Hacker-Software/ARIA-AGENT
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 267 Wiki Activity

fix(xtts): Flotte zeigt die echte Karte statt "GPU void"

Browse Source 
Das NVIDIA-Container-Toolkit ersetzt NVIDIA_VISIBLE_DEVICES im Prozess durch
"void" — whisper/f5tts/voxtral meldeten das als GPU. Die Compose gibt die
Karte jetzt zusaetzlich als ARIA_GPU mit (nur Anzeige).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
duffyduck
2026-10-25 12:51:41 +01:00
co-authored by Claude Opus 5.5
parent 7b4a1f4429
commit 5a59ea3930
4 changed files with 18 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
xtts/docker-compose.yml
+3
View File
@@ -43,6 +43,7 @@ services:
environment: environment:
# GPU-Wahl: NVIDIA_VISIBLE_DEVICES (mehrere via "0,1"). Default GPU 0. # GPU-Wahl: NVIDIA_VISIBLE_DEVICES (mehrere via "0,1"). Default GPU 0.
- NVIDIA_VISIBLE_DEVICES=${F5TTS_GPU:-0} - NVIDIA_VISIBLE_DEVICES=${F5TTS_GPU:-0}
- ARIA_GPU=${F5TTS_GPU:-0} # nur Anzeige (Toolkit macht aus NVIDIA_VISIBLE_DEVICES "void")
- NVIDIA_DRIVER_CAPABILITIES=compute,utility - NVIDIA_DRIVER_CAPABILITIES=compute,utility
- NODE_NAME=${NODE_NAME:-node} # erscheint in Diagnostic + Logs - NODE_NAME=${NODE_NAME:-node} # erscheint in Diagnostic + Logs
# Bootstrap-only — alle anderen F5-TTS-Settings (Modell, cfg_strength, # Bootstrap-only — alle anderen F5-TTS-Settings (Modell, cfg_strength,
@@ -70,6 +71,7 @@ services:
runtime: nvidia runtime: nvidia
environment: environment:
- NVIDIA_VISIBLE_DEVICES=${WHISPER_GPU:-1} - NVIDIA_VISIBLE_DEVICES=${WHISPER_GPU:-1}
- ARIA_GPU=${WHISPER_GPU:-1} # nur Anzeige (Toolkit macht aus NVIDIA_VISIBLE_DEVICES "void")
- NVIDIA_DRIVER_CAPABILITIES=compute,utility - NVIDIA_DRIVER_CAPABILITIES=compute,utility
- NODE_NAME=${NODE_NAME:-node} - NODE_NAME=${NODE_NAME:-node}
- RVS_HOST=${RVS_HOST} - RVS_HOST=${RVS_HOST}
@@ -105,6 +107,7 @@ services:
- ./voice-id:/voice-id # Speaker-Fingerprint (wie whisper) - ./voice-id:/voice-id # Speaker-Fingerprint (wie whisper)
environment: environment:
- NVIDIA_VISIBLE_DEVICES=${VOXTRAL_GPU:-1} # STT-3B ~9 GB → 12-GB-Karte - NVIDIA_VISIBLE_DEVICES=${VOXTRAL_GPU:-1} # STT-3B ~9 GB → 12-GB-Karte
- ARIA_GPU=${VOXTRAL_GPU:-1} # nur Anzeige (Toolkit macht aus NVIDIA_VISIBLE_DEVICES "void")
- NVIDIA_DRIVER_CAPABILITIES=compute,utility - NVIDIA_DRIVER_CAPABILITIES=compute,utility
- NODE_NAME=${NODE_NAME:-node} - NODE_NAME=${NODE_NAME:-node}
- RVS_HOST=${RVS_HOST} - RVS_HOST=${RVS_HOST}
xtts/f5tts/bridge.py
+5 -1
View File
@@ -78,7 +78,11 @@ F5TTS_DEVICE = os.getenv("F5TTS_DEVICE", "cuda") # nur Bootstrap
# ── Compute-Fleet: Worker-Identitaet & Registrierung ────────────── # ── Compute-Fleet: Worker-Identitaet & Registrierung ──────────────
# Meldet sich bei der aria-bridge (worker_hello) + periodischer worker_ping. # Meldet sich bei der aria-bridge (worker_hello) + periodischer worker_ping.
