/libtorchaudio.so
Expected in: <421678CD-1041-32CD-92EF-29D22242240C> /arima/tts-generation-webui/installer_files/env/lib/libtorch_cpu.dylib

Same error for - SeamlessM4Tv2Model, Bark Voice Clone, Bark TTS tab

Tried reinstalling torch/torchaudio - still same error

[rsxdalv]

Loading Demucs tab... Failed to load Demucs tab. Please check your configuration. Error: dlopen(/tts-generation-webui/installer_files/env/lib/python3.10/site-packages/torchaudio/lib/libtorchaudio.so, 0x0006): Symbol not found: __ZN2at8internal15invoke_parallelExxxRKNSt3__18functionIFvxxEEE Referenced from: /tts-generation-webui/installer_files/env/lib/python3.10/site-packages/torchaudio/lib/libtorchaudio.so Expected in: <421678CD-1041-32CD-92EF-29D22242240C> /arima/tts-generation-webui/installer_files/env/lib/libtorch_cpu.dylib

Same error for - SeamlessM4Tv2Model, Bark Voice Clone, Bark TTS tab

Tried reinstalling torch/torchaudio - still same error

Thanks for reporting!

Could you please do:

installer_scripts/conda_env_bash.sh

pip show torch torchaudio

And installer_scripts/diagnostics.sh

[StaticJunkk]

Output for pip show

bash-3.2$ pip show torch torchaudio
Name: torch
Version: 2.3.1
Summary: Tensors and Dynamic neural networks in Python with strong GPU acceleration
Home-page: https://pytorch.org/
Author: PyTorch Team
Author-email: [email protected]
License: BSD-3
Location: /Users/yuvi.dhelawat/Desktop/arima/tts-generation-webui/installer_files/env/lib/python3.10/site-packages
Requires: filelock, fsspec, jinja2, networkx, sympy, typing-extensions
Required-by: accelerate, aeiou, audiocraft, audiolm-pytorch, auraloss, clean-fid, clip-anytorch, dctorch, demucs, descript-audio-codec, descript-audiotools, dora_search, ema-pytorch, encodec, fairseq, flashy, julius, k-diffusion, kornia, lion-pytorch, local-attention, openai-whisper, openunmix, pytorch-lightning, rotary-embedding-torch, rvc, stable-audio-tools, styletts2, suno-bark, torch-stoi, torchaudio, torchcrepe, torchdiffeq, torchmetrics, torchsde, torchtext, torchvision, v-diffusion-pytorch, valle_x, vector-quantize-pytorch, vocos, x-transformers, xformers
---
Name: torchaudio
Version: 2.3.1
Summary: An audio package for PyTorch
Home-page: https://github.com/pytorch/audio
Author: Soumith Chintala, David Pollack, Sean Naren, Peter Goldsborough, Moto Hira, Caroline Chen, Jeff Hwang, Zhaoheng Ni, Xiaohui Zhang
Author-email: [email protected]
License: 
Location: /Users/yuvi.dhelawat/Desktop/arima/tts-generation-webui/installer_files/env/lib/python3.10/site-packages
Requires: torch
Required-by: aeiou, audiocraft, audiolm-pytorch, demucs, descript-audio-codec, descript-audiotools, encodec, fairseq, openunmix, stable-audio-tools, styletts2, torch-stoi, valle_x, vocos

diagnostic.txt

[rsxdalv]

My only guess is that these two don't match:

conda list:
# packages in environment at /Users/yuvi.dhelawat/Desktop/arima/tts-generation-webui/installer_files/env:
#
# Name                    Version                   Build  Channel
pytorch                   2.3.1           cpu_generic_py310hb190f2a_1    conda-forge
torchaudio                2.3.1                 py310_cpu    pytorch

And it could be that it's refusing to reinstall, since all the versions are 2.3.1
However, we could still force it to reinstall by uninstalling pytorch and then installing it again.

Did you reinstall using conda or pip?
Looking above my guess would be conda.

Here is the command with force reinstall flag included.
conda install -y -k --force-reinstall pytorch::torch=2.3.1 pytorch::torchvision pytorch::torchaudio cpuonly -c pytorch

To have a higher chance of success, you can run uninstall before but it probably will be slow. You can also 1. try reinstall, and if it fails then 2. uninstall 3. reinstall again.

To check for sanity we can do conda list at the end of it to see if these two packages have changed successfully.