dream

1
services:
2
  agent:
3
    command: sh -c 'bin/wait && python -m deeppavlov_agent.run agent.pipeline_config=assistant_dists/document_based_qa_transformers/pipeline_conf.json'
4
    environment:
5
      WAIT_HOSTS: "sentseg:8011, combined-classification:8087, ranking-based-response-selector:8002,
6
        sentence-ranker:8128, transformers-lm-mistral-7b-128k:8185, doc-retriever:8165, dff-document-qa-transformers-llm-skill:8186"
7
      WAIT_HOSTS_TIMEOUT: ${WAIT_TIMEOUT:-1000}
8
      HIGH_PRIORITY_INTENTS: 1
9
      RESTRICTION_FOR_SENSITIVE_CASE: 1
10
      ALWAYS_TURN_ON_ALL_SKILLS: 0
11
      LANGUAGE: EN
12

13
  files:
14
    image: julienmeerschart/simple-file-upload-download-server
15

16
  sentseg:
17
    env_file: [ .env ]
18
    build:
19
      context: ./annotators/SentSeg/
20
      dockerfile: Dockerfile-test
21
    command: flask run -h 0.0.0.0 -p 8011
22
    environment:
23
      - FLASK_APP=server
24
    deploy:
25
      resources:
26
        limits:
27
          memory: 1.5G
28
        reservations:
29
          memory: 1.5G
30

31
  doc-retriever:
32
    env_file: [ .env ]
33
    build:
34
      context: .
35
      dockerfile: ./annotators/doc_retriever/Dockerfile
36
      args:
37
        SERVICE_PORT: 8165
38
        SERVICE_NAME: doc_retriever
39
        CONFIG_PATH: ./doc_retriever_config.json
40
        DOC_PATH_OR_LINK: http://files.deeppavlov.ai/dream_data/documents_for_qa/test_file_dream_repo.html,http://files.deeppavlov.ai/dream_data/documents_for_qa/alphabet_financial_report.txt,http://files.deeppavlov.ai/dream_data/documents_for_qa/test_file_jurafsky_chatbots.pdf
41
        PARAGRAPHS_NUM: 5
42
        FILE_SERVER_TIMEOUT: 30
43
    command: python -m flask run -h 0.0.0.0 -p 8165
44
    environment:
45
      - FLASK_APP=server
46
      - CUDA_VISIBLE_DEVICES=0
47
    deploy:
48
      resources:
49
        limits:
50
          memory: 5G
51
        reservations:
52
          memory: 5G
53

54
  combined-classification:
55
    env_file: [ .env ]
56
    build:
57
      args:
58
        CONFIG: combined_classifier.json
59
        SERVICE_PORT: 8087
60
      context: .
61
      dockerfile: ./annotators/combined_classification/Dockerfile
62
    command: gunicorn --workers=1 server:app -b 0.0.0.0:8087 --timeout 600
63
    environment:
64
      - CUDA_VISIBLE_DEVICES=0
65
    deploy:
66
      resources:
67
        limits:
68
          memory: 2G
69
        reservations:
70
          memory: 2G
71

72
  ranking-based-response-selector:
73
    env_file: [ .env ]
74
    build:
75
      args:
76
        SERVICE_PORT: 8002
77
        SERVICE_NAME: response_selector
78
        LANGUAGE: EN
79
        SENTENCE_RANKER_ANNOTATION_NAME: sentence_ranker
80
        SENTENCE_RANKER_SERVICE_URL: http://sentence-ranker:8128/respond
81
        SENTENCE_RANKER_TIMEOUT: 3
82
        N_UTTERANCES_CONTEXT: 5
83
        FILTER_TOXIC_OR_BADLISTED: 1
84
      context: .
85
      dockerfile: ./response_selectors/ranking_based_response_selector/Dockerfile
86
    command: flask run -h 0.0.0.0 -p 8002
87
    environment:
88
      - FLASK_APP=server
89
    deploy:
90
      resources:
91
        limits:
92
          memory: 100M
93
        reservations:
94
          memory: 100M
95

96
  sentence-ranker:
97
    env_file: [ .env ]
98
    build:
99
      args:
100
        SERVICE_PORT: 8128
101
        SERVICE_NAME: sentence_ranker
102
        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
103
      context: ./services/sentence_ranker/
104
    command: flask run -h 0.0.0.0 -p 8128
105
    environment:
106
      - CUDA_VISIBLE_DEVICES=0
107
      - FLASK_APP=server
108
    deploy:
109
      resources:
110
        limits:
111
          memory: 3G
112
        reservations:
113
          memory: 3G
114

115
  transformers-lm-mistral-7b-128k:
116
    env_file: [ .env ]
117
    build:
118
      args:
119
        SERVICE_PORT: 8185
120
        SERVICE_NAME: transformers_lm_mistral_7b_128k
121
        PRETRAINED_MODEL_NAME_OR_PATH: NousResearch/Yarn-Mistral-7b-128k
122
        HALF_PRECISION: 1
123
        USE_FLASH_ATTENTION_2: 1
124
      context: .
125
      dockerfile: ./services/transformers_lm/Dockerfile
126
    command: flask run -h 0.0.0.0 -p 8185
127
    environment:
128
      - CUDA_VISIBLE_DEVICES=0
129
      - FLASK_APP=server
130
    deploy:
131
      resources:
132
        limits:
133
          memory: 50G
134
        reservations:
135
          memory: 50G
136

137
  dff-document-qa-transformers-llm-skill:
138
    env_file: [ .env ]
139
    build:
140
      args:
141
        SERVICE_PORT: 8186
142
        SERVICE_NAME: dff_document_qa_llm_skill
143
        GENERATIVE_SERVICE_URL: http://transformers-lm-mistral-7b-128k:8185/respond
144
        GENERATIVE_SERVICE_CONFIG: transformers_mistral.json
145
        GENERATIVE_TIMEOUT: 120
146
        N_UTTERANCES_CONTEXT: 7
147
        FILE_SERVER_TIMEOUT: 30
148
        DOCUMENT_PROMPT_FILE: common/prompts/document_qa_instruction.json
149
      context: .
150
      dockerfile: ./skills/dff_document_qa_llm_skill/Dockerfile
151
    deploy:
152
      resources:
153
        limits:
154
          memory: 128M
155
        reservations:
156
          memory: 128M
157

158
version: '3.7'
159