search

haoheliu / versatile_audio_super_resolution

Versatile audio super resolution (any -> 48kHz) with AudioSR.
MIT License
1.11k stars 107 forks source link

Divide by zero and killed #18

Open Mrw33554432 opened 1 year ago

Mrw33554432 commented 1 year ago

Below is the full log. The model is installed via pip (with pytorch/cuda already configured). No error reported in pip install. System: Ubuntu 22.04 LTS

$ audiosr -i f.wav
Loading AudioSR: basic
Loading model on cuda:0
xxx/lib/python3.11/site-packages/torch/functional.py:504: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at ../aten/src/ATen/native/TensorShape.cpp:3483.)
  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]
xxx/lib/python3.11/site-packages/torchaudio/transforms/_transforms.py:611: UserWarning: Argument 'onesided' has been deprecated and has no influence on the behavior of this module.
  warnings.warn(
DiffusionWrapper has 258.20 M params.
xxx/lib/python3.11/site-packages/audiosr/latent_diffusion/models/ddpm.py:237: RuntimeWarning: divide by zero encountered in divide
  "sqrt_recip_alphas_cumprod", to_torch(np.sqrt(1.0 / alphas_cumprod))
xxx/lib/python3.11/site-packages/audiosr/latent_diffusion/models/ddpm.py:240: RuntimeWarning: divide by zero encountered in divide
  "sqrt_recipm1_alphas_cumprod", to_torch(np.sqrt(1.0 / alphas_cumprod - 1))
Killed
susanin-ivan commented 1 year ago

How much VRAM do you have?

While starting, I've seen it takes up to 16 GB VRAM at some moments, then drops.

This is regular VRAM usage on my machine when audiosr is processing files from the list:

| 0 N/A N/A 31681 C /usr/bin/python3 12792MiB |

Hope this helps

shikida commented 3 months ago

same error here. killed without any further info