Qwen3-TTS-Openai-Fastapi/docker-compose.yml at main · StompZone/Qwen3-TTS-Openai-Fastapi · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
version: '3.8'

services:
  # GPU-enabled service with official backend (default)
  qwen3-tts-gpu:
    build:
      context: .
      dockerfile: Dockerfile
      target: production
    container_name: qwen3-tts-api
    network_mode: host
    ports:
      - "8880:8880"
    environment:
      - HOST=0.0.0.0
      - PORT=8880
      - WORKERS=1
      - CORS_ORIGINS=*
      - TTS_BACKEND=official
      - TTS_WARMUP_ON_START=false
      # When using device_ids filter, the GPU appears as device 0 inside container
      - CUDA_VISIBLE_DEVICES=0
      - NVIDIA_DRIVER_CAPABILITIES=compute,utility
    volumes:
      # Mount model cache for persistence
      - ~/.cache/huggingface:/home/appuser/.cache/huggingface
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              device_ids: ['2']
              capabilities: [gpu]
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8880/health"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 120s

  # GPU-enabled service with vLLM-Omni backend
  qwen3-tts-vllm:
    build:
      context: .
      dockerfile: Dockerfile.vllm
    container_name: qwen3-tts-api-vllm
    network_mode: host
    ports:
      - "8880:8880"
    environment:
      - HOST=0.0.0.0
      - PORT=8880
      - WORKERS=1
      - CORS_ORIGINS=*
      - TTS_BACKEND=vllm_omni
      - TTS_WARMUP_ON_START=true
      - TTS_MODEL_NAME=Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice
      - VLLM_WORKER_MULTIPROC_METHOD=spawn
      - NVIDIA_DRIVER_CAPABILITIES=compute,utility
      - HF_HOME=/root/.cache/huggingface
    volumes:
      # Mount model cache for persistence
      - ~/.cache/huggingface:/root/.cache/huggingface
    ipc: host
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              device_ids: ['2']
              capabilities: [gpu]
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8880/health"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 180s
    profiles:
      - vllm

  # CPU-only service
  qwen3-tts-cpu:
    build:
      context: .
      dockerfile: Dockerfile
      target: cpu-base
    container_name: qwen3-tts-api-cpu
    network_mode: host
    ports:
      - "8880:8880"
    environment:
      - HOST=0.0.0.0
      - PORT=8880
      - WORKERS=1
      - CORS_ORIGINS=*
      - TTS_BACKEND=official
      - TTS_MODEL_NAME=Qwen/Qwen3-TTS-12Hz-1.7B-Base
    volumes:
      - ~/.cache/huggingface:/home/appuser/.cache/huggingface
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8880/health"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 180s
    profiles:
      - cpu

# To run GPU version with official backend: docker-compose up qwen3-tts-gpu
# To run GPU version with vLLM backend: docker-compose --profile vllm up qwen3-tts-vllm
# To run CPU version: docker-compose --profile cpu up qwen3-tts-cpu