RVC crashing when a model is loaded

(RVC) [USER@main Retrieval-based-Voice-Conversion-WebUI]$ python infer-web.py
2024-07-26 10:51:30 | INFO | configs.config | Found GPU AMD Radeon RX 7900 XTX
2024-07-26 10:51:30 | INFO | configs.config | Half-precision floating-point: True, device: cuda:0
/home/USER/RVC/lib64/python3.8/site-packages/gradio_client/documentation.py:106: UserWarning: Could not get documentation group for <class 'gradio.mix.Parallel'>: No known documentation group for module 'gradio.mix'
warnings.warn(f"Could not get documentation group for {cls}: {exc}")
/home/USER/RVC/lib64/python3.8/site-packages/gradio_client/documentation.py:106: UserWarning: Could not get documentation group for <class 'gradio.mix.Series'>: No known documentation group for module 'gradio.mix'
warnings.warn(f"Could not get documentation group for {cls}: {exc}")
2024-07-26 10:51:31 | INFO | main | Use Language: en_US
Running on local URL: http://0.0.0.0:7865
2024-07-26 10:52:12 | INFO | infer.modules.vc.modules | Get sid: Madonna2333333.pth
2024-07-26 10:52:12 | INFO | infer.modules.vc.modules | Loading: assets/weights/Madonna2333333.pth
Speicherzugriffsfehler (Speicherabzug geschrieben)

I am running RVC on Fedora 40.
I created a virtual env for RVC and successfully installed all requirements.
I can start it according to tutorial and i can see the webinterface, browse it, it detects my GPU, all is fine.
But when i load a .pth-file from /assets/weighs/my.pth, and i tried a lot of them, i get this error and the python process crashes.
The model-files all lay on a NFS-share and have user-permissions for the same User that runs RVC.
I also have Stable Diffusion on same machine which works without a problem.

Start with infer-web.py and the model-loading path in infer/modules/vc/modules, then reproduce the crash using a .pth file from assets/weights on Fedora 40 with the AMD Radeon RX 7900 XTX setup described. Compare loading from the NFS share and a local path, and use the resulting logs or crash data to define a reproducible fix and successful model loading.