{ "nbformat": 4, "nbformat_minor": 0, "metadata": { "colab": { "provenance": [] }, "kernelspec": { "name": "python3", "display_name": "Python 3" }, "language_info": { "name": "python" } }, "cells": [ { "cell_type": "markdown", "source": [ "## Решение дает 12 баллов из 15 при отправке в систему проверки stepik.\n", "\n" ], "metadata": { "id": "v_eGuOtAv_9G" } }, { "cell_type": "code", "execution_count": null, "metadata": { "id": "Yk-K_mHMs7ns", "colab": { "base_uri": "https://localhost:8080/" }, "outputId": "503662e2-b124-4f83-9c61-80ff1a412d8f" }, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m268.3/268.3 kB\u001b[0m \u001b[31m2.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m817.7/817.7 kB\u001b[0m \u001b[31m8.3 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.8/1.8 MB\u001b[0m \u001b[31m28.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m1.9/1.9 MB\u001b[0m \u001b[31m46.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m163.3/163.3 kB\u001b[0m \u001b[31m13.1 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m27.0/27.0 MB\u001b[0m \u001b[31m28.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m75.6/75.6 kB\u001b[0m \u001b[31m6.5 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m287.5/287.5 kB\u001b[0m \u001b[31m14.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m113.0/113.0 kB\u001b[0m \u001b[31m2.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m49.4/49.4 kB\u001b[0m \u001b[31m926.4 kB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m77.9/77.9 kB\u001b[0m \u001b[31m2.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m58.3/58.3 kB\u001b[0m \u001b[31m4.2 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m53.0/53.0 kB\u001b[0m \u001b[31m1.8 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m144.8/144.8 kB\u001b[0m \u001b[31m1.7 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[?25h" ] } ], "source": [ "!pip install openai langchain tiktoken langchain-openai langchain-community sentence_transformers faiss-cpu -q" ] }, { "cell_type": "code", "source": [ "!wget https://raw.githubusercontent.com/a-milenkin/LLM_practical_course/main/notebooks/utils.py" ], "metadata": { "id": "Keq0WAjLuPgJ", "colab": { "base_uri": "https://localhost:8080/" }, "outputId": "8fb2fde7-0152-4fc4-f52c-4406df5e548c" }, "execution_count": null, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "--2024-04-15 15:41:31-- https://raw.githubusercontent.com/a-milenkin/LLM_practical_course/main/notebooks/utils.py\n", "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.110.133, ...\n", "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.\n", "HTTP request sent, awaiting response... 200 OK\n", "Length: 10823 (11K) [text/plain]\n", "Saving to: ‘utils.py’\n", "\n", "\rutils.py 0%[ ] 0 --.-KB/s \rutils.py 100%[===================>] 10.57K --.-KB/s in 0s \n", "\n", "2024-04-15 15:41:31 (65.0 MB/s) - ‘utils.py’ saved [10823/10823]\n", "\n" ] } ] }, { "cell_type": "code", "source": [ "from utils import ChatOpenAI\n", "from getpass import getpass\n", "\n", "#course_api_key= \"Введите ваш API ключ с курса\"\n", "course_api_key = getpass(prompt='API key')\n", "\n", "# Инициализируем языковую модель\n", "llm = ChatOpenAI(temperature=0.0, course_api_key=course_api_key)" ], "metadata": { "id": "PCS_WedHtPRg", "colab": { "base_uri": "https://localhost:8080/" }, "outputId": "db7db222-e7a3-408b-adf6-6702dfa05e46" }, "execution_count": null, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "API key··········\n" ] } ] }, { "cell_type": "markdown", "source": [ "Загружаем файлы" ], "metadata": { "id": "0eRICp2Lq8s-" } }, { "cell_type": "code", "source": [ "!wget https://stepik.org/media/attachments/lesson/1084288/The_Daughter_of_The_Commandant.pdf" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "VWFprDaxqy9Y", "outputId": "cb74ccae-a601-40c8-f7fa-2169886236a5" }, "execution_count": null, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "--2024-04-15 15:46:10-- https://stepik.org/media/attachments/lesson/1084288/The_Daughter_of_The_Commandant.pdf\n", "Resolving stepik.org (stepik.org)... 