"Invalid Value Encountered in Divide" when running preprocessing code.

hayeong0 / DDDM-VC

Official Pytorch Implementation for "DDDM-VC: Decoupled Denoising Diffusion Models with Disentangled Representation and Prior Mixup for Verified Robust Voice Conversion" (AAAI 2024)

https://hayeong0.github.io/DDDM-VC-demo/

160 stars 18 forks source link

"Invalid Value Encountered in Divide" when running preprocessing code. #9

Closed cryptowooser closed 2 months ago

cryptowooser commented 3 months ago

Hello,

I'm encountering the following runtime warning when running the preprocessing code against my dataset. This appears to show up for every wav I throw at it, and these wavs are typical 16khz so I'm not sure what the issue might be.

/home/gradientascent/miniconda3/envs/DDDM-VC-2/lib/python3.9/site-packages/amfm_decompy/pYAAPT.py:970: RuntimeWarning: invalid value encountered in divide phi[lag_min:lag_max] = formula_nume/np.sqrt(formula_denom)

Is this a known issue?

hayeong0 commented 3 months ago

Hello,

It seems that this warning occurs when using the YAAPT algorithm to extract F0. Could you let me know what type of data you are using? This issue is likely to arise if 1) there are significant periods of silence in the audio, or 2) the audio signal has a low volume. I have encountered this warning with singing data or real-world samples. Normalizing the audio to a consistent scale might help mitigate the problem :)

cryptowooser commented 3 months ago

Thank you for your quick reply! It appears to have been caused by brief silences at the start and end of the audio. I'll try removing them with pydub and see if that helps.