Feat/llm-based and ranking-based dm (#405)

* feat: create files for llm based response selection * feat: tests * feat: algorithm * fix: filter bad or toxic uttrs * feat: ranking based response selector * feat: model cards * feat: ranking and refactoring * feat: component files * feat: component description * feat: component description * feat: use gpt-j resp selector * feat: context * feat: utilize hypotheses correctly * fix: configs * fix: configs * fix: mapping * fix: utilize ranker based selector * fix: decay dummy * fix: decay dummy * fix: response * fix: remove old selector * fix: logs * fix: getting annotations * fix: code style * fix: deepy uses resp selector * fix: no language * fix: ru usage * fix: usage of formatter * fix: codestyle * fix: logs * fix: after review * fix: sync
deeppavlov · Apr 25, 2023 · e2dfd33 · e2dfd33
1 parent ae14c7a
commit e2dfd33
Show file tree

Hide file tree

Showing 62 changed files with 2,323 additions and 587 deletions.
diff --git a/assistant_dists/ai_faq_assistant/dev.yml b/assistant_dists/ai_faq_assistant/dev.yml
@@ -10,12 +10,12 @@ services:
       - "./annotators/SentSeg:/src"
     ports:
       - 8011:8011
-  no-restrictions-selector:
+  ranking-based-response-selector:
     volumes:
-      - "./response_selectors/convers_evaluation_based_selector:/src"
+      - "./response_selectors/ranking_based_response_selector:/src"
       - "./common:/src/common"
     ports:
-      - 8009:8009
+      - 8002:8002
   badlisted-words:
     volumes:
       - "./annotators/BadlistedWordsDetector:/src"

diff --git a/assistant_dists/ai_faq_assistant/docker-compose.override.yml b/assistant_dists/ai_faq_assistant/docker-compose.override.yml
@@ -2,7 +2,7 @@ services:
   agent:
     command: sh -c 'bin/wait && python -m deeppavlov_agent.run agent.pipeline_config=assistant_dists/ai_faq_assistant/pipeline_conf.json'
     environment:
-      WAIT_HOSTS: "sentseg:8011, no-restrictions-selector:8009, badlisted-words:8018, combined-classification:8087, 
+      WAIT_HOSTS: "sentseg:8011, ranking-based-response-selector:8002, badlisted-words:8018, combined-classification:8087, 
         sentence-ranker:8128, prompt-selector:8135, transformers-lm-oasst12b:8158, 
         dff-ai-faq-prompted-skill:8143"
       WAIT_HOSTS_TIMEOUT: ${WAIT_TIMEOUT:-1000}
@@ -38,32 +38,21 @@ services:
         reservations:
           memory: 2G
 
-  no-restrictions-selector:
+  ranking-based-response-selector:
     env_file: [ .env ]
     build:
       args:
-        TAG_BASED_SELECTION: 1
-        CALL_BY_NAME_PROBABILITY: 0.5
-        PROMPT_PROBA: 0
-        ACKNOWLEDGEMENT_PROBA: 0.3
-        PRIORITIZE_WITH_REQUIRED_ACT: 0
-        PRIORITIZE_NO_DIALOG_BREAKDOWN: 0
-        PRIORITIZE_WITH_SAME_TOPIC_ENTITY: 0
-        IGNORE_DISLIKED_SKILLS: 0
-        GREETING_FIRST: 0
-        RESTRICTION_FOR_SENSITIVE_CASE: 0
-        PRIORITIZE_PROMTS_WHEN_NO_SCRIPTS: 0
-        MAX_TURNS_WITHOUT_SCRIPTS: 100
-        ADD_ACKNOWLEDGMENTS_IF_POSSIBLE: 0
-        PRIORITIZE_SCRIPTED_SKILLS: 0
-        CONFIDENCE_STRENGTH: 0.8
-        CONV_EVAL_STRENGTH: 0.4
-        PRIORITIZE_HUMAN_INITIATIVE: 1
-        QUESTION_TO_QUESTION_DOWNSCORE_COEF: 0.8
+        SERVICE_PORT: 8002
+        SERVICE_NAME: response_selector
         LANGUAGE: EN
+        SENTENCE_RANKER_ANNOTATION_NAME: sentence_ranker
+        SENTENCE_RANKER_SERVICE_URL: http://sentence-ranker:8128/respond
+        SENTENCE_RANKER_TIMEOUT: 3
+        N_UTTERANCES_CONTEXT: 5
+        FILTER_TOXIC_OR_BADLISTED: 1
       context: .
-      dockerfile: ./response_selectors/convers_evaluation_based_selector/Dockerfile
-    command: flask run -h 0.0.0.0 -p 8009
+      dockerfile: ./response_selectors/ranking_based_response_selector/Dockerfile
+    command: flask run -h 0.0.0.0 -p 8002
     environment:
       - FLASK_APP=server
     deploy:

diff --git a/assistant_dists/ai_faq_assistant/pipeline_conf.json b/assistant_dists/ai_faq_assistant/pipeline_conf.json
@@ -273,18 +273,18 @@
                 "connector": {
                     "protocol": "http",
                     "timeout": 1.0,
-                    "url": "http://no-restrictions-selector:8009/respond"
+                    "url": "http://ranking-based-response-selector:8002/respond"
                 },
-                "dialog_formatter": "state_formatters.dp_formatters:full_history_dialog",
+                "dialog_formatter": "state_formatters.dp_formatters:cropped_dialog",
                 "response_formatter": "state_formatters.dp_formatters:base_response_selector_formatter_service",
                 "previous_services": [
                     "candidate_annotators"
                 ],
                 "state_manager_method": "add_bot_utterance",
                 "is_enabled": true,
                 "source": {
-                    "directory": "response_selectors/convers_evaluation_based_selector",
-                    "container": "no-restrictions-selector",
+                    "directory": "response_selectors/ranking_based_response_selector",
+                    "container": "ranking-based-response-selector",
                     "endpoint": "respond"
                 }
             }

diff --git a/assistant_dists/deeppavlov_assistant/dev.yml b/assistant_dists/deeppavlov_assistant/dev.yml
@@ -5,12 +5,12 @@ services:
       - ".:/dp-agent"
     ports:
       - 4242:4242
-  no-restrictions-selector:
+  ranking-based-response-selector:
     volumes:
-      - "./response_selectors/convers_evaluation_based_selector:/src"
+      - "./response_selectors/ranking_based_response_selector:/src"
       - "./common:/src/common"
     ports:
-      - 8009:8009
+      - 8002:8002
   sentence-ranker:
     volumes:
       - "./services/sentence_ranker:/src"

diff --git a/assistant_dists/deeppavlov_assistant/docker-compose.override.yml b/assistant_dists/deeppavlov_assistant/docker-compose.override.yml
@@ -2,40 +2,29 @@ services:
   agent:
     command: sh -c 'bin/wait && python -m deeppavlov_agent.run agent.pipeline_config=assistant_dists/deeppavlov_assistant/pipeline_conf.json'
     environment:
-      WAIT_HOSTS: "no-restrictions-selector:8009, sentence-ranker:8128, prompt-selector:8135, openai-api-chatgpt:8145,
+      WAIT_HOSTS: "ranking-based-response-selector:8002, sentence-ranker:8128, prompt-selector:8135, openai-api-chatgpt:8145,
         dff-deeppavlov-prompted-skill:8146"
       WAIT_HOSTS_TIMEOUT: ${WAIT_TIMEOUT:-1000}
       HIGH_PRIORITY_INTENTS: 1
       RESTRICTION_FOR_SENSITIVE_CASE: 1
       ALWAYS_TURN_ON_ALL_SKILLS: 0
       LANGUAGE: EN
 
