C:\Users\USER\Desktop\RVC1006Nvidia>runtime\python.exe infer-web.py --pycmd runtime\python.exe --port 7897
C:\Users\USER\Desktop\RVC1006Nvidia\runtime\lib\site-packages\torch\cuda_init_.py:173: UserWarning:
NVIDIA GeForce RTX 5090 with CUDA capability sm_120 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_61 sm_70 sm_75 sm_80 sm_86 sm_90 compute_37.
If you want to use the NVIDIA GeForce RTX 5090 GPU with PyTorch, please check the instructions at Start Locally | PyTorch
warnings.warn(incompatible_device_warn.format(device_name, capability, " ".join(arch_list), device_name))
2025-04-10 20:05:22 | INFO | configs.config | Found GPU NVIDIA GeForce RTX 5090
is_half:True, device:cuda:0
2025-04-10 20:05:24 | INFO | main | Use Language: en_US
Running on local URL: http://0.0.0.0:7897
2025-04-10 20:05:33 | INFO | infer.modules.vc.modules | Get sid: kikiV1.pth
2025-04-10 20:05:33 | INFO | infer.modules.vc.modules | Loading: assets/weights/kikiV1.pth
2025-04-10 20:05:34 | INFO | infer.modules.vc.modules | Select index: logs\kikiV1.index
2025-04-10 20:05:44 | INFO | infer.modules.vc.pipeline | Loading rmvpe model,assets/rmvpe/rmvpe.pt
C:\Users\USER\Desktop\RVC1006Nvidia>pause
Press any key to continue . . .
The problem could be with the software or the 5090 card.
Because when I change the card to a 4090, it works.
Is there a solution? I’ve tried everything.
To use PyTorch for Windows on NVIDIA 5080, 5090 Blackwell RTX GPUs use the latest nightly builds, or the command below.
pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128
There are also several discussions on Github about this, for example:
“CUDA applications built using CUDA Toolkit versions 2.1 through 12.8 are compatible with Blackwell GPUs as long as they are built to include PTX versions of their kernels. This can be tested by forcing the PTX to JIT-compile at application load time with following the steps:”
Which kinda made me think: it works even with cuda 2.1? Not likely but maybe cuda 11.8 still works with it?
On pytorch get started site, I can choose cuda 11.8. I tried this but this didn’t work, as cuda 12.8 is the first cuda version that made sm_120 or blackwell gpus supported, so earlier versions just can’t support it somehow.
Weirdly I couldn’t find a nice nvidia table for this, the best one I saw was wikipedia: