04_CAG_GC.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "! pip3 install -qU langchain-upstage"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "%load_ext dotenv\n",
    "%dotenv\n",
    "# set UPSTAGE_API_KEY"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "\n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'What is the method for scaling large language models (LLMs) called, which encompasses depthwise scaling and continued pretraining, and is introduced as an alternative to methods using mixture-of-experts?'"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from langchain_core.prompts import PromptTemplate\n",
    "from langchain_upstage import ChatUpstage\n",
    "from langchain_core.output_parsers import StrOutputParser\n",
    "\n",
    "\n",
    "llm = ChatUpstage()\n",
    "\n",
    "prompt_template = PromptTemplate.from_template(\n",
    "    \"\"\"\n",
    "    Please provide one question from the following text \n",
    "    regarding \"{keyword}\":\n",
    "    \n",
    "    ---\n",
    "    {text}\n",
    "    \"\"\"\n",
    ")\n",
    "chain = prompt_template | llm | StrOutputParser()\n",
    "keyword = \"DUS\"\n",
    "text = \"\"\"\n",
    "SOLAR 10.7B: Scaling Large Language Models with Simple yet Effective Depth Up-Scaling\n",
    "\n",
    "We introduce SOLAR 10.7B, a large language model (LLM) with 10.7 billion parameters, \n",
    "demonstrating superior performance in various natural language processing (NLP) tasks. \n",
    "Inspired by recent efforts to efficiently up-scale LLMs, \n",
    "we present a method for scaling LLMs called depth up-scaling (DUS), \n",
    "which encompasses depthwise scaling and continued pretraining.\n",
    "In contrast to other LLM up-scaling methods that use mixture-of-experts, \n",
    "DUS does not require complex changes to train and inference efficiently. \n",
    "We show experimentally that DUS is simple yet effective \n",
    "in scaling up high-performance LLMs from small ones. \n",
    "Building on the DUS model, we additionally present SOLAR 10.7B-Instruct, \n",
    "a variant fine-tuned for instruction-following capabilities, \n",
    "surpassing Mixtral-8x7B-Instruct. \n",
    "SOLAR 10.7B is publicly available under the Apache 2.0 license, \n",
    "promoting broad access and application in the LLM field.\n",
    "\"\"\"\n",
    "\n",
    "chain.invoke({\"keyword\": keyword, \"text\": text})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'SOLAR 10.7B is available under the Apache 2.0 license.'"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from langchain_core.prompts import PromptTemplate\n",
    "\n",
    "prompt_template = PromptTemplate.from_template(\n",
    "    \"\"\"\n",
    "    Please provide most correct answer from the following context. \n",
    "    If the answer is not present in the context, please write \"The information is not present in the context.\"\n",
    "    ---\n",
    "    Question: {question}\n",
    "    ---\n",
    "    Context: {Context}\n",
    "    \"\"\"\n",
    ")\n",
    "chain = prompt_template | llm | StrOutputParser()\n",
    "context = \"\"\"\n",
    "We introduce SOLAR 10.7B, a large language model (LLM) with 10.7 billion parameters, \n",
    "    demonstrating superior performance in various natural language processing (NLP) tasks. \n",
    "    Inspired by recent efforts to efficiently up-scale LLMs, \n",
    "    we present a method for scaling LLMs called depth up-scaling (DUS), \n",
    "    which encompasses depthwise scaling and continued pretraining.\n",
    "    In contrast to other LLM up-scaling methods that use mixture-of-experts, \n",
    "    DUS does not require complex changes to train and inference efficiently. \n",
    "    We show experimentally that DUS is simple yet effective \n",
    "    in scaling up high-performance LLMs from small ones. \n",
    "    Building on the DUS model, we additionally present SOLAR 10.7B-Instruct, \n",
    "    a variant fine-tuned for instruction-following capabilities, \n",
    "    surpassing Mixtral-8x7B-Instruct. \n",
    "    SOLAR 10.7B is publicly available under the Apache 2.0 license, \n",
    "    promoting broad access and application in the LLM field.\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'SOLAR 10.7B is available under the Apache 2.0 license.'"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "chain.invoke({\"question\": \"What is the license under which SOLAR 10.7B is available?\", \"Context\": context})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'The information is not present in the context.'"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "chain.invoke({\"question\": \"Who is the created SOLAR 10.7B?\", \"Context\": context})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'The information is not present in the context.'"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "chain.invoke({\"question\": \"Did Google provide resources for the SOLAR 10.7B project?\", \"Context\": context})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'DUS is good because it is a simple yet effective method for scaling up high-performance LLMs from small ones. It encompasses depthwise scaling and continued pretraining, and does not require complex changes to train and inference efficiently.'"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "chain.invoke({\"question\": \"Why DUS is good\", \"Context\": context})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The name of the variant fine-tuned for instruction-following capabilities is SOLAR 10.7B-Instruct.\n"
     ]
    }
   ],
   "source": [
    "answer = chain.invoke({\"question\": \"What is the name of the variant fine-tuned for instruction-following capabilities?\", \"Context\": context})\n",
    "print(answer)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Groundedness Check with LangChain and Upstage\n",
    "![Groundedness](./figures/gc.png)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The name of the variant fine-tuned for instruction-following capabilities is SOLAR 10.7B-Instruct.\n",
      "grounded\n",
      "✅ Groundedness check passed\n"
     ]
    }
   ],
   "source": [
    "# GC\n",
    "from langchain_upstage import UpstageGroundednessCheck\n",
    "\n",
    "groundedness_check = UpstageGroundednessCheck()\n",
    "\n",
    "answer = chain.invoke(\n",
    "    {\n",
    "        \"question\": \"What is the name of the variant fine-tuned for instruction-following capabilities?\",\n",
    "        \"Context\": context,\n",
    "    }\n",
    ")\n",
    "print(answer)\n",
    "gc_result = groundedness_check.invoke({\"context\": context, \"answer\": answer})\n",
    "\n",
    "print(gc_result)\n",
    "if gc_result.lower().startswith(\"grounded\"):\n",
    "    print(\"✅ Groundedness check passed\")\n",
    "else:\n",
    "    print(\"❌ Groundedness check failed\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "notGrounded\n",
      "❌ Groundedness check failed\n"
     ]
    }
   ],
   "source": [
    "# GC\n",
    "from langchain_upstage import UpstageGroundednessCheck\n",
    "\n",
    "groundedness_check = UpstageGroundednessCheck()\n",
    "\n",
    "answer = \"Solar 10.7B is available to the public with a non-commercial license.\"\n",
    "gc_result = groundedness_check.invoke({\"context\": context, \"answer\": answer})\n",
    "\n",
    "print(gc_result)\n",
    "if gc_result.lower().startswith(\"grounded\"):\n",
    "    print(\"✅ Groundedness check passed\")\n",
    "else:\n",
    "    print(\"❌ Groundedness check failed\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Excercise \n",
    "\n",
    "Write code to check GC and make LLM more secure. For example, check GC several times until LLM gives you a reliable answer."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}