week2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 从简单任务开始"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "1955-56, 1956-57, 1957-58, 1958-59, 1959-60, 1965-66, 1997-98, 1999-2000, 2001-02, 2013-14, 2015-16, 2016-17, 2017-18.\n"
     ]
    }
   ],
   "source": [
    "import openai\n",
    "import os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "COMPLETION_MODEL = \"text-davinci-003\"\n",
    "\n",
    "\n",
    "def get_response(prompt):\n",
    "    completions = openai.Completion.create (\n",
    "        engine=COMPLETION_MODEL,\n",
    "        prompt=prompt,\n",
    "        max_tokens=512,\n",
    "        n=1,\n",
    "        stop=None,\n",
    "        temperature=0.0,        \n",
    "    )\n",
    "    message = completions['choices'][0].text # type: ignore\n",
    "    return message\n",
    "\n",
    "# 少了没有数据的21-22\n",
    "print(get_response(\"皇家马德里获得过几次欧洲冠军联赛冠军，分别是哪几年？\")) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "Imagen算法是一种基于深度学习的图像分类算法，它可以自动识别图像中的物体，并将其分类为不同的类别。它使用卷积神经网络（CNN）来提取图像的特征，并使用支持向量机（SVM）来进行分类。\n"
     ]
    }
   ],
   "source": [
    "# 基本上会胡扯\n",
    "print(get_response(\"Imagen算法是什么？\"))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 任务驱动，而不是问答驱动"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{\n",
      "    \"title\": \"Glow-in-the-Dark Inflatable PVC Frog Night Market Hot Selling Water Toy\",\n",
      "    \"selling_points\": [\n",
      "        \"Made of durable PVC material\",\n",
      "        \"Glow-in-the-dark design for night play\",\n",
      "        \"Inflatable design for easy storage and transport\",\n",
      "        \"Perfect for pool, beach, and lake play\",\n",
      "        \"Great for children of all ages\"\n",
      "    ],\n",
      "    \"price_range\": \"$10 - $20\"\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "import openai\n",
    "import os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "COMPLETION_MODEL = \"text-davinci-003\"\n",
    "\n",
    "\n",
    "prompt = \"\"\"\n",
    "Consideration proudct : 工厂现货PVC充气青蛙夜市地摊热卖充气玩具发光蛙儿童水上玩具\n",
    "\n",
    "1. Compose human readale product title used on Amazon in english within 20 words.\n",
    "2. Write 5 selling points for the products in Amazon.\n",
    "3. Evaluate a price range for this product in U.S.\n",
    "\n",
    "Output the result in json format with three properties called title, selling_points and price_range\n",
    "\"\"\"\n",
    "\n",
    "def get_response(prompt):\n",
    "    completions = openai.Completion.create (\n",
    "        engine=COMPLETION_MODEL,\n",
    "        prompt=prompt,\n",
    "        max_tokens=512,\n",
    "        n=1,\n",
    "        stop=None,\n",
    "        temperature=0.0,        \n",
    "    )\n",
    "    message = completions['choices'][0].text # type: ignore\n",
    "    return message\n",
    "\n",
    "print(get_response(prompt)) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 试一下命名实体识别"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{\n",
      "    \"实体\": [\"Man Utd\", \"Ten Hag\", \"League Cup\"]\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "sentence = \"Man Utd must win trophies, says Ten Hag ahead of League Cup final\"\n",
    "prompt = f\"请提取以下句子中的命名实体，并以json形式输出\\n{sentence} \\n\"\n",
    "\n",
    "print(get_response(prompt))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{\n",
      "    \"Organization\": [\"Man Utd\", \"League Cup\"],\n",
      "    \"Person\": [\"Ten Hag\"]\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "prompt = f\"请提取以下句子中的命名实体，并以json形式输出，json的key是实体的类型。\\n{sentence} \\n\"\n",
    "print(get_response(prompt))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "{\n",
      "    \"订单号\": \"202303251200ABC\"\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "order_sentence = \"你好，我有一个订单一直没有收到，订单号是202303251200ABC\"\n",
    "prompt = f\"请提取以下句子中的命名实体，并以json形式输出，json的key是实体的类型。\\n{order_sentence} \\n\"\n",
    "print(get_response(prompt))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 情感分析"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "正面的。\n"
     ]
    }
   ],
   "source": [
    "prompts = \"\"\"判断一下以下用户的评论情感上是正面的还是负面的.\n",
    "\n",
    "买的银色版真的很好看，一天就到了，晚上就开始拿起来完系统很丝滑流畅，做工扎实，手感细腻，很精致哦苹果一如既往的好品质。\"\"\"\n",
    "\n",
    "print(get_response(prompts))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 更好的Prompt来控制结果输出"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "正面\n"
     ]
    }
   ],
   "source": [
    "prompts = \"\"\"判断一下用户的评论情感上是正面的还是负面的.\n",
    "\n",
    "评论：买的银色版真的很好看，一天就到了，晚上就开始拿起来完系统很丝滑流畅，做工扎实，手感细腻，很精致哦苹果一如既往的好品质。\n",
    "情感：\"\"\"\n",
    "\n",
    "print(get_response(prompts))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Few-shot Learning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "正面\n"
     ]
    }
   ],
   "source": [
    "prompts = \"\"\"判断一下用户的评论情感上是正面的还是负面的\n",
    "评论：买的银色版真的很好看，一天就到了，晚上就开始拿起来完系统很丝滑流畅，做工扎实，手感细腻，很精致哦苹果一如既往的好品质\n",
    "情感：正面\n",
    "\n",
    "评论：随意降价，不予价保，服务态度差\n",
    "情感：负面\n",
    "\"\"\"\n",
    "\n",
    "good_case = prompts + \"\"\"\n",
    "评论：外形外观：苹果审美一直很好，金色非常漂亮\n",
    "拍照效果：14pro升级的4800万像素真的是没的说，太好了，\n",
    "运行速度：苹果的反应速度好，用上三五年也不会卡顿的，之前的7P用到现在也不卡\n",
    "其他特色：14pro的磨砂金真的太好看了，不太高调，也不至于没有特点，非常耐看，很好的\n",
    "情感：\n",
    "\"\"\"\n",
    "\n",
    "print(get_response(good_case))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 对于Few-Shot Learning的封装"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prompt:\n",
      "判断一下用户的评论情感上是正面的还是负面的\n",
      "\n",
      "评论： 买的银色版真的很好看，一天就到了，晚上就开始拿起来完系统很丝滑流畅，做工扎实，手感细腻，很精致哦苹果一如既往的好品质\n",
      "情感： 正面\n",
      "\n",
      "评论： 随意降价，不予价保，服务态度差\n",
      "情感： 负面\n",
      "\n",
      "评论： 外形外观：苹果审美一直很好，金色非常漂亮\n",
      "拍照效果：14pro升级的4800万像素真的是没的说，太好了，\n",
      "运行速度：苹果的反应速度好，用上三五年也不会卡顿的，之前的7P用到现在也不卡\n",
      "其他特色：14pro的磨砂金真的太好看了，不太高调，也不至于没有特点，非常耐看，很好的\n",
      "情感： \n",
      "\n",
      "\n",
      "\n",
      "Response:\n",
      " 正面\n"
     ]
    }
   ],
   "source": [
    "class Example:\n",
    "    def __init__(self, question, answer) -> None:\n",
    "        self.question = question\n",
    "        self.answer = answer\n",
    "\n",
    "class PromptTemplate:\n",
    "    def __init__(self, prompt, examples, question_prefix = \"Q: \", answer_prefix = \"A: \") -> None:\n",
    "        self.prompt = prompt\n",
    "        self.examples = examples\n",
    "        self.question_prefix = question_prefix\n",
    "        self.answer_prefix = answer_prefix\n",
    "\n",
    "    def get_prompt(self, question):\n",
    "        result = self.prompt + \"\\n\\n\"\n",
    "        for example in self.examples:\n",
    "            result += f\"{self.question_prefix} {example.question}\\n{self.answer_prefix} {example.answer}\\n\\n\"\n",
    "        result += f\"{self.question_prefix} {question}\\n{self.answer_prefix} \"\n",
    "        return result\n",
    "\n",
    "example1 = Example(\"买的银色版真的很好看，一天就到了，晚上就开始拿起来完系统很丝滑流畅，做工扎实，手感细腻，很精致哦苹果一如既往的好品质\", \"正面\")\n",
    "example2 = Example(\"随意降价，不予价保，服务态度差\", \"负面\")\n",
    "\n",
    "prompt = PromptTemplate(\"判断一下用户的评论情感上是正面的还是负面的\", [example1, example2], \"评论：\", \"情感：\")\n",
    "\n",
    "question = \"\"\"外形外观：苹果审美一直很好，金色非常漂亮\n",
    "拍照效果：14pro升级的4800万像素真的是没的说，太好了，\n",
    "运行速度：苹果的反应速度好，用上三五年也不会卡顿的，之前的7P用到现在也不卡\n",
    "其他特色：14pro的磨砂金真的太好看了，不太高调，也不至于没有特点，非常耐看，很好的\"\"\"\n",
    "\n",
    "print(f\"Prompt:\")\n",
    "print(prompt.get_prompt(question))\n",
    "print(\"\\n\\n\")\n",
    "print(f\"Response:\")\n",
    "print(get_response(prompt.get_prompt(question)))\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Completions接口"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### max_tokens, temperature, top_p, n 以及 stop 参数的演示"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import openai\n",
    "import os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "COMPLETION_MODEL = \"text-davinci-003\"\n",
    "\n",
    "def get_response(prompt, max_tokens=512, temperature=0.0, n=1, stop=None):\n",
    "    completions = openai.Completion.create(\n",
    "        engine=COMPLETION_MODEL,\n",
    "        prompt=prompt,\n",
    "        max_tokens=max_tokens,\n",
    "        n=n,\n",
    "        stop=stop,\n",
    "        temperature=temperature,        \n",
    "    )\n",
    "    return completions\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"choices\": [\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 0,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u7684\\u901a\\u77e5\\u60a8\\uff0c\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\\u7ed3\\u675f\\uff0c\\u611f\\u8c22\\u60a8\\u7684\\u5173\\u6ce8\\uff0c\\u671f\\u5f85\\u60a8\\u7684\\u4e0b\\u6b21\\u5149\\u4e34\\uff01\"\n",
      "    }\n",
      "  ],\n",
      "  \"created\": 1686981724,\n",
      "  \"id\": \"cmpl-7SJIa1Dl2RLaZqd4wGrY3XxSOow5y\",\n",
      "  \"model\": \"text-davinci-003\",\n",
      "  \"object\": \"text_completion\",\n",
      "  \"usage\": {\n",
      "    \"completion_tokens\": 97,\n",
      "    \"prompt_tokens\": 60,\n",
      "    \"total_tokens\": 157\n",
      "  }\n",
      "}\n"
     ]
    }
   ],
   "source": [
    "prompt = \"\"\"请你用淘宝客服的语气告诉用户，这个洗面奶的优惠已经结束了\"\"\"\n",
    "\n",
    "answer = get_response(prompt)\n",
    "print(answer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "您好，非常抱歉的通知您，洗面奶的优惠活动已经结束，感谢您的关注，期待您的下次光临！\n"
     ]
    }
   ],
   "source": [
    "print(answer.choices[0].text) # type: ignore"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"choices\": [\n",
      "    {\n",
      "      \"finish_reason\": \"length\",\n",
      "      \"index\": 0,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\"\n",
      "    }\n",
      "  ],\n",
      "  \"created\": 1686981791,\n",
      "  \"id\": \"cmpl-7SJJfNCGqDk4KNqJ5ecWYWq1vzax0\",\n",
      "  \"model\": \"text-davinci-003\",\n",
      "  \"object\": \"text_completion\",\n",
      "  \"usage\": {\n",
      "    \"completion_tokens\": 10,\n",
      "    \"prompt_tokens\": 60,\n",
      "    \"total_tokens\": 70\n",
      "  }\n",
      "}\n",
      "\n",
      "\n",
      "您好，\n"
     ]
    }
   ],
   "source": [
    "answer = get_response(prompt, max_tokens=10)\n",
    "print(answer)\n",
    "print(answer.choices[0].text) # type: ignore"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"choices\": [\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 0,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u7ed9\\u60a8\\u5e26\\u6765\\u4e0d\\u4fbf\\uff0c\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\\u7ed3\\u675f\\uff0c\\u611f\\u8c22\\u60a8\\u7684\\u5173\\u6ce8\\uff0c\\u671f\\u5f85\\u60a8\\u7684\\u4e0b\\u6b21\\u5149\\u4e34\\uff01\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 1,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u7ed9\\u60a8\\u5e26\\u6765\\u4e0d\\u4fbf\\uff0c\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\\u7ed3\\u675f\\uff0c\\u611f\\u8c22\\u60a8\\u7684\\u5173\\u6ce8\\uff0c\\u671f\\u5f85\\u60a8\\u7684\\u4e0b\\u6b21\\u5149\\u4e34\\uff01\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 2,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u7ed9\\u60a8\\u5e26\\u6765\\u4e0d\\u4fbf\\uff0c\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\\u7ed3\\u675f\\uff0c\\u611f\\u8c22\\u60a8\\u7684\\u5173\\u6ce8\\uff0c\\u671f\\u5f85\\u60a8\\u7684\\u4e0b\\u6b21\\u5149\\u4e34\\uff01\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 3,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u7ed9\\u60a8\\u5e26\\u6765\\u4e0d\\u4fbf\\uff0c\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\\u7ed3\\u675f\\uff0c\\u611f\\u8c22\\u60a8\\u7684\\u5173\\u6ce8\\uff0c