dream

1
services:
2
  agent:
3
    command: sh -c 'bin/wait && python -m deeppavlov_agent.run agent.pipeline_config=assistant_dists/nutrition_assistant/pipeline_conf.json'
4
    environment:
5
      WAIT_HOSTS: "sentseg:8011, ranking-based-response-selector:8002, combined-classification:8087, 
6
        sentence-ranker:8128, prompt-selector:8135, transformers-lm-gptjt:8161, 
7
        dff-nutrition-prompted-skill:8141"
8
      WAIT_HOSTS_TIMEOUT: ${WAIT_TIMEOUT:-1000}
9
      HIGH_PRIORITY_INTENTS: 1
10
      RESTRICTION_FOR_SENSITIVE_CASE: 1
11
      ALWAYS_TURN_ON_ALL_SKILLS: 0
12
      LANGUAGE: EN
13
      FALLBACK_FILE: fallbacks_dream_en.json
14

15
  sentseg:
16
    env_file: [ .env ]
17
    build:
18
      context: ./annotators/SentSeg/
19
    command: flask run -h 0.0.0.0 -p 8011
20
    environment:
21
      - FLASK_APP=server
22
    deploy:
23
      resources:
24
        limits:
25
          memory: 1.5G
26
        reservations:
27
          memory: 1.5G
28

29
  combined-classification:
30
    env_file: [ .env ]
31
    build:
32
      args:
33
        CONFIG: combined_classifier.json
34
        SERVICE_PORT: 8087
35
      context: .
36
      dockerfile: ./annotators/combined_classification/Dockerfile
37
    command: gunicorn --workers=1 server:app -b 0.0.0.0:8087 --timeout 600
38
    environment:
39
      - CUDA_VISIBLE_DEVICES=0
40
    deploy:
41
      resources:
42
        limits:
43
          memory: 2G
44
        reservations:
45
          memory: 2G
46

47
  ranking-based-response-selector:
48
    env_file: [ .env ]
49
    build:
50
      args:
51
        SERVICE_PORT: 8002
52
        SERVICE_NAME: response_selector
53
        LANGUAGE: EN
54
        SENTENCE_RANKER_ANNOTATION_NAME: sentence_ranker
55
        SENTENCE_RANKER_SERVICE_URL: http://sentence-ranker:8128/respond
56
        SENTENCE_RANKER_TIMEOUT: 3
57
        N_UTTERANCES_CONTEXT: 5
58
        FILTER_TOXIC_OR_BADLISTED: 1
59
      context: .
60
      dockerfile: ./response_selectors/ranking_based_response_selector/Dockerfile
61
    command: flask run -h 0.0.0.0 -p 8002
62
    environment:
63
      - FLASK_APP=server
64
    deploy:
65
      resources:
66
        limits:
67
          memory: 100M
68
        reservations:
69
          memory: 100M
70

71
  prompt-selector:
72
    env_file: [ .env ]
73
    build:
74
      args:
75
        SERVICE_PORT: 8135
76
        SERVICE_NAME: prompt_selector
77
        SENTENCE_RANKER_SERVICE_URL: http://sentence-ranker:8128/respond
78
        N_SENTENCES_TO_RETURN: 3
79
        PROMPTS_TO_CONSIDER: nutrition
80
      context: .
81
      dockerfile: ./annotators/prompt_selector/Dockerfile
82
    command: flask run -h 0.0.0.0 -p 8135
83
    environment:
84
      - FLASK_APP=server
85
    deploy:
86
      resources:
87
        limits:
88
          memory: 100M
89
        reservations:
90
          memory: 100M
91

92
  sentence-ranker:
93
    env_file: [ .env ]
94
    build:
95
      args:
96
        SERVICE_PORT: 8128
97
        SERVICE_NAME: sentence_ranker
98
        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
99
      context: ./services/sentence_ranker/
100
    command: flask run -h 0.0.0.0 -p 8128
101
    environment:
102
      - CUDA_VISIBLE_DEVICES=0
103
      - FLASK_APP=server
104
    deploy:
105
      resources:
106
        limits:
107
          memory: 3G
108
        reservations:
109
          memory: 3G
110

111
  transformers-lm-gptjt:
112
    env_file: [ .env ]
113
    build:
114
      args:
115
        SERVICE_PORT: 8161
116
        SERVICE_NAME: transformers_lm_gptjt
117
        PRETRAINED_MODEL_NAME_OR_PATH: togethercomputer/GPT-JT-6B-v1
118
        HALF_PRECISION: 1
119
      context: .
120
      dockerfile: ./services/transformers_lm/Dockerfile
121
    command: flask run -h 0.0.0.0 -p 8161
122
    environment:
123
      - CUDA_VISIBLE_DEVICES=0
124
      - FLASK_APP=server
125
    deploy:
126
      resources:
127
        limits:
128
          memory: 50G
129
        reservations:
130
          memory: 50G
131

132
  dff-nutrition-prompted-skill:
133
    env_file: [ .env ]
134
    build:
135
      args:
136
        SERVICE_PORT: 8141
137
        SERVICE_NAME: dff_nutrition_prompted_skill
138
        PROMPT_FILE: common/prompts/nutrition.json
139
        GENERATIVE_SERVICE_URL: http://transformers-lm-gptjt:8161/respond
140
        GENERATIVE_SERVICE_CONFIG: default_generative_config.json
141
        GENERATIVE_TIMEOUT: 120
142
        N_UTTERANCES_CONTEXT: 7
143
      context: .
144
      dockerfile: ./skills/dff_template_prompted_skill/Dockerfile
145
    deploy:
146
      resources:
147
        limits:
148
          memory: 128M
149
        reservations:
150
          memory: 128M
151

152
version: '3.7'
153