-  no-restrictions-selector:
+  ranking-based-response-selector:
     env_file: [ .env ]
     build:
       args:
-        TAG_BASED_SELECTION: 1
-        CALL_BY_NAME_PROBABILITY: 0.5
-        PROMPT_PROBA: 0
-        ACKNOWLEDGEMENT_PROBA: 0.3
-        PRIORITIZE_WITH_REQUIRED_ACT: 0
-        PRIORITIZE_NO_DIALOG_BREAKDOWN: 0
-        PRIORITIZE_WITH_SAME_TOPIC_ENTITY: 0
-        IGNORE_DISLIKED_SKILLS: 0
-        GREETING_FIRST: 0
-        RESTRICTION_FOR_SENSITIVE_CASE: 0
-        PRIORITIZE_PROMTS_WHEN_NO_SCRIPTS: 0
-        MAX_TURNS_WITHOUT_SCRIPTS: 100
-        ADD_ACKNOWLEDGMENTS_IF_POSSIBLE: 0
-        PRIORITIZE_SCRIPTED_SKILLS: 0
-        CONFIDENCE_STRENGTH: 0.8
-        CONV_EVAL_STRENGTH: 0.4
-        PRIORITIZE_HUMAN_INITIATIVE: 1
-        QUESTION_TO_QUESTION_DOWNSCORE_COEF: 0.8
+        SERVICE_PORT: 8002
+        SERVICE_NAME: response_selector
         LANGUAGE: EN
+        SENTENCE_RANKER_ANNOTATION_NAME: sentence_ranker
+        SENTENCE_RANKER_SERVICE_URL: http://sentence-ranker:8128/respond
+        SENTENCE_RANKER_TIMEOUT: 3
+        N_UTTERANCES_CONTEXT: 5
+        FILTER_TOXIC_OR_BADLISTED: 1
       context: .
-      dockerfile: ./response_selectors/convers_evaluation_based_selector/Dockerfile
-    command: flask run -h 0.0.0.0 -p 8009
+      dockerfile: ./response_selectors/ranking_based_response_selector/Dockerfile
+    command: flask run -h 0.0.0.0 -p 8002
     environment:
       - FLASK_APP=server
     deploy:

diff --git a/assistant_dists/deeppavlov_assistant/pipeline_conf.json b/assistant_dists/deeppavlov_assistant/pipeline_conf.json
@@ -159,18 +159,18 @@
                 "connector": {
                     "protocol": "http",
                     "timeout": 1.0,
-                    "url": "http://no-restrictions-selector:8009/respond"
+                    "url": "http://ranking-based-response-selector:8002/respond"
                 },
-                "dialog_formatter": "state_formatters.dp_formatters:full_history_dialog",
+                "dialog_formatter": "state_formatters.dp_formatters:cropped_dialog",
                 "response_formatter": "state_formatters.dp_formatters:base_response_selector_formatter_service",
                 "previous_services": [
                     "candidate_annotators"
                 ],
                 "state_manager_method": "add_bot_utterance",
                 "is_enabled": true,
                 "source": {
-                    "directory": "response_selectors/convers_evaluation_based_selector",
-                    "container": "no-restrictions-selector",
+                    "directory": "response_selectors/ranking_based_response_selector",
+                    "container": "ranking-based-response-selector",
                     "endpoint": "respond"
                 }
             }

diff --git a/assistant_dists/deepy_assistant/cpu.yml b/assistant_dists/deepy_assistant/cpu.yml
@@ -1,9 +1,5 @@
 version: '3.7'
 services:
-  combined-classification:
-    environment:
-      DEVICE: cpu
-      CUDA_VISIBLE_DEVICES: ""
   sentence-ranker:
     environment:
       DEVICE: cpu

diff --git a/assistant_dists/deepy_assistant/dev.yml b/assistant_dists/deepy_assistant/dev.yml
@@ -5,29 +5,12 @@ services:
       - ".:/dp-agent"
     ports:
       - 4242:4242
-  sentseg:
-    volumes:
-      - "./annotators/SentSeg:/src"
-    ports:
-      - 8011:8011
-  no-restrictions-selector:
-    volumes:
-      - "./response_selectors/convers_evaluation_based_selector:/src"
-      - "./common:/src/common"
-    ports:
-      - 8009:8009
-  badlisted-words:
-    volumes:
-      - "./annotators/BadlistedWordsDetector:/src"
-      - "./common:/src/common"
-    ports:
-      - 8018:8018
-  combined-classification:
+  ranking-based-response-selector:
     volumes:
+      - "./response_selectors/ranking_based_response_selector:/src"
       - "./common:/src/common"
-      - "./annotators/combined_classification:/src"
     ports:
-      - 8087:8087
+      - 8002:8002
   sentence-ranker:
     volumes:
       - "./services/sentence_ranker:/src"

