InnerVoice/docker-compose.yml at main · arkano1dev/InnerVoice · 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
networks:
  TelegramNet:
    driver: bridge

services:
  whisper:
    # GPU (AMD ROCm). Do not apply low block I/O limits: model load needs normal disk I/O.
    # If Whisper exits 139 or restarts, see docs/GPU-TROUBLESHOOTING.md
    build:
      context: ./whisper
      dockerfile: Dockerfile
    container_name: innervoice-whisper
    restart: unless-stopped
    networks:
      - TelegramNet
    ports:
      - "9000:9000"
    volumes:
      - whisper_cache:/root/.cache
    environment:
      - WHISPER_MODEL=medium
      - WHISPER_FP16=true
      - WHISPER_BEAM_SIZE=1
      - WHISPER_CONDITION_ON_PREVIOUS_TEXT=false
      - WHISPER_TORCH_COMPILE=true
      - PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True
      - VRAM_THRESHOLD_FREE_MB=1536
      - ROCR_VISIBLE_DEVICES=0
      - HSA_OVERRIDE_GFX_VERSION=11.0.0
    # Mark this container as low-priority for the kernel OOM killer
    oom_score_adj: 1000
    deploy:
      resources:
        limits:
          memory: 5G
          cpus: "4"
    security_opt:
      - seccomp:unconfined
    cap_add:
      - SYS_PTRACE
    ipc: host
    shm_size: "2g"
    devices:
      - /dev/kfd
      - /dev/dri
    group_add:
      - video
      - render

  bot:
    build:
      context: ./bot
      dockerfile: Dockerfile
    container_name: innervoice-bot
    restart: unless-stopped
    depends_on:
      - whisper
    networks:
      - TelegramNet
    volumes:
      - ./.env:/app/.env:ro
      - audio_temp:/app/audios
    environment:
      - BOT_TOKEN=${BOT_TOKEN}
      - WHISPER_API_URL=http://whisper:9000
      - TZ=UTC
    deploy:
      resources:
        limits:
          memory: 512M
          cpus: "0.5"

volumes:
  audio_temp:
    driver: local
  whisper_cache:
    driver: local