178.248.239.111\n", "Connecting to stepik.org (stepik.org)|178.248.239.111|:443... connected.\n", "HTTP request sent, awaiting response... 200 OK\n", "Length: 198595 (194K) [application/pdf]\n", "Saving to: ‘The_Daughter_of_The_Commandant.pdf’\n", "\n", "The_Daughter_of_The 100%[===================>] 193.94K 444KB/s in 0.4s \n", "\n", "2024-04-15 15:46:12 (444 KB/s) - ‘The_Daughter_of_The_Commandant.pdf’ saved [198595/198595]\n", "\n" ] } ] }, { "cell_type": "code", "source": [ "import pandas as pd\n", "from tqdm import tqdm\n", "\n", "df = pd.read_csv('https://stepik.org/media/attachments/lesson/1084288/pushkin_questions.csv')" ], "metadata": { "id": "dBAX9Ldxq3uC" }, "execution_count": null, "outputs": [] }, { "cell_type": "markdown", "source": [ "Соберем все компоненты для RAG" ], "metadata": { "id": "gjrto68msT-A" } }, { "cell_type": "markdown", "source": [ "\n", "
\n" ], "metadata": { "id": "zDg0bWUhr3dz" } }, { "cell_type": "code", "source": [ "!pip install pypdf -q" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "uKIFOOMF4dpR", "outputId": "dbef3bd2-719d-46b4-e1af-df1dcd931e63" }, "execution_count": null, "outputs": [ { "output_type": "stream", "name": "stdout", "text": [ "\u001b[?25l \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m0.0/290.4 kB\u001b[0m \u001b[31m?\u001b[0m eta \u001b[36m-:--:--\u001b[0m\r\u001b[2K \u001b[91m━━━━━\u001b[0m\u001b[91m╸\u001b[0m\u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m41.0/290.4 kB\u001b[0m \u001b[31m1.1 MB/s\u001b[0m eta \u001b[36m0:00:01\u001b[0m\r\u001b[2K \u001b[91m━━━━━━━━━━━━━━━━\u001b[0m\u001b[91m╸\u001b[0m\u001b[90m━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m122.9/290.4 kB\u001b[0m \u001b[31m1.8 MB/s\u001b[0m eta \u001b[36m0:00:01\u001b[0m\r\u001b[2K \u001b[91m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[90m╺\u001b[0m \u001b[32m286.7/290.4 kB\u001b[0m \u001b[31m2.9 MB/s\u001b[0m eta \u001b[36m0:00:01\u001b[0m\r\u001b[2K \u001b[90m━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━\u001b[0m \u001b[32m290.4/290.4 kB\u001b[0m \u001b[31m2.6 MB/s\u001b[0m eta \u001b[36m0:00:00\u001b[0m\n", "\u001b[?25h" ] } ] }, { "cell_type": "code", "source": [ "# Определите Document loader для pdf\n", "from langchain_community.document_loaders import PyPDFLoader\n", "\n", "loader = PyPDFLoader(\"/content/The_Daughter_of_The_Commandant.pdf\")\n", "docs = loader.load() # метод load сразу разбивает текст на документы по страницам" ], "metadata": { "id": "WOwhANxRr3Ik" }, "execution_count": null, "outputs": [] }, { "cell_type": "markdown", "source": [ "Следующие две ячейки можно пропустить, если нас устраивает сплит в методе load" ], "metadata": { "id": "WmsW0czZYBV8" } }, { "cell_type": "code", "source": [ "docs = [docs[i].page_content for i in range(len(docs))] # собираем текст, если хотим использовать свой сплитер" ], "metadata": { "id": "P5d-XyjimVxb" }, "execution_count": null, "outputs": [] }, { "cell_type": "code", "source": [ "# Определите подходящий Splitter\n", "from langchain.text_splitter import (\n", " CharacterTextSplitter,\n", " RecursiveCharacterTextSplitter,\n", ")\n", "\n", "splitter = CharacterTextSplitter(\n", " separator=\"\\n\\n\", # символ-разделитель, по умолчанию переход к новому абзацу '\\n\\n'\n", " chunk_size=500, # размер документа в символах\n", " chunk_overlap=100, # насколько соседние документы могут перекрывать друг-друга\n", " length_function=len, # функция, по которой считается размер документа\n", " is_separator_regex=False, # является ли разделитель регулярным выражением\n", ")\n", "\n", "\n", "split_documents = splitter.create_documents(docs)\n", "len(split_documents)" ], "metadata": { "colab": { "base_uri": "https://localhost:8080/" }, "id": "g8kNIzR8h3i-", "outputId": "d8c6a59b-e0e3-4a45-dd4c-988f077705f7" }, "execution_count": null, "outputs": [ { "output_type": "execute_result", "data": { "text/plain": [ "4" ] }, "metadata": {}, "execution_count": 38 } ] }, { "cell_type": "markdown", "source": [ "# Выбираем Embedding models" ], "metadata": { "id": "IClLxZZytsQ0" } }, { "cell_type": "markdown", "source": [ "## 🔱 Эмбеддинги от OpenAI (API)" ], "metadata": { "id": "QNFHkIi9t4JE" } }, { "cell_type": "code", "source": [ "# Если используете ключ курса, запустите эту ячейку\n", "from utils import OpenAIEmbeddings\n", "\n", "embeddings_api_model = OpenAIEmbeddings(course_api_key=course_api_key)" ], "metadata": { "id": "a3loV8ySuB7e" }, "execution_count": null, "outputs": [] }, { "cell_type": "markdown", "source": [ "# Выбираем 🗂 Vector Store" ], "metadata": { "id": "mxBZ5FBMut1j" } }, { "cell_type": "code", "source": [ "from langchain.vectorstores import FAISS\n", "\n", "embeddings = OpenAIEmbeddings(course_api_key=course_api_key)\n", "db = FAISS.from_documents(\n", " split_documents[:4], embeddings\n", ")\n", "\n", "db.save_local(\"faiss_db\") # можно сохранить базу локально, указав путь" ], "metadata": { "id": "hRKdQT8QuPPT" }, "execution_count": null, "outputs": [] }, { "cell_type": "markdown", "source": [ "# Определим 🎣 Retriever" ], "metadata": { "id": "bHeBYyV-vPmu" } }, { "cell_type": "code", "source": [ "# Самый частый кейс - использование векторного хранилища и его методов для получения документов\n", "retriever = db.as_retriever(\n", " search_type=\"similarity\", # тип поиска похожих документов\n", " k=4, # количество возвращаемых документов (Default: 4)\n", " score_threshold=None, # минимальный порог для поиска \"similarity_score_threshold\"\n", ")" ], "metadata": { "id": "pcljBDDBvB2s" }, "execution_count": null, "outputs": [] }, { "cell_type": "markdown", "source": [ "# 🚰 RAG Pipeline - подключаем RAG к LLM" ], "metadata": { "id": "YzCDWTNdvy_J" } }, { "cell_type": "code", "source": [ "from langchain.schema import StrOutputParser\n", "from langchain_core.prompts import ChatPromptTemplate\n", "from langchain_core.runnables import RunnablePassthrough\n", "\n", "# Создаём простой шаблон\n", "template = \"\"\"\n", "Answer the question based only on the following context:\n", "\n", "{context}\n", "\n", "Question: {question}\n", "\"\"\"\n", "# Создаём промпт из шаблона\n", "prompt = ChatPromptTemplate.from_template(template)\n", "\n", "\n", "# Объявляем функцию, которая будет собирать строку из полученных документов\n", "def format_docs(docs):\n", " return \"\\n\\n\".join([d.page_content for d in docs])\n", "\n", "\n", "# Создаём цепочку\n", "chain = (\n", " {\"context\": retriever | format_docs, \"question\": RunnablePassthrough()}\n", " | prompt\n", " | llm\n", " | StrOutputParser()\n", ")" ], "metadata": { "id": "119g8AIqvB0x" }, "execution_count": null, "outputs": [] }, { "cell_type": "code", "source": [ "answers = [] # Список, где будем хранить ответы модели\n", "\n", "for text_input in tqdm(df['question']):\n", " answer = chain.invoke(text_input)\n", " answers.append(answer) # Добавляем ответ в список\n", " break # Для отладки. Уберите, когда убедитесь, что на одном примере работает" ], "metadata": { "id": "z5b0Ii9kvBx2", "colab": { "base_uri": "https://localhost:8080/" }, "outputId": "5a0e6f74-ecda-4100-9594-e3f87d2e3e1b" }, "execution_count": null, "outputs": [ { "output_type": "stream", "name": "stderr", "text": [ " 0%| | 0/15 [00:02