[ISSUE for v2]: v.2.0.76-beta There is a load on the CPU when converting audio with a 'pth' file.

[Rakkyotan]

Voice Changer Version

vcclient_win_cuda_2.0.76-beta.zip

Operational System

Windows 11

GPU

RTX4070Ti

CUDA Version

12.6

Read carefully and check the options

• If you use win_cuda_torch_cuda edition, setup cuda? see here
• If you use win_cuda edition, setup cuda and cudnn? see here
• If you use mac edition, client is not launched automatically. Use chrome to open application.?
• I've tried to change the Chunk Size
• I've tried to set the Index to zero
• I've read the tutorial
• I've tried to extract to another folder (or re-extract) the .zip file

Does pre-installed model work?

YES

Model Type

RVC

Issue Description

In version 2.0.72, the CPU load rate hardly increased during voice conversion, but in 2.0.76, the CPU was loaded and the CPU load rate was over 80%. I had the same symptoms when I used VU last time, so I may not have gone to the GPU. Thank you for your understanding.

バージョン2.0.72の時は音声変換の際CPU負荷率がほぼ上がりませんでしたが、2.0.76ではCPUに負荷がかかっておりCPU負荷率が80％以上になっていました。前回のVUの時も同じ症状がありましたので、GPUの方へいっていないかもしれません。確認の程、よろしくお願いいたします。

Application Screenshot

CPU load rate while talking　Ver.2.0.73

CPU load rate while talking　Ver.2.0.76

Logs on console

Ver.2.0.73
2025-02-23 21:46:27,242 - vcclient - gpu_device_manager - INFO - Reloading GPU info - vcclient_dev\voice_changer\gpu_device_manager\gpu_device_manager.py - 120
2025-02-23 21:46:27,288 - vcclient - gpu_device_manager - ERROR - Failed to reload_gpu_info_for_win. Failed to resolve Python.Runtime.Loader.Initialize from C:\RVC2073_internal\pythonnet\runtime\Python.Runtime.dll. If you are using torch_cuda, this error can be ignored. - vcclient_dev\voice_changer\gpu_device_manager\gpu_device_manager.py - 50
2025-02-23 21:46:27,288 - vcclient - gpu_device_manager - INFO - GPU[sys]:[GPUInfo(name='cpu', device_id='-1', adapter_ram=0, device_id_int=-1, cuda_compute_version_major=-1, cuda_compute_version_minor=-1)] - vcclient_dev\voice_changer\gpu_device_manager\gpu_device_manager.py - 123
2025-02-23 21:46:27,379 - vcclient - gpu_device_manager - INFO - Sun Feb 23 21:46:27 2025 - vcclient_dev\voice_changer\gpu_device_manager\gpu_device_manager.py - 96

Ver.2.0.76
2025-02-23 21:52:05,415 - vcclient - gpu_device_manager - INFO - Reloading GPU info - vcclient_dev\voice_changer\gpu_device_manager\gpu_device_manager.py - 120
2025-02-23 21:52:05,459 - vcclient - gpu_device_manager - ERROR - Failed to reload_gpu_info_for_win. Failed to resolve Python.Runtime.Loader.Initialize from C:\RVC2076_internal\pythonnet\runtime\Python.Runtime.dll. If you are using torch_cuda, this error can be ignored. - vcclient_dev\voice_changer\gpu_device_manager\gpu_device_manager.py - 50
2025-02-23 21:52:05,459 - vcclient - gpu_device_manager - INFO - GPU[sys]:[GPUInfo(name='cpu', device_id='-1', adapter_ram=0, device_id_int=-1, cuda_compute_version_major=-1, cuda_compute_version_minor=-1)] - vcclient_dev\voice_changer\gpu_device_manager\gpu_device_manager.py - 123
2025-02-23 21:52:05,553 - vcclient - gpu_device_manager - INFO - Sun Feb 23 21:52:05 2025 - vcclient_dev\voice_changer\gpu_device_manager\gpu_device_manager.py - 96

[lenore-puwawa]

Possible reasons for increased CPU load include the following:

• Increasing the index value to greater than 0:
  　→ On Windows, setting the index value higher than 0 increases CPU load.
• Using an F0 extraction algorithm that doesn't utilize the GPU:
  　→ For example, harvest/dio uses the CPU, while rmvpe uses the GPU.
• Setting "Gpu:" (under "Extra:") to cpu[-1]:
  　→ Please select your GPU.

Including a screenshot of the VC Client in operation may help obtain a more accurate response.

[Rakkyotan]

When I checked, the index had gone up for some reason.
Thank you for letting me know!

確認しましたところindexがなぜか上がっていました
教えていただきありがとうございます