NODE_NAME = os.getenv("NODE_NAME", "node").strip() or "node" NODE_NAME = os.getenv("NODE_NAME", "node").strip() or "node"
GPU_IDS = os.getenv("NVIDIA_VISIBLE_DEVICES", "").strip() # Welche Karte(n) dieser Dienst nutzt — fuer die Anzeige. ARIA_GPU setzt die
# Compose; NVIDIA_VISIBLE_DEVICES ersetzt das Container-Toolkit im Prozess
# durch "void", taugt also nicht zur Anzeige.
GPU_IDS = (os.getenv("ARIA_GPU", "").strip()
or os.getenv("NVIDIA_VISIBLE_DEVICES", "").strip().replace("void", ""))
WORKER_SERVICE = "f5tts" WORKER_SERVICE = "f5tts"
INSTANCE_ID = f"{WORKER_SERVICE}@{NODE_NAME}" INSTANCE_ID = f"{WORKER_SERVICE}@{NODE_NAME}"
WORKER_PING_INTERVAL_S = int(os.getenv("WORKER_PING_INTERVAL_S", "10")) WORKER_PING_INTERVAL_S = int(os.getenv("WORKER_PING_INTERVAL_S", "10"))
xtts/voxtral/bridge.py
+5 -1
View File
@@ -78,7 +78,11 @@ VOXTRAL_DEVICE = os.getenv("VOXTRAL_DEVICE", "cuda")
# Registry per periodischem worker_ping frisch. INSTANCE_ID adressiert diesen # Registry per periodischem worker_ping frisch. INSTANCE_ID adressiert diesen
# Worker bei Redundanz (targetInstance-Routing, Stage 3). # Worker bei Redundanz (targetInstance-Routing, Stage 3).
NODE_NAME = os.getenv("NODE_NAME", "node").strip() or "node" NODE_NAME = os.getenv("NODE_NAME", "node").strip() or "node"
GPU_IDS = os.getenv("NVIDIA_VISIBLE_DEVICES", "").strip() # Welche Karte(n) dieser Dienst nutzt — fuer die Anzeige. ARIA_GPU setzt die
# Compose; NVIDIA_VISIBLE_DEVICES ersetzt das Container-Toolkit im Prozess
# durch "void", taugt also nicht zur Anzeige.
GPU_IDS = (os.getenv("ARIA_GPU", "").strip()
or os.getenv("NVIDIA_VISIBLE_DEVICES", "").strip().replace("void", ""))
WORKER_SERVICE = "voxtral" WORKER_SERVICE = "voxtral"
INSTANCE_ID = f"{WORKER_SERVICE}@{NODE_NAME}" INSTANCE_ID = f"{WORKER_SERVICE}@{NODE_NAME}"
WORKER_PING_INTERVAL_S = int(os.getenv("WORKER_PING_INTERVAL_S", "10")) WORKER_PING_INTERVAL_S = int(os.getenv("WORKER_PING_INTERVAL_S", "10"))
xtts/whisper/bridge.py
+5 -1
View File
@@ -77,7 +77,11 @@ WHISPER_LANGUAGE = os.getenv("WHISPER_LANGUAGE", "de")
# ── Compute-Fleet: Worker-Identitaet & Registrierung ────────────── # ── Compute-Fleet: Worker-Identitaet & Registrierung ──────────────
# Meldet sich bei der aria-bridge (worker_hello) + periodischer worker_ping. # Meldet sich bei der aria-bridge (worker_hello) + periodischer worker_ping.
NODE_NAME = os.getenv("NODE_NAME", "node").strip() or "node" NODE_NAME = os.getenv("NODE_NAME", "node").strip() or "node"
GPU_IDS = os.getenv("NVIDIA_VISIBLE_DEVICES", "").strip() # Welche Karte(n) dieser Dienst nutzt — fuer die Anzeige. ARIA_GPU setzt die
# Compose; NVIDIA_VISIBLE_DEVICES ersetzt das Container-Toolkit im Prozess
# durch "void", taugt also nicht zur Anzeige.
GPU_IDS = (os.getenv("ARIA_GPU", "").strip()
or os.getenv("NVIDIA_VISIBLE_DEVICES", "").strip().replace("void", ""))
WORKER_SERVICE = "whisper" WORKER_SERVICE = "whisper"
INSTANCE_ID = f"{WORKER_SERVICE}@{NODE_NAME}" INSTANCE_ID = f"{WORKER_SERVICE}@{NODE_NAME}"
WORKER_PING_INTERVAL_S = int(os.getenv("WORKER_PING_INTERVAL_S", "10")) WORKER_PING_INTERVAL_S = int(os.getenv("WORKER_PING_INTERVAL_S", "10"))