Logo
Explore Help
Register Sign In
m/mGPUmanager
1
0
Fork 0
You've already forked mGPUmanager
Code Issues 13 Pull Requests Actions Packages Projects Releases Wiki Activity
Labels Milestones New Issue
13 Open 0 Closed
Label
Use alt + click/enter to exclude labels
All labels No label
Milestone
All milestones No milestones
Project
All projects No project
Author
All users
Assignee
Assigned to nobody Assigned to anybody
m
Sort
Newest Oldest Recently updated Least recently updated Most commented Least commented Nearest due date Farthest due date
Label
13 Open 0 Closed
Label
Clear labels
Milestone
No milestone
Projects
Clear projects
Assignee
Clear assignees
No assignee
m
The broker can reclaim VRAM it did not allocate, but nothing stops a direct client taking it — a bare ollama run still OOM-kills whisper
#13 opened 2026-07-29 10:34:34 +00:00 by mAi
comfyui's VRAM need is not a constant: it varies cold-vs-warm (8410 vs 6170 MiB) and with the requested resolution
#12 opened 2026-07-29 10:06:38 +00:00 by mAi
1
comfyui's vram_budget_mib is 11000, measured usage is 4346 — the broker refuses image jobs that would fit comfortably
#11 opened 2026-07-28 23:46:53 +00:00 by mAi
3
STT glossary drift: consumers.yaml stt_prompt is hand-maintained and diverges from mVoice lexicon.yaml (missing mBrian → 'mBryan')
#10 opened 2026-07-25 09:30:31 +00:00 by mAi
2
STT forces German on every utterance: -l de hardcoded + language never forwarded, so English voice is translated to German
#9 opened 2026-07-25 09:30:31 +00:00 by mAi
3
unload(): whisper-server systemd branch marks consumer unloaded without stopping the unit (latent VRAM-not-freed bug)
#8 opened 2026-07-14 16:14:54 +00:00 by mAi
1
STT big-model via GPU manager: mVoice eager-loads at startup + PWA bypasses the broker → big model OOM-crashes on a full GPU
#7 opened 2026-07-14 15:29:30 +00:00 by mAi
3
Broker must be the SOLE VRAM authority — close bypass gaps so out-of-band ollama holds get evicted (imagen starved by 24h keep-alive)
#6 opened 2026-07-10 10:51:06 +00:00 by mAi
1
Scheduler must evict ollama for higher-priority image leases (vram_managed guard defeats the manager's purpose)
#5 opened 2026-07-06 10:35:26 +00:00 by mAi
5
Broker reports per-consumer gpu_resident_mib=0 for externally-started consumers → eviction finds "no candidates" and never reclaims their VRAM
#4 opened 2026-06-07 13:12:26 +00:00 by mAi
1
mGPUmanager: ollama not evictable + no game-awareness + OLLAMA_KEEP_ALIVE=24h starve the game/FLUX
#3 opened 2026-06-07 09:14:08 +00:00 by mAi
Add a generic GPU-lease primitive (acquire/renew/release) for long-running async consumers (ComfyUI/FLUX)
#2 opened 2026-06-07 08:49:29 +00:00 by mAi
2
Phase 1: GPU-Manager — Schritte 0–5 (ComfyUI systemd → mVoice load/unload → Broker MVP → wa.sh migrate → Queue → LRU-Eviction)
#1 opened 2026-05-11 11:18:04 +00:00 by mAi
4
Powered by Gitea Version: 1.24.4 Page: 39ms Template: 7ms
English
Bahasa Indonesia Deutsch English Español Français Gaeilge Italiano Latviešu Magyar nyelv Nederlands Polski Português de Portugal Português do Brasil Suomi Svenska Türkçe Čeština Ελληνικά Български Русский Українська فارسی മലയാളം 日本語 简体中文 繁體中文(台灣) 繁體中文(香港) 한국어
Licenses API