The GPU acceleration status can be verified by following these steps:
- Ensure that you have installed the dependencies the NVIDIA GPU way (onnxruntime-gpu and PyTorch CUDA version)
- Execute the verify command in the activated virtual environment:
python -c "import torch; print(f'CUDA available: {torch.cuda.is_available()}')" - If the output
CUDA available: TrueIndicates successful configuration - Runtime can be performed using the
nvidia-smiCommand to observe GPU utilization
Troubleshooting: If it returns False, please check the driver version of your graphics card, the installation of CUDA toolkit, and whether PyTorch has installed the wheel package corresponding to the CUDA version.
This answer comes from the articleKitten-TTS-Server: a self-deployable lightweight text-to-speech serviceThe
May not be reproduced without permission:AI productivity tools " How do I verify that GPU acceleration was successfully enabled?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil