Merge pull request #394 from deeppavlov/dev

Release v0.18.0
deeppavlov · Apr 17, 2023 · 9ee02d4 · 9ee02d4
2 parents 254d4cb + bdaa9d0
commit 9ee02d4
Show file tree

Hide file tree

Showing 21 changed files with 70 additions and 141 deletions.
diff --git a/Jenkinsfile b/Jenkinsfile
@@ -256,105 +256,6 @@ pipeline {
         }
       }
     }
-
-    stage('Build-ML') {
-      steps {
-        script{
-          startTime = currentBuild.duration
-          Exception ex = null
-          catchError(buildResult: 'FAILURE', stageResult: 'FAILURE') {
-            try {
-              sh '''
-                tests/runtests_russian.sh MODE=clean
-                tests/runtests_multilingual.sh MODE=build
-              '''
-            }
-            catch (Exception e) {
-              int duration = (currentBuild.duration - startTime) / 1000
-              throw e
-            }
-          }
-        }
-      }
-      post {
-        failure {
-          script {
-            sh 'tests/runtests_multilingual.sh MODE=clean'
-          }
-        }
-        success {
-          script {
-            int duration = (currentBuild.duration - startTime) / 1000
-          }
-        }
-      }
-    }
-
-    stage('Start-ML') {
-      steps {
-        script {
-          startTime = currentBuild.duration
-          Exception ex = null
-          catchError(buildResult: 'FAILURE', stageResult: 'FAILURE') {
-            try {
-              sh 'tests/runtests_multilingual.sh MODE=clean && tests/runtests_multilingual.sh MODE=start'
-            }
-            catch (Exception e) {
-              int duration = (currentBuild.duration - startTime) / 1000
-              throw e
-            }
-          }
-        }
-      }
-      post {
-        failure {
-          script {
-            sh 'tests/runtests_multilingual.sh MODE=clean'
-          }
-        }
-        success {
-          script {
-            started = true
-            int duration = (currentBuild.duration - startTime) / 1000
-          }
-        }
-        aborted {
-          script {
-            sh 'tests/runtests_multilingual.sh MODE=clean'
-          }
-        }
-      }
-    }
-
-    stage('Test skills-ML') {
-      steps {
-        script {
-          startTime = currentBuild.duration
-          Exception ex = null
-          catchError(buildResult: 'FAILURE', stageResult: 'FAILURE') {
-            try {
-              sh label: 'test skills', script: 'tests/runtests_multilingual.sh MODE=test_skills'
-            }
-            catch (Exception e) {
-              int duration = (currentBuild.duration - startTime) / 1000
-              throw e
-            }
-          }
-        }
-      }
-      post {
-        success {
-          script {
-            int duration = (currentBuild.duration - startTime) / 1000
-          }
-        }
-        aborted {
-          script {
-            sh 'tests/runtests_multilingual.sh MODE=clean'
-          }
-        }
-      }
-    }
   }
   post {
     aborted {

diff --git a/annotators/relative_persona_extractor/test.py b/annotators/relative_persona_extractor/test.py
@@ -19,19 +19,19 @@ def main():
     }
     gold = [
         {
-            "max_similarity": 0.695,
+            "max_similarity": 0.551,
             "persona": [
                 "I like Italian food especially pasta and pizza.",
+                "I like visiting interesting places virtually.",
                 "I like to watch football and basketball on TV.",
-                "I like watching travel video blogs.",
             ],
         },
         {
-            "max_similarity": 0.645,
+            "max_similarity": 0.474,
             "persona": [
+                "I like visiting interesting places virtually.",
                 "I like watching travel video blogs.",
-                "I like to watch football and basketball on TV.",
-                "I like Italian food especially pasta and pizza.",
+                "I like travelling.",
             ],
         },
     ]

diff --git a/assistant_dists/ai_faq_assistant/docker-compose.override.yml b/assistant_dists/ai_faq_assistant/docker-compose.override.yml
@@ -133,7 +133,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:
@@ -177,7 +177,7 @@ services:
         GENERATIVE_SERVICE_URL: http://transformers-lm-oasst12b:8158/respond
         GENERATIVE_SERVICE_CONFIG: default_generative_config.json
         GENERATIVE_TIMEOUT: 5
-        N_UTTERANCES_CONTEXT: 3
+        N_UTTERANCES_CONTEXT: 7
       context: .
       dockerfile: ./skills/dff_template_prompted_skill/Dockerfile
     command: gunicorn --workers=1 server:app -b 0.0.0.0:8143 --reload

