dialoGPT Persona

annotators/sentence_ranker/Dockerfile

@@ -0,0 +1,23 @@
+# syntax=docker/dockerfile:experimental
+
+FROM pytorch/pytorch:1.5-cuda10.1-cudnn7-runtime
+
+WORKDIR /src
+
+ARG PRETRAINED_MODEL_NAME_OR_PATH
+ENV PRETRAINED_MODEL_NAME_OR_PATH ${PRETRAINED_MODEL_NAME_OR_PATH}
+ARG SERVICE_PORT
+ENV SERVICE_PORT ${SERVICE_PORT}
+# ARG N_HYPOTHESES_TO_GENERATE
+# ENV N_HYPOTHESES_TO_GENERATE ${N_HYPOTHESES_TO_GENERATE}
+
+
+COPY ./requirements.txt /src/requirements.txt
+COPY ./persona_sentences.txt /src/persona_sentences.txt
+RUN pip install -r /src/requirements.txt
+
+RUN python -c "from sentence_transformers import SentenceTransformer;SentenceTransformer('${PRETRAINED_MODEL_NAME_OR_PATH}')"
+
+COPY . /src
+
+CMD gunicorn --workers=1 server:app -b 0.0.0.0:${SERVICE_PORT} --timeout=300

annotators/sentence_ranker/persona_sentences.txt

@@ -0,0 +1,30 @@
+I used to be ADDICTED to Diet Coke. 
+I hate puzzles. 
+I was in a commercial for my aunt and uncle’s store. 
+I have a scar on my left hand from where I cat bit me when I was babysitting when I was 12.
+I used to try to sleep in my splits when I was a dancer because I wanted to have perfect flexibility.
+I love being outside, but I hate smelling like outside.
+I am afraid of jack-in-the-boxes even though I know when they are going to pop out.
+In second grade, I fell off the counter I was standing on trying to get a mug for milk and I had to wear an ice pack in my pants to school for a couple weeks. 
+For the third grade play, the part I wanted the most was to play the mom.
+I almost went to the University of Iowa to major in English and be on the dance team.
+The high school I graduated from, I only went to for my senior year. 
+My sister, Caroline, and I get asked if we are twins all the time, and when people find out we aren’t actually twins, they always think she’s older.
+I’m really close to my Dad.
+When my parents first met Cam’s parents, they kept saying how Cam’s mom looked so familiar, but couldn’t figure out where they’d seen her before. 
+Cam and I planned for him to ask me out on November 15th because 15 was his favorite number.
+When I was younger I was OBSESSED with Jesse McCartney in the way that some people nowadays are obsessed with One Direction.
+I either get told I look like Hilary Duff or Leighton Meester. I’m ok with either of those!!
+A lot of people can’t figure out what nationality I am. Some people say I look 100% American and some people think I’m Asian. I am 50% Honduran and my mom is from the states.
+No, I am not fluent in Spanish.
+I haven’t had un-painted toenails in 6 years. I just don’t like the way my toenails look without polish haha!!
+I was a drill team officer in high school and it was one of the most challenging and rewarding positions.
+I’m a literature nerd. I LOVE Shakespeare, Jane Austen, Charlotte Bronte, and anything Greek mythology.
+I ALMOST got certified to be a Zumba instructor, but then I was transferring so I didn’t. 
+Oddly enough, I think my favorite food ever is tuna salad.
+I’m obsessed with libraries and churches. When I’m in a new town, I like to wander around those if I can find them!
+My favorite place in the world is Eureka Springs, Arkansas.
+I didn’t get a cell phone until my freshman year of high school, and I didn’t get an iPhone until I was a junior in high school.
+If I could have any job in the world, I would own a coffee shop that also sold art, flowers, and stationary. 
+I plan on working for myself at some point in my life. 
+I’m insanely good with state capitols.

