Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Bug report: Application exits when trying to clone voice in Bark Voice Clone #416

Open
Aamir3d opened this issue Nov 11, 2024 · 7 comments

Comments

@Aamir3d
Copy link

Aamir3d commented Nov 11, 2024

Steps

  1. Start Webui
  2. Switch to BARK voice clone tab
  3. Add an MP3 file (via upload)
  4. Click Generate
    image

Expected result - new voice file created from MP3
Actual result - Application exits with no error messages
image

@rsxdalv
Copy link
Owner

rsxdalv commented Nov 11, 2024

I will test it out. What about non mp3, does that work?

@rsxdalv
Copy link
Owner

rsxdalv commented Nov 11, 2024

I was able to do it, so there's a more complicated issue. MP3 does not seem to be the issue either.
I can suggest two methods: trying it again without GPU, or trying it after the computer has been restarted.

Generally if it crashes without any info that's a high-level crash, like PC running out of memory.

@Aamir3d
Copy link
Author

Aamir3d commented Nov 11, 2024

I was able to do it, so there's a more complicated issue. MP3 does not seem to be the issue either. I can suggest two methods: trying it again without GPU, or trying it after the computer has been restarted.

Generally if it crashes without any info that's a high-level crash, like PC running out of memory.

Thanks - I'm not out of memory (12 GB GPU, but it shows 50% free when the model loads). However, I'll try a restart later and see how it goes.

I just tested F5 TTS through the WebUI and that works flawlessly.
I also tested the Audio/Music Generation section with Medium and Audiogen models, they work without issues too.

@Aamir3d
Copy link
Author

Aamir3d commented Nov 11, 2024

So I turned off the GPU, but still no good. The application crashed.
image

However, when going through the startup log, I noticed this (before the Gradio WebUI started). Not sure if that's got anything to do with the crash.
image

@rsxdalv
Copy link
Owner

rsxdalv commented Nov 12, 2024

Thanks for the debug info. It really seems like a black box error. It's possible that it happens with fairseq, but the root cause might be another component, such as xformers.

As for the database error, it shouldn't cause the problem.

I'm wondering if RVC works or not. If not, it could really be a fairseq issue. Then we can try reinstalling fairseq or building from source.

@Aamir3d
Copy link
Author

Aamir3d commented Nov 17, 2024

So after some tests today

[2024-11-17 04:28:05,775] [INFO] [real_accelerator.py:219:get_accelerator] Setting ds_accelerator to cuda (auto detect)
test.c
LINK : fatal error LNK1181: cannot open input file 'aio.lib'
test.c
LINK : fatal error LNK1181: cannot open input file 'cufile.lib'
W1117 04:28:30.292000 20468 torch\distributed\elastic\multiprocessing\redirects.py:27] NOTE: Redirects are currently not supported in Windows or MacOs.
USER\\anaconda3\lib\site-packages\fairseq\tasks\multires_hubert_pretraining.py:154: SyntaxWarning: "is not" with a literal. Did you mean "!="?
  dictionaries = [ (Dictionary.load(f"{label_dir}/dict.{label}.txt") if label is not "" else None ) for label in self.cfg.labels]
""
""
App exitted or crashed.
Starting command prompt for user to run commands in case of failure...
""
AI\tts-generation-webui-main>

Not sure if that helps.
To rule out errors with @gitmylo's project, I have an earlier install and ran the Audio WebUI successfully (I was able to create an NPZ voice using Suno).

@rsxdalv
Copy link
Owner

rsxdalv commented Nov 23, 2024

So just to go over some of these:

  • LINK : fatal error LNK1181: cannot open input file 'aio.lib'
    This is just because of how deepspeed works on Windows, should not be the core issue.
  • W1117 04:28:30.292000 20468 torch\distributed\elastic\multiprocessing\redirects.py:27] NOTE: Redirects are currently not supported in Windows or MacOs.
    Almost the same as above
  • USER\anaconda3\lib\site-packages\fairseq\tasks\multires_hubert_pretraining.py:154: SyntaxWarning: "is not" with a literal. Did you mean "!="?
    dictionaries = [ (Dictionary.load(f"{label_dir}/dict.{label}.txt") if label is not "" else None ) for label in self.cfg.labels]
    This is because fairseq is old and has this warning-raising code.

But I can see what is probably the real problem - USER\anaconda3 - the anaconda install on PC is interfering with the tts-webui's anaconda. This could be confirmed with a couple of checks:

diagnostics.bat
conda_env_cmd.bat then doing
where.exe conda
and
echo %PATH%
This should indicate if the system conda is not letting the tts-webui conda to work (which appears to be the case).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants