Open kazumi371 opened 3 weeks ago
my friend keep getting this and its slow
Did you pick Tensor gpu from the list? Did you also setup it in Google collab settings?
If you're talking about the t4 one then yes they said it glitches out specifically when they use that one
In that error code just keeps showing up its not working on the t4 gpu
There is a button that may make it easier for you to download the whole log and show the missing parts from the screenshot. I am not a programmer but I am sure that Mr. Okada will probably be able to take a better look.
logs.txt here is the log my friend sent me this is the error they run into
I have just tried the colab one and it works perfectly fine with the exception that, after few minutes, it stops working because I am not paying for any of the premium plans, which is expected. Other than that, I managed to use the default models to convert an audio clip in real time. It was a bit choppy at first and the output would cut every second or so but after a while it started to behave correctly, then the server stopped due the free limitations.
Well it doesn't work specifically for the t4 gpu they mentioned how it works fine with just a normal CPU but it's really slow but for the t4 GPU it shows that error message it's either not reading that GPU being used or it's running an error because there's something wrong with cuda
I used it perfectly fine with the T4 GPU.
Then I don't know what's going on with theirs
Try a different browser maybe?
I get some errors in the console too but the conversion still works. I can't test deeper because I am free user so I barely can use it for around 5 minutes.
I used it perfectly fine with the T4 GPU.
Which link are you choosing to collab proxy or Ngrok URL? And are you using the version that was updated 3 days ago or my friend is trying to figure it out
I just used the link in the main page of RVC. Then I click on "open in colab" button. https://github.com/w-okada/voice-changer/blob/v.2/w_okada's_Voice_Changer_version_2_x.ipynb After that, I just follow the instructions given, enter my Ngrok token, and run it. I open the Ngrok URL that ends in "app" and I am done.
I just used the link in the main page of RVC. Then I click on "open in colab" button. https://github.com/w-okada/voice-changer/blob/v.2/w_okada's_Voice_Changer_version_2_x.ipynb After that, I just follow the instructions given, enter my Ngrok token, and run it. I open the Ngrok URL that ends in "app" and I am done.
it still does not work
It's sometimes works for them but they keep getting an error message when they speak into the mic and that's the problem it's not outputting anything it just keeps running into that error message something about cuda not finalized and it has something to do with the t4 GPU
I don't know why it's not working for them
Cuz sometimes when you open it it's set to CPU and using the CPU it doesn't work for them it's an error within the files or something with a cuda programming or cuda version it's running
I hope this is helpful for this issue to potentially be fixed
its still not working for them
its still not working for them
Are they paying premium service?
its still not working for them
Are they paying premium service?
No but it usually works without it
For some reason there's can run for hours
Well now you know why it doesn't work then. Google stop the processes after few minutes, that's why there is no output sound. It only works for few minutes then you need to wait for hours before being able to try again.
Well now you know why it doesn't work then. Google stop the processes after few minutes, that's why there is no output sound. It only works for few minutes then you need to wait for hours before being able to try again.
That doesn't explain the error messages something is wrong with a specific file or folder
Well now you know why it doesn't work then. Google stop the processes after few minutes, that's why there is no output sound. It only works for few minutes then you need to wait for hours before being able to try again.
That doesn't explain the error messages something is wrong with a specific file or folder
That doesn't matter. As mentioned already, the software works regardless of the error. I get the error too but the software works and even you stated that it worked at first then stopped working. That's because it will only do for few minutes unless you pay. It won't work for free. And the colab version is the same for everyone since is ran remotely so it either fails for everyone or works for everyone. So, unless something is done wrongly at time of running it, it should still work. You can try running it through a different browser and you will see that after few minutes the software will give an error because the server has simply stopped processing. The UI will remain and you may be even able to click on the buttons but you won't get any output at all. I am not sure what else could be done other than paying for a paid service. @w-okada may check and see about that error, but other users are using the colab version just fine already with the difference that they are paying for it.
Voice Changer Version
vcclient
Operational System
windows 10
GPU
none
CUDA Version
none
Read carefully and check the options
Does pre-installed model work?
No
Model Type
rvc
Issue Description
the colab version is not outputting anything and its not reading inputs
Application Screenshot
Logs on console
File "vcclient_dev/voice_changer/voice_change_manager/models/rvc/attentions.py", line 52, in forward File "torch/nn/modules/module.py", line 1553, in _wrapped_call_impl return self._call_impl(*args, kwargs) File "torch/nn/modules/module.py", line 1562, in _call_impl return forward_call(*args, *kwargs) File "vcclient_dev/voice_changer/voice_change_manager/models/rvc/attentions.py", line 181, in forward File "torch/nn/modules/module.py", line 1553, in _wrapped_call_impl return self._call_impl(args, kwargs) File "torch/nn/modules/module.py", line 1562, in _call_impl return forward_call(*args, **kwargs) File "torch/nn/modules/conv.py", line 308, in forward return self._conv_forward(input, self.weight, self.bias) File "torch/nn/modules/conv.py", line 304, in _conv_forward return F.conv1d(input, weight, bias, self.stride, RuntimeError: CUDNN_BACKEND_TENSOR_DESCRIPTOR cudnnFinalize failed cudnn_status: CUDNN_STATUS_NOT_INITIALIZED During handling of the above exception, another exception occurred: Traceback (most recent call last): File "vcclient_dev/server/validation_error_logging_route.py", line 17, in custom_route_handler File "fastapi/routing.py", line 278, in app File "fastapi/routing.py", line 191, in run_endpoint_function File "vcclient_dev/server/vcserver_rest_api_voice_changaer.py", line 88, in post_convert_chunk File "vcclient_dev/voice_changer/voice_change_manager/voice_changer.py", line 390, in convert_chunk File "vcclient_dev/voice_changer/voice_change_manager/voice_changer.py", line 516, in _convert_chunk File "vcclient_dev/voice_changer/voice_change_manager/vc_pipelines/rvc_pipeline.py", line 320, in run File "vcclient_dev/voice_changer/voice_change_manager/inferencer/rvc_inverencer_v2_f0.py", line 99, in infer RuntimeError: ('Exeption in RVCInferencerv2F0', RuntimeError('CUDNN_BACKEND_TENSOR_DESCRIPTOR cudnnFinalize failed cudnn_status: CUDNN_STATUS_NOT_INITIALIZED'))