minuteman/docker-compose.yml at master · fkmjec/minuteman · 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
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
version: '3.9'

services:
  flask:
    build: ./flask
    environment:
      # OPTION: if torchserve is not running, this must be kept to "true". This turns off the polling for available models
      MOCK_ML_MODELS: "false"
      POSTGRES_DB_URL: "postgresql://postgres:postgres@postgres:5432/"
      # TODO: supply your flask secret key here; generate a long random string and paste it here
      FLASK_SECRET_KEY: "[YOUR_SECRET_KEY]"
      ETHERPAD_API_URL: "http://etherpad:9001/api"
      # TODO: set this to the domain where you will be running your etherpad instance. To be reliable,
      # it should be on the same domain as the Flask frontend to enable iframing of the editor.
      ETHERPAD_URL: "[YOUR_ETHERPAD_URL]"
      TORCHSERVE_MANAGEMENT_URL: "http://torchserve:8085"
      PYTHONBUFFERED: 1
    command: gunicorn main:app -w 2 --threads 2 -b 0.0.0.0:7777 --timeout 3600 --log-level info
    restart: always
    volumes:
      - /SOME_LOCAL_FOLDER_FOR_STORING_AUDIO_DATA:/audio
    depends_on:
      - rabbitmq
      - postgres
    # OPTION: set the running port for the application.
    ports:
      - 7777:7777

  trsc-worker:
    build: ./transcription_worker
    environment:
      # OPTION: you can select a more or less powerful model here if needed and possible with your hardware
      WHISPER_MODEL: "large-v3"
    command: python transcription_worker.py
    restart: always
    depends_on:
      - rabbitmq
      - translation-worker
    volumes:
      - /SOME_LOCAL_WHISPER_MODEL_FOLDER:/whisper_model
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

    translation-worker:
    build: ./translation_worker
    environment:
      MOCK_ML_MODELS: "false"
      CT_MODEL_PATH: "/nllb-200-600M-int8"
      SP_MODEL_PATH: "/flores200_sacrebleu_tokenizer_spm.model"
      PYTHONBUFFERED: 1
      DATA_FOLDER_PATH: "/minuteman_data"
    command: gunicorn main:app -w 1 --threads 1 -b 0.0.0.0:7778 --timeout 3600 --log-level info
    restart: always
    # OPTION: mount the model and tokenizer files
    volumes:
      - /SOME_LOCAL_NLLB_MODEL_FOLDER/nllb-200-600M-int8:/nllb-200-600M-int8
      - /SOME_LOCAL_NLLB_MODEL_FOLDER/flores200_sacrebleu_tokenizer_spm.model:/flores200_sacrebleu_tokenizer_spm.model
      - /SOME_LOCAL_WHISPER_MODEL_FOLDER:/whisper_model
      - /SOME_LOCAL_DATA_FOLDER:/minuteman_data
    # OPTION: set the running port for the application.
    ports:
      - 7778:7778
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              device_ids: ["0", "1"]
              capabilities: [gpu]

  summ-worker:
    build: ./summarization_worker
    environment:
      # OPTION: when set to true, this replaces requests to torchserve models with a mock string
      PYTHONBUFFERED: 1
      MOCK_ML_MODELS: "false"
      TORCH_BACKEND_URL: "http://torchserve:8084"
    command: python summarization_worker.py
    restart: always
    depends_on:
      - rabbitmq
      - translation-worker

  torchserve:
    build: ./torchserve
    # OPTION: you can provide additional models here if you created them
    command: torchserve --ts-config /torch-config.cnf --model-store /torch_model_dir --models bart=lidiya-bart.mar --log-config /torchserve-log4j2.xml --start
    volumes:
     - ./torch_model_dir:/torch_model_dir
     - ./torch-config.cnf:/torch-config.cnf
     - ./torchserve-log4j2.xml:/torchserve-log4j2.xml
    deploy:
      resources:
        reservations:
         devices:
           - driver: nvidia
             device_ids: ["1"]
             capabilities: [gpu]

  postgres:
    image: postgres:15-alpine
    restart: always
    environment:
      POSTGRES_USER: postgres
      POSTGRES_PASSWORD: postgres
      PGDATA: /var/lib/postgresql/data/pgdata
    volumes:
      - pgdata:/var/lib/postgresql/data
      - ./flask/db/create_db.sql:/docker-entrypoint-initdb.d/create_tables.sql
    command: ["postgres", "-c", "log_statement=none", "-c", "log_checkpoints=off"]

  etherpad:
    build: ./etherpad-lite
    ports:
      - 9001:9001
    environment:
      TITLE: Minuteman
      # TODO: set the Etherpad admin password
      ADMIN_PASSWORD: "YOUR ADMIN PASSWORD"
      ADMIN_USER: admin
      DB_TYPE: postgres
      DB_HOST: postgres
      DB_PORT: 5432
      DB_USER: postgres
      DB_PASS: postgres
      DB_NAME: postgres
      DB_CHARSET: utf8mb4
    restart: no
    depends_on:
      - postgres
      - rabbitmq
    volumes:
      - plugins:/opt/etherpad-lite/src/plugin_packages
      - etherpad-var:/opt/etherpad-lite/var

  rabbitmq:
    build: ./rabbitmq
    image: rabbitmq
    volumes:
      - ./rabbitmq:/var/lib/rabbitmq
    # OPTION: rabbitMQ ports are forwarded to the host by default, as it is often the case that you want to stream a transcript into the editor from an external script
    ports:
      - 5672:5672
      - 15672:15672

volumes:
  pgdata:
  plugins:
  etherpad-var: