RVC-Project / Retrieval-based-Voice-Conversion-WebUI

Easily train a good VC model with voice data <= 10 mins!
MIT License
24.51k stars 3.61k forks source link

Crash on inference with longer file #1113

Open Acrivec opened 1 year ago

Acrivec commented 1 year ago

obraz Short files were converting correctly. I've tried to use a longer song and this keeps happening.

G:\Downloads\RVC0813>runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897 --dml
No supported Nvidia GPU found
use DirectML instead
Use Language: en_US
Running on local URL:  http://0.0.0.0:7897
loading weights/ilutest6.pth
gin_channels: 256 self.spk_embed_dim: 109
<All keys matched successfully>
loading rmvpe model
2023-08-28 18:40:20.7873460 [W:onnxruntime:, session_state.cc:1169 onnxruntime::VerifyEachNodeIsAssignedToAnEp] Some nodes were not assigned to the preferred execution providers which may or may not have an negative impact on performance. e.g. ORT explicitly assigns shape related ops to CPU to improve perf.
2023-08-28 18:40:20.7920688 [W:onnxruntime:, session_state.cc:1171 onnxruntime::VerifyEachNodeIsAssignedToAnEp] Rerunning with verbose output on a non-minimal build will show node assignments.

G:\Downloads\RVC0813>pause
Press any key to continue . . .
Acrivec commented 1 year ago

Happens only with rmvpe selected.

Acrivec commented 1 year ago

https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/1026

Qesse75 commented 1 year ago

obraz Short files were converting correctly. I've tried to use a longer song and this keeps happening.

G:\Downloads\RVC0813>runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897 --dml
No supported Nvidia GPU found
use DirectML instead
Use Language: en_US
Running on local URL:  http://0.0.0.0:7897
loading weights/ilutest6.pth
gin_channels: 256 self.spk_embed_dim: 109
<All keys matched successfully>
loading rmvpe model
2023-08-28 18:40:20.7873460 [W:onnxruntime:, session_state.cc:1169 onnxruntime::VerifyEachNodeIsAssignedToAnEp] Some nodes were not assigned to the preferred execution providers which may or may not have an negative impact on performance. e.g. ORT explicitly assigns shape related ops to CPU to improve perf.
2023-08-28 18:40:20.7920688 [W:onnxruntime:, session_state.cc:1171 onnxruntime::VerifyEachNodeIsAssignedToAnEp] Rerunning with verbose output on a non-minimal build will show node assignments.

G:\Downloads\RVC0813>pause
Press any key to continue . . .
Chen-Ed commented 10 months ago

我也是遇到同样的情况, 同样的音频 同样的模型 在pm可以正常运行 但在rmvpe会崩溃

KittyKite commented 7 months ago

Getting the same issue here, pm works fine, rmvpe crashes with this issue

A233S commented 4 months ago

same problem