diff --git a/assistant_dists/deepy_assistant/docker-compose.override.yml b/assistant_dists/deepy_assistant/docker-compose.override.yml
@@ -2,68 +2,29 @@ services:
   agent:
     command: sh -c 'bin/wait && python -m deeppavlov_agent.run agent.pipeline_config=assistant_dists/deepy_assistant/pipeline_conf.json'
     environment:
-      WAIT_HOSTS: "sentseg:8011, no-restrictions-selector:8009, badlisted-words:8018, combined-classification:8087, 
-                   sentence-ranker:8128, prompt-selector:8135, openai-api-davinci3:8131,
-                   dff-deepy-prompted-skill:8148"
+      WAIT_HOSTS: "ranking-based-response-selector:8002, sentence-ranker:8128, prompt-selector:8135, openai-api-davinci3:8131,
+        dff-deepy-prompted-skill:8148"
       WAIT_HOSTS_TIMEOUT: ${WAIT_TIMEOUT:-1000}
+      HIGH_PRIORITY_INTENTS: 1
+      RESTRICTION_FOR_SENSITIVE_CASE: 1
+      ALWAYS_TURN_ON_ALL_SKILLS: 0
+      LANGUAGE: EN
 
-  sentseg:
-    env_file: [ .env ]
-    build:
-      context: ./annotators/SentSeg/
-    command: flask run -h 0.0.0.0 -p 8011
-    environment:
-      - FLASK_APP=server
-    deploy:
-      resources:
-        limits:
-          memory: 1.5G
-        reservations:
-          memory: 1.5G
-
-  combined-classification:
-    env_file: [ .env ]
-    build:
-      args:
-        CONFIG: combined_classifier.json
-        SERVICE_PORT: 8087
-      context: .
-      dockerfile: ./annotators/combined_classification/Dockerfile
-    environment:
-      - CUDA_VISIBLE_DEVICES=0
-    deploy:
-      resources:
-        limits:
-          memory: 2G
-        reservations:
-          memory: 2G
-
-  no-restrictions-selector:
+  ranking-based-response-selector:
     env_file: [ .env ]
     build:
       args:
-        TAG_BASED_SELECTION: 1
-        CALL_BY_NAME_PROBABILITY: 0.5
-        PROMPT_PROBA: 0
-        ACKNOWLEDGEMENT_PROBA: 0.3
-        PRIORITIZE_WITH_REQUIRED_ACT: 0
-        PRIORITIZE_NO_DIALOG_BREAKDOWN: 0
-        PRIORITIZE_WITH_SAME_TOPIC_ENTITY: 0
-        IGNORE_DISLIKED_SKILLS: 0
-        GREETING_FIRST: 0
-        RESTRICTION_FOR_SENSITIVE_CASE: 0
-        PRIORITIZE_PROMTS_WHEN_NO_SCRIPTS: 0
-        MAX_TURNS_WITHOUT_SCRIPTS: 100
-        ADD_ACKNOWLEDGMENTS_IF_POSSIBLE: 0
-        PRIORITIZE_SCRIPTED_SKILLS: 0
-        CONFIDENCE_STRENGTH: 0.8
-        CONV_EVAL_STRENGTH: 0.4
-        PRIORITIZE_HUMAN_INITIATIVE: 1
-        QUESTION_TO_QUESTION_DOWNSCORE_COEF: 0.8
+        SERVICE_PORT: 8002
+        SERVICE_NAME: response_selector
         LANGUAGE: EN
+        SENTENCE_RANKER_ANNOTATION_NAME: sentence_ranker
+        SENTENCE_RANKER_SERVICE_URL: http://sentence-ranker:8128/respond
+        SENTENCE_RANKER_TIMEOUT: 3
+        N_UTTERANCES_CONTEXT: 5
+        FILTER_TOXIC_OR_BADLISTED: 1
       context: .
-      dockerfile: ./response_selectors/convers_evaluation_based_selector/Dockerfile
-    command: flask run -h 0.0.0.0 -p 8009
+      dockerfile: ./response_selectors/ranking_based_response_selector/Dockerfile
+    command: flask run -h 0.0.0.0 -p 8002
     environment:
       - FLASK_APP=server
     deploy:
@@ -73,23 +34,6 @@ services:
         reservations:
           memory: 100M
 
-  badlisted-words:
-    env_file: [ .env ]
-    build:
-      args:
-        SERVICE_PORT: 8018
-        SERVICE_NAME: badlisted_words
-      context: annotators/BadlistedWordsDetector/
-    command: flask run -h 0.0.0.0 -p 8018
-    environment:
-      - FLASK_APP=server
-    deploy:
-      resources:
-        limits:
-          memory: 256M
-        reservations:
-          memory: 256M
-
   prompt-selector:
     env_file: [ .env ]
     build: