search

dnhkng / GlaDOS

This is the Personality Core for GLaDOS, the first steps towards a real-life implementation of the AI from the Portal series by Valve.
MIT License
3.04k stars 285 forks source link

CPU usage during idle listening 50% #43

Closed l33tkr3w closed 6 months ago

l33tkr3w commented 6 months ago

I see that when GlaDOS is sitting idle and just "Listening..." my CPU usage is hovering at 50% usage on a 14900k. Is this expected? Im just tryiong to determine if this is expected or if im not using a Cuda library somewhere. This is with the recent windows installer.

dnhkng commented 6 months ago

Seems like its inferencing the VAD on CPU, but that's still too much CPU...

l33tkr3w commented 6 months ago

Im still troubleshooting. Seems like everything is running on cudda except for onnxruntime. When I check my device info with ```

import onnxruntime as rt rt.get_device() 'GPU'

dnhkng commented 6 months ago

Make sure you have uninstalled onnxruntime before installing onnxruntime-gpu.

l33tkr3w commented 6 months ago

Make sure you have uninstalled onnxruntime before installing onnxruntime-gpu.

while in venv I did pip uninstall onnxruntime pip uninstall onnxruntime-gpu pip cache purge pip install onnxruntime-gpu

Still sits at 45-50% usage. It sees the device as GPU but somehow sets the provider to CPU.

Cuda compilation tools, release 12.2, V12.2.128

dnhkng commented 6 months ago

Seems this is CPU only.

https://github.com/snakers4/silero-vad/discussions/388