Hi, I have Home Assistant installed in docker on a virtual machine. To leverage a GPU, I have installed faster-whisper on a different machine. I have it set up to use cuda with the following:
version: '3'
services:
wyoming-whisper:
image: rhasspy/wyoming-whisper:latest
ports:
- "10300:10300"
volumes:
- /home/user/whisper-data:/data
command: --model medium-int8 --language en --beam-size 5 --device cuda
restart: unless-stopped
runtime: nvidia
deploy:
resources:
reservations:
devices:
- driver: nvidia
count: 1
capabilities: [gpu]
The container starts no problem, I see it occupy memory when I use “nvidia-smi”, and the docker log shows:
INFO:main:Ready
I have Home assistant set up with a pipeline using faster-whisper and piper. Piper easily speaks text.
However, I open my phone app, or use the computer. Open the assistant, click the microphone icon so that it starts blinking, but nothing shows up on the faster-whisper logs. I do not see any errors in Home Assistant. The microphone just continues to flash.
No clue even where to start troubleshooting since there is no indication of a problem. I have tried removing the CUDA argument and just running it on the PC and I get the same outcome.
Any help would be greatly appreciated in figuring out where thie issue lies.
Thanks!