annotators/sentence_ranker/requirements.txt

@@ -0,0 +1,11 @@
+transformers==4.20.1
+flask==1.1.1
+itsdangerous==2.0.1
+gunicorn==19.9.0
+requests==2.22.0
+sentry-sdk[flask]==0.14.1
+healthcheck==1.3.3
+jinja2<=3.0.3
+Werkzeug<=2.0.3
+sentence-transformers==2.2.2
+huggingface-hub==0.4.0

annotators/sentence_ranker/server.py

@@ -0,0 +1,131 @@
+from typing import List, Tuple
+from sentence_transformers import SentenceTransformer
+import logging
+import time
+import os
+
+import sentry_sdk
+import torch
+from flask import Flask, request, jsonify
+from sentry_sdk.integrations.flask import FlaskIntegration
+sentry_sdk.init(dsn=os.getenv("SENTRY_DSN"), integrations=[FlaskIntegration()])
+
+class SentenceRanker:
+    def __init__(self, 
+        persona_sentences=None, 
+        sentence_model=None
+    ):
+        """_summary_
+
+        Args:
+            persona_sentences (List[str]): список предложений составляющие полную персону. Defaults to None.
+            sentence_model (SentenceTransformer): модель для перевода предложения в вектор. Defaults to None.
+        """
+        self.persona_sentences = persona_sentences
+        self.sentence_model = sentence_model
+        self.sentence_embeddings = self.sentence_model.encode(
+            persona_sentences, 
+            convert_to_tensor=True
+        )
+        # для кеширования похожих запросов
+        self.ranked_sentences = {}
+
+    def rank_sentences(self, query, k):
+        """возвращает топ k предложений которые похожи на query
+
+        Args:
+            query (str): предложение, на основе которого ищем похожие
+            k (int): количество возвращаемых предложений. Defaults to 5.
+
+        Returns:
+            List[List[str], float]: отранжированные предложения и максимальное косинусное расстояние среди всех 
+        """
+        key = f"{query}_{k}"
+        if self.ranked_sentences.get(key, False):
+            return self.ranked_sentences[key]
+
+        user_sentence_embeddings = self.sentence_model.encode(query, convert_to_tensor=True)
+
+        cos_sim_ranks = self.cos_sim(
+            user_sentence_embeddings,
+            self.sentence_embeddings
+        )
+
+        top_indices = torch.argsort(cos_sim_ranks, descending=True)
+        max_similarity = float(cos_sim_ranks[top_indices][0])
+        top_indices = list(top_indices[:k].cpu().numpy())
+        similar_sentences = [self.persona_sentences[idx] for idx in top_indices]
+        self.ranked_sentences[key] = similar_sentences, max_similarity 
+        return [similar_sentences, max_similarity]
+
+    def cos_sim(self, a, b):
+        """возвращает косинусное расстояние 
+
+        K - количество предложений для сравнения
+        N - размерность возвращаемого вектора
+        Args:
+            a (torch.FloatTensor): shape (1, N)
+            b (torch.FloatTensor): shape (K, N)
+
+        Returns:
+            torch.FloatTensor: shape (1, K) тензор с косинусными расстояниями
+        """
+        a_norm = torch.nn.functional.normalize(a, p=2, dim=1)
+        b_norm = torch.nn.functional.normalize(b, p=2, dim=1)
+        return torch.sum(a_norm * b_norm, dim=1)
+
+logging.basicConfig(format="%(asctime)s - %(name)s - %(levelname)s - %(message)s", level=logging.INFO)
+# logging.getLogger("werkzeug").setLevel("INFO")
+logger = logging.getLogger(__name__)
+
+PRETRAINED_MODEL_NAME_OR_PATH = os.environ.get("PRETRAINED_MODEL_NAME_OR_PATH")
+# logging.info(f'PRETRAINED_MODEL_NAME_OR_PATH = {PRETRAINED_MODEL_NAME_OR_PATH}')
+DEFAULT_CONFIDENCE = 10
+N_HYPOTHESES_TO_GENERATE = int(os.environ.get("N_HYPOTHESES_TO_GENERATE", 1))
+ZERO_CONFIDENCE = 0.0
+MAX_HISTORY_DEPTH = 3
+TOP_SIMILAR_SENTENCES = 5
+
+try:
+    sentence_model = SentenceTransformer(PRETRAINED_MODEL_NAME_OR_PATH)
+
+    persona = open("./persona_sentences.txt").read()
+    persona_sentences = persona.split("\n")
+    persona_sentences = [item for item in persona_sentences if len(item) > 0]
+
+    sentence_ranker = SentenceRanker(
+        persona_sentences=persona_sentences,
+        sentence_model=sentence_model
+    )
+    logger.info("sentence_ranker is ready")
+except Exception as e:
+    sentry_sdk.capture_exception(e)
+    logger.exception(e)
+    raise e
+
+app = Flask(__name__)
+
+@app.route("/response", methods=["POST"])
+def respond():
+    try:
+        dialogs = request.json.get("dialogs", [])
+        process_result = []
+        # берем последнюю реплику, затем забираем у нее результат работы аннотатора sentseg
+        context_str = dialogs[0]["human_utterances"][-1]["annotations"]['sentseg']['punct_sent']
+        max_likelihood_sentences, max_sentence_similarity = sentence_ranker.rank_sentences(
+            [context_str], 
+            k=TOP_SIMILAR_SENTENCES
+        )
+
+        process_result.append([
+            max_likelihood_sentences, 
+            max_sentence_similarity
+        ])
+
+    except Exception as exc:
+        logger.exception(exc)
+        sentry_sdk.capture_exception(exc)
+
+    return jsonify(
+        process_result
+    )

