Hynes/Void-Homelab
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(workers): free Ollama VRAM before loading Whisper on the GPU

Browse Source 
Whisper (CT 311) and Ollama (CT 102) share one A2000. Before loading
Whisper on CUDA, ask Ollama to unload its models (GET /api/ps then POST
/api/generate keep_alive:0) and wait for the card to clear, so the GPU
load has headroom. Best-effort and stdlib-only; Ollama reloads
cooperatively, and the existing CUDA->CPU fallback covers any failure.
Toggle via OLLAMA_FREE_BEFORE_STT; endpoint via OLLAMA_URL.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
root
2026-06-05 21:12:05 +10:00
parent c2569cad76
commit a9191cee00
5 changed files with 145 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
workers/void_workers/config.py
View File

@@ -15,6 +15,12 @@ WHISPER_MODEL = env("WHISPER_MODEL", "small.en")
WHISPER_CACHE = env("WHISPER_CACHE", "/var/lib/void/whisper-models")
ALLOW_PRIVATE = env("VOID_INGEST_ALLOW_PRIVATE", "false") == "true"
# GPU sharing: Whisper and Ollama (CT 102) share one A2000. Before loading
# Whisper on the GPU, ask Ollama to unload its models to make room (it reloads
# cooperatively on its next request). Best-effort; CPU fallback covers failure.
OLLAMA_URL = env("OLLAMA_URL", "http://192.168.1.185:11434")
OLLAMA_FREE_BEFORE_STT = env("OLLAMA_FREE_BEFORE_STT", "true") == "true"
CONCURRENCY = {
"extract.pdf": env_int("VOID_CONCURRENCY_EXTRACT_PDF", 2),
"extract.image": env_int("VOID_CONCURRENCY_EXTRACT_IMAGE", 2),