Merge pull request #612 from deeppavlov/feat/add-gpt-4o

Add new model gpt-4o

assistant_dists/universal_prompted_assistant/dev.yml

@@ -71,6 +71,12 @@ services:
       - "./common:/src/common"
     ports:
       - 8180:8180
+  openai-api-gpt4o:
+    volumes:
+      - "./services/openai_api_lm:/src"
+      - "./common:/src/common"
+    ports:
+      - 8181:8181
   gigachat-api:
     volumes:
       - "./services/gigachat_api_lm:/src"

assistant_dists/universal_prompted_assistant/docker-compose.override.yml

@@ -6,8 +6,8 @@ services:
         sentence-ranker:8128, 
         transformers-lm-gptjt:8161, openai-api-chatgpt:8145, openai-api-davinci3:8131,
         openai-api-gpt4:8159, openai-api-gpt4-32k:8160, openai-api-chatgpt-16k:8167,
-        openai-api-gpt4-turbo:8180, gigachat-api:8187, gigachat-pro-api:8188, gigachat-plus-api:8189, 
-        dff-universal-prompted-skill:8147"
+        openai-api-gpt4-turbo:8180, openai-api-gpt4o:8181, gigachat-api:8187, gigachat-pro-api:8188, 
+        gigachat-plus-api:8189, dff-universal-prompted-skill:8147"
       WAIT_HOSTS_TIMEOUT: ${WAIT_TIMEOUT:-1000}
 
   sentseg:
@@ -220,6 +220,25 @@ services:
         reservations:
           memory: 100M
 
+  openai-api-gpt4o:
+    env_file: [ .env ]
+    build:
+      args:
+        SERVICE_PORT: 8181
+        SERVICE_NAME: openai_api_gpt4o
+        PRETRAINED_MODEL_NAME_OR_PATH: gpt-4o
+      context: .
+      dockerfile: ./services/openai_api_lm/Dockerfile
+    command: flask run -h 0.0.0.0 -p 8181
+    environment:
+      - FLASK_APP=server
+    deploy:
+      resources:
+        limits:
+          memory: 500M
+        reservations:
+          memory: 100M
+
   gigachat-api:
     env_file: [ .env ]
     build:

assistant_dists/universal_ru_prompted_assistant/dev.yml

@@ -67,6 +67,12 @@ services:
       - "./common:/src/common"
     ports:
       - 8180:8180
+  openai-api-gpt4o:
+    volumes:
+      - "./services/openai_api_lm:/src"
+      - "./common:/src/common"
+    ports:
+      - 8181:8181
   gigachat-api:
     volumes:
       - "./services/gigachat_api_lm:/src"

assistant_dists/universal_ru_prompted_assistant/docker-compose.override.yml

@@ -6,8 +6,8 @@ services:
         dialogrpt-ru:8122, transformers-lm-ruxglm:8171, transformers-lm-rugpt35:8178, 
         openai-api-chatgpt:8145, openai-api-davinci3:8131,
         openai-api-gpt4:8159, openai-api-gpt4-32k:8160, openai-api-chatgpt-16k:8167,
-        openai-api-gpt4-turbo:8180, gigachat-api:8187, gigachat-pro-api:8188, gigachat-plus-api:8189, 
-        dff-universal-ru-prompted-skill:8161"
+        openai-api-gpt4-turbo:8180, openai-api-gpt4o:8181, gigachat-api:8187, gigachat-pro-api:8188, 
+        gigachat-plus-api:8189, dff-universal-ru-prompted-skill:8161"
       WAIT_HOSTS_TIMEOUT: ${WAIT_TIMEOUT:-1800}
       HIGH_PRIORITY_INTENTS: 0
       RESTRICTION_FOR_SENSITIVE_CASE: 0
@@ -214,6 +214,25 @@ services:
         reservations:
           memory: 100M
 
+  openai-api-gpt4o:
+    env_file: [ .env ]
+    build:
+      args:
+        SERVICE_PORT: 8181
+        SERVICE_NAME: openai_api_gpt4o
+        PRETRAINED_MODEL_NAME_OR_PATH: gpt-4o
+      context: .
+      dockerfile: ./services/openai_api_lm/Dockerfile
+    command: flask run -h 0.0.0.0 -p 8181
+    environment:
+      - FLASK_APP=server
+    deploy:
+      resources:
+        limits:
+          memory: 500M
+        reservations:
+          memory: 100M
+
   gigachat-api:
     env_file: [ .env ]
     build:

components.tsv

@@ -184,7 +184,7 @@
 8178    transformers-lm-rugpt35
 8179	dff-robot-prompted-skill
 8180    openai-api-gpt4-turbo
-8181
+8181    openai-api-gpt-4o
 8182
 8183	external-integration-skill
 8184	external-fake-server

components/jaWueh23hfsdias9cdjasdf.yml

@@ -0,0 +1,25 @@
+name: openai_api_gpt4o
+display_name: GPT-4o
+component_type: Generative
+model_type: NN-based
+is_customizable: false
+author: [email protected]
+description: GPT-4o (“o” for “omni”) is our most advanced model. It is multimodal (accepting text or image inputs and 
+  outputting text), and it has the same high intelligence as GPT-4 Turbo but is much more efficient—it generates text 
+  2x faster and is 50% cheaper. The maximum number of output tokens for this model is 128000.
+ram_usage: 100M
+gpu_usage: null
+group: services
+connector:
+  protocol: http
+  timeout: 120.0
+  url: http://openai-api-gpt4o:8181/respond
+dialog_formatter: null
+response_formatter: null
+previous_services: null
+required_previous_services: null
+state_manager_method: null
+tags: null
+endpoint: respond
+service: services/openai_api_lm/service_configs/openai-api-gpt-4o
+date_created: '2024-05-15T09:45:32'

services/openai_api_lm/server.py

@@ -32,7 +32,7 @@
     "gpt-4-32k": json.load(open("common/generative_configs/openai-chatgpt.json", "r")),
     "gpt-4-1106-preview": json.load(open("common/generative_configs/openai-chatgpt.json", "r")),
 }
-CHAT_COMPLETION_MODELS = ["gpt-3.5-turbo", "gpt-3.5-turbo-16k", "gpt-4", "gpt-4-32k", "gpt-4-1106-preview"]
+CHAT_COMPLETION_MODELS = ["gpt-3.5-turbo", "gpt-3.5-turbo-16k", "gpt-4", "gpt-4-32k", "gpt-4-1106-preview", "gpt-4o"]
 
 
 def generate_responses(context, openai_api_key, openai_org, prompt, generation_params, continue_last_uttr=False):

services/openai_api_lm/service_configs/openai-api-gpt-4o/environment.yml

@@ -0,0 +1,4 @@
+SERVICE_PORT: 8181
+SERVICE_NAME: openai-api-gpt4o
+PRETRAINED_MODEL_NAME_OR_PATH: gpt-4o
+FLASK_APP: server

services/openai_api_lm/service_configs/openai-api-gpt-4o/service.yml

@@ -0,0 +1,30 @@
+name: openai-api-gpt4o
+endpoints:
+- respond
+- generate_goals
+compose:
+  env_file:
+  - .env
+  build:
+    args:
+      SERVICE_PORT: 8181
+      SERVICE_NAME: openai_api_gpt4_o
+      PRETRAINED_MODEL_NAME_OR_PATH: gpt-4o
+      FLASK_APP: server
+    context: .
+    dockerfile: ./services/openai_api_lm/Dockerfile
+  command: flask run -h 0.0.0.0 -p 8181
+  environment:
+  - FLASK_APP=server
+  deploy:
+    resources:
+      limits:
+        memory: 100M
+      reservations:
+        memory: 100M
+  volumes:
+  - ./services/openai_api_lm:/src
+  - ./common:/src/common
+  ports:
+  - 8181:8181
+proxy: null