assistant_dists/dream_mini/cpu.yml

@@ -8,6 +8,10 @@ services:
     environment:
       DEVICE: cpu
       CUDA_VISIBLE_DEVICES: ""
+  dialogpt-persona:
+    environment:
+      DEVICE: cpu
+      CUDA_VISIBLE_DEVICES: ""
   intent-catcher:
     environment:
       DEVICE: cpu

assistant_dists/dream_mini/dev.yml

@@ -57,4 +57,17 @@ services:
       - "./services/dialogpt:/src"
     ports:
       - 8125:8125
+
+  dialogpt-persona:
+    volumes:
+      - "./services/dialogpt_persona:/src"
+    ports:
+      - 8131:8131
+
+  sentence-ranker:
+    volumes:
+      - "./annotators/sentence_ranker:/src"
+    ports:
+      - 8130:8130
+
 version: "3.7"

assistant_dists/dream_mini/docker-compose.override.yml

@@ -2,12 +2,10 @@ services:
   agent:
     command: sh -c 'bin/wait && python -m deeppavlov_agent.run -ch http_client -pl assistant_dists/dream_mini/pipeline_conf.json --cors'
     environment:
-      WAIT_HOSTS: "convers-evaluator-annotator:8004, dff-program-y-skill:8008, sentseg:8011, convers-evaluation-selector:8009, 
-          dff-intent-responder-skill:8012, intent-catcher:8014, badlisted-words:8018,
-          spelling-preprocessing:8074, dialogpt:8125"
+      WAIT_HOSTS: "convers-evaluator-annotator:8004, dff-program-y-skill:8008, sentseg:8011, convers-evaluation-selector:8009, dff-intent-responder-skill:8012, intent-catcher:8014, badlisted-words:8018, spelling-preprocessing:8074, dialogpt:8125, dialogpt-persona:8131, sentence-ranker:8130"
       WAIT_HOSTS_TIMEOUT: ${WAIT_TIMEOUT:-480}
   convers-evaluator-annotator:
-    env_file: [.env]
+    env_file: [ .env ]
     build:
       args:
         CONFIG: conveval.json
@@ -27,7 +25,7 @@ services:
           memory: 2G
 
   dff-program-y-skill:
-    env_file: [.env]
+    env_file: [ .env ]
     build:
       args:
         SERVICE_PORT: 8008
@@ -43,9 +41,8 @@ services:
         reservations:
           memory: 1024M
 
-
   sentseg:
-    env_file: [.env]
+    env_file: [ .env ]
     build:
       context: ./annotators/SentSeg/
     command: flask run -h 0.0.0.0 -p 8011
@@ -59,18 +56,18 @@ services:
           memory: 1.5G
 
   convers-evaluation-selector:
-    env_file: [.env]
+    env_file: [ .env ]
     build:
       args:
-        TAG_BASED_SELECTION: 1
+        TAG_BASED_SELECTION: 0
         CALL_BY_NAME_PROBABILITY: 0.5
-        PROMPT_PROBA: 0.3
+        PROMPT_PROBA: 0
         ACKNOWLEDGEMENT_PROBA: 0.3
         PRIORITIZE_WITH_REQUIRED_ACT: 1
         PRIORITIZE_NO_DIALOG_BREAKDOWN: 0
-        PRIORITIZE_WITH_SAME_TOPIC_ENTITY: 1
+        PRIORITIZE_WITH_SAME_TOPIC_ENTITY: 0
         IGNORE_DISLIKED_SKILLS: 0
-        GREETING_FIRST: 1
+        GREETING_FIRST: 0
         RESTRICTION_FOR_SENSITIVE_CASE: 1
         PRIORITIZE_PROMTS_WHEN_NO_SCRIPTS: 0
         ADD_ACKNOWLEDGMENTS_IF_POSSIBLE: 1
@@ -105,15 +102,15 @@ services:
           memory: 128M
 
   intent-catcher:
-    env_file: [.env]
+    env_file: [ .env ]
     build:
       context: .
       dockerfile: ./annotators/IntentCatcherTransformers/Dockerfile
       args:
         SERVICE_PORT: 8014
         CONFIG_NAME: intents_model_dp_config.json
         INTENT_PHRASES_PATH: intent_phrases.json
-    command:  python -m flask run -h 0.0.0.0 -p 8014
+    command: python -m flask run -h 0.0.0.0 -p 8014
     environment:
       - FLASK_APP=server
       - CUDA_VISIBLE_DEVICES=0
@@ -125,7 +122,7 @@ services:
           memory: 3.5G
 
   badlisted-words:
-    env_file: [.env]
+    env_file: [ .env ]
     build:
       context: annotators/BadlistedWordsDetector/
     command: flask run -h 0.0.0.0 -p 8018
@@ -139,7 +136,7 @@ services:
           memory: 256M
 
   spelling-preprocessing:
-    env_file: [.env]
+    env_file: [ .env ]
     build:
       context: ./annotators/spelling_preprocessing/
     command: flask run -h 0.0.0.0 -p 8074
@@ -172,4 +169,42 @@ services:
         reservations:
           memory: 2G
 
+  dialogpt-persona:
+    env_file: [ .env ]
+    build:
+      args:
+        SERVICE_PORT: 8131
+        SERVICE_NAME: dialogpt_persona
+        PRETRAINED_MODEL_NAME_OR_PATH: dim/dialogpt-medium-persona-chat
+      context: ./services/dialogpt_persona/
+    command: flask run -h 0.0.0.0 -p 8131
+    environment:
+      - CUDA_VISIBLE_DEVICES=0
+      - FLASK_APP=server
+    deploy:
+      resources:
+        limits:
+          memory: 2G
+        reservations:
+          memory: 2G
+
+  sentence-ranker:
+    env_file: [ .env ]
+    build:
+      args:
+        SERVICE_PORT: 8130
+        SERVICE_NAME: sentence_ranker
+        PRETRAINED_MODEL_NAME_OR_PATH: 'sentence-transformers/nli-distilroberta-base-v2'
+      context: ./annotators/sentence_ranker/
+    command: flask run -h 0.0.0.0 -p 8130
+    environment:
+      - CUDA_VISIBLE_DEVICES=0
+      - FLASK_APP=server
+    deploy:
+      resources:
+        limits:
+          memory: 1G
+        reservations:
+          memory: 1G
+
 version: '3.7'