diff --git a/assistant_dists/da_costa_clothes_assistant/docker-compose.override.yml b/assistant_dists/da_costa_clothes_assistant/docker-compose.override.yml
@@ -133,7 +133,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:
@@ -177,7 +177,7 @@ services:
         GENERATIVE_SERVICE_URL: http://transformers-lm-oasst12b:8158/respond
         GENERATIVE_SERVICE_CONFIG: default_generative_config.json
         GENERATIVE_TIMEOUT: 5
-        N_UTTERANCES_CONTEXT: 3
+        N_UTTERANCES_CONTEXT: 7
       context: .
       dockerfile: ./skills/dff_template_prompted_skill/Dockerfile
     command: gunicorn --workers=1 server:app -b 0.0.0.0:8144 --reload

diff --git a/assistant_dists/deepy_assistant/docker-compose.override.yml b/assistant_dists/deepy_assistant/docker-compose.override.yml
@@ -133,7 +133,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:
@@ -176,7 +176,7 @@ services:
         GENERATIVE_SERVICE_URL: http://openai-api-davinci3:8131/respond
         GENERATIVE_SERVICE_CONFIG: openai-text-davinci-003-long.json
         GENERATIVE_TIMEOUT: 7
-        N_UTTERANCES_CONTEXT: 3
+        N_UTTERANCES_CONTEXT: 7
         ENVVARS_TO_SEND: OPENAI_API_KEY,OPENAI_ORGANIZATION
       context: .
       dockerfile: ./skills/dff_template_prompted_skill/Dockerfile

diff --git a/assistant_dists/dream/docker-compose.override.yml b/assistant_dists/dream/docker-compose.override.yml
@@ -1276,7 +1276,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:

diff --git a/assistant_dists/dream_mini/docker-compose.override.yml b/assistant_dists/dream_mini/docker-compose.override.yml
@@ -177,7 +177,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:

diff --git a/assistant_dists/dream_mini_persona_based/docker-compose.override.yml b/assistant_dists/dream_mini_persona_based/docker-compose.override.yml
@@ -179,7 +179,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:

diff --git a/assistant_dists/dream_multimodal/docker-compose.override.yml b/assistant_dists/dream_multimodal/docker-compose.override.yml
@@ -181,7 +181,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:

diff --git a/assistant_dists/dream_persona_openai_prompted/docker-compose.override.yml b/assistant_dists/dream_persona_openai_prompted/docker-compose.override.yml
@@ -133,7 +133,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:
@@ -176,7 +176,7 @@ services:
         GENERATIVE_SERVICE_URL: http://openai-api-davinci3:8131/respond
         GENERATIVE_SERVICE_CONFIG: openai-text-davinci-003.json
         GENERATIVE_TIMEOUT: 7
-        N_UTTERANCES_CONTEXT: 3
+        N_UTTERANCES_CONTEXT: 7
         ENVVARS_TO_SEND: OPENAI_API_KEY,OPENAI_ORGANIZATION
       context: .
       dockerfile: ./skills/dff_template_prompted_skill/Dockerfile

diff --git a/assistant_dists/dream_persona_prompted/docker-compose.override.yml b/assistant_dists/dream_persona_prompted/docker-compose.override.yml
@@ -133,7 +133,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:
@@ -177,7 +177,7 @@ services:
         GENERATIVE_SERVICE_URL: http://transformers-lm-gptj:8130/respond
         GENERATIVE_SERVICE_CONFIG: default_generative_config.json
         GENERATIVE_TIMEOUT: 5
-        N_UTTERANCES_CONTEXT: 3
+        N_UTTERANCES_CONTEXT: 7
       context: .
       dockerfile: ./skills/dff_template_prompted_skill/Dockerfile
     command: gunicorn --workers=1 server:app -b 0.0.0.0:8134 --reload

diff --git a/assistant_dists/empathetic_marketing_assistant/docker-compose.override.yml b/assistant_dists/empathetic_marketing_assistant/docker-compose.override.yml
@@ -133,7 +133,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:
@@ -177,7 +177,7 @@ services:
         GENERATIVE_SERVICE_URL: http://transformers-lm-gptj:8130/respond
         GENERATIVE_SERVICE_CONFIG: default_generative_config.json
         GENERATIVE_TIMEOUT: 5
-        N_UTTERANCES_CONTEXT: 3
+        N_UTTERANCES_CONTEXT: 7
       context: .
       dockerfile: ./skills/dff_template_prompted_skill/Dockerfile
     command: gunicorn --workers=1 server:app -b 0.0.0.0:8138 --reload

