rorapp/cluster-knaster
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

TTS-Slots (Piper CPU, Qwen3-TTS auf Ada), M4000 außer Betrieb, Engine-Grenzen im Placement (ADR-0006)
ci / test (push) Successful in 24s
Details

Browse Source 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
rorapp
2026-10-12 02:12:30 +02:00
co-authored by Claude Opus 5.5
parent 0533f224f1
commit f90b7df31c
25 changed files with 586 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sites/knaster/inventory.yaml
+4 -4
View File
@@ -14,15 +14,15 @@ nodes:
memory_gb: 32
disks: [{name: sda, size_gb: 128}, {name: sdb, size_gb: 256}]
gpus:
- {index: 0, name: "Quadro M4000", memory_gb: 8, compute_cap: 5.2}
- {index: 1, name: "Quadro M4000", memory_gb: 8, compute_cap: 5.2}
- {index: 0, name: "Quadro M4000", memory_gb: 8, compute_cap: 5.2, disabled: true, note: "außer Betrieb bis llama-bench + Strommessung (ADR-0006)"}
- {index: 1, name: "Quadro M4000", memory_gb: 8, compute_cap: 5.2, disabled: true, note: "außer Betrieb bis llama-bench + Strommessung (ADR-0006)"}
- name: pve2
cpu: {model: "Intel Core i7 (7. Gen) 3.6 GHz", cores: 4, threads: 8}
memory_gb: 32
disks: [{name: sda, size_gb: 128}, {name: sdb, size_gb: 256}, {name: sdc, size_gb: 256}]
gpus:
- {index: 0, name: "Quadro M4000", memory_gb: 8, compute_cap: 5.2}
- {index: 1, name: "Quadro M4000", memory_gb: 8, compute_cap: 5.2}
- {index: 0, name: "Quadro M4000", memory_gb: 8, compute_cap: 5.2, disabled: true, note: "außer Betrieb bis llama-bench + Strommessung (ADR-0006)"}
- {index: 1, name: "Quadro M4000", memory_gb: 8, compute_cap: 5.2, disabled: true, note: "außer Betrieb bis llama-bench + Strommessung (ADR-0006)"}
- name: pve3
cpu: {model: "Intel Core i7 (7. Gen) 3.6 GHz", cores: 4, threads: 8}
memory_gb: 32
sites/knaster/registry.yaml
+23
View File
@@ -80,4 +80,27 @@ slots:
kind: rerank
ops: {docker_host: gpu-embed-1, container: llamacpp-rerank, pve_node: pve3, vmid: "221"} # VM-IDs/Hosts: ANNAHME bis Deployment
# --- Sprachausgabe (ADR-0006) ---
tts-de:
tier: 1
device: cpu
engine: piper
node: infra
litellm_model: tts-de
base_model: piper-de-thorsten-medium
max_context: 5000 # Zeichen pro Anfrage (TTS-Dienst-Limit)
kind: tts
ops: {docker_host: infra, container: tts-piper, pve_node: pve1, vmid: "101"} # ANNAHME bis Deployment
tts-de-hq:
tier: 1
device: gpu
engine: pytorch
node: gpu-flex # teilt sich die Ada mit tier1-flex; während Training unavailable
litellm_model: tts-de-hq
base_model: qwen3-tts-1.7b
max_context: 5000
kind: tts
ops: {docker_host: gpu-flex, container: qwen-tts, pve_node: pve-ada, vmid: "202"} # ANNAHME bis Deployment
adapters: []