diff --git a/assistant_dists/fairytale_assistant/docker-compose.override.yml b/assistant_dists/fairytale_assistant/docker-compose.override.yml
@@ -133,7 +133,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:
@@ -176,7 +176,7 @@ services:
         GENERATIVE_SERVICE_URL: http://openai-api-chatgpt:8145/respond
         GENERATIVE_SERVICE_CONFIG: openai-chatgpt.json
         GENERATIVE_TIMEOUT: 7
-        N_UTTERANCES_CONTEXT: 3
+        N_UTTERANCES_CONTEXT: 7
         ENVVARS_TO_SEND: OPENAI_API_KEY,OPENAI_ORGANIZATION
       context: .
       dockerfile: ./skills/dff_template_prompted_skill/Dockerfile

diff --git a/assistant_dists/nutrition_assistant/docker-compose.override.yml b/assistant_dists/nutrition_assistant/docker-compose.override.yml
@@ -133,7 +133,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:
@@ -177,7 +177,7 @@ services:
         GENERATIVE_SERVICE_URL: http://transformers-lm-oasst12b:8158/respond
         GENERATIVE_SERVICE_CONFIG: default_generative_config.json
         GENERATIVE_TIMEOUT: 5
-        N_UTTERANCES_CONTEXT: 3
+        N_UTTERANCES_CONTEXT: 7
       context: .
       dockerfile: ./skills/dff_template_prompted_skill/Dockerfile
     command: gunicorn --workers=1 server:app -b 0.0.0.0:8141 --reload

diff --git a/assistant_dists/rhodes_coaching_assistant/docker-compose.override.yml b/assistant_dists/rhodes_coaching_assistant/docker-compose.override.yml
@@ -133,7 +133,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:
@@ -176,7 +176,7 @@ services:
         GENERATIVE_SERVICE_URL: http://openai-api-davinci3:8131/respond
         GENERATIVE_SERVICE_CONFIG: openai-text-davinci-003.json
         GENERATIVE_TIMEOUT: 7
-        N_UTTERANCES_CONTEXT: 3
+        N_UTTERANCES_CONTEXT: 7
         ENVVARS_TO_SEND: OPENAI_API_KEY,OPENAI_ORGANIZATION
       context: .
       dockerfile: ./skills/dff_template_prompted_skill/Dockerfile

diff --git a/assistant_dists/universal_prompted_assistant/docker-compose.override.yml b/assistant_dists/universal_prompted_assistant/docker-compose.override.yml
@@ -114,7 +114,7 @@ services:
       args:
         SERVICE_PORT: 8128
         SERVICE_NAME: sentence_ranker
-        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+        PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
       context: ./services/sentence_ranker/
     command: flask run -h 0.0.0.0 -p 8128
     environment:
@@ -216,7 +216,7 @@ services:
         SERVICE_PORT: 8147
         SERVICE_NAME: dff_universal_prompted_skill
         GENERATIVE_TIMEOUT: 7
-        N_UTTERANCES_CONTEXT: 3
+        N_UTTERANCES_CONTEXT: 7
       context: .
       dockerfile: ./skills/dff_universal_prompted_skill/Dockerfile
     command: gunicorn --workers=1 server:app -b 0.0.0.0:8147 --reload

diff --git a/services/openai_api_lm/server.py b/services/openai_api_lm/server.py
@@ -71,8 +71,10 @@ def generate_responses(context, openai_api_key, openai_org, prompt, generation_p
     elif isinstance(response, str):
         outputs = [response.strip()]
 
-    outputs = [GENERATIVE_ROBOT_TEMPLATE.sub("\n", resp).strip() for resp in outputs]
-    outputs = [resp.split("\n")[0] for resp in outputs]
+    if PRETRAINED_MODEL_NAME_OR_PATH != "gpt-3.5-turbo":
+        # post-processing of the responses by all models except of ChatGPT
+        outputs = [GENERATIVE_ROBOT_TEMPLATE.sub("\n", resp).strip() for resp in outputs]
+        outputs = [resp.split("\n")[0] for resp in outputs]
     return outputs
 
 

diff --git a/services/sentence_ranker/component.yml b/services/sentence_ranker/component.yml
@@ -8,8 +8,8 @@ sentence-ranker:
   author: DeepPavlov
   description: ranking model given as `PRETRAINED_MODEL_NAME_OR_PATH` which for a
     pair os sentences returns a float score of correspondence
-  ram_usage: 1.2G
-  gpu_usage: 2.1G
+  ram_usage: 1.5G
+  gpu_usage: 800M
   port: 8128
   endpoints:
   - group: candidate_annotators
@@ -19,7 +19,7 @@ sentence-ranker:
   build_args:
     SERVICE_PORT: 8128
     SERVICE_NAME: sentence_ranker
-    PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/bert-base-nli-mean-tokens
+    PRETRAINED_MODEL_NAME_OR_PATH: sentence-transformers/all-MiniLM-L6-v2
   date_created: '2023-03-16T09:45:32'
 sentence-ranker-multilingual:
   name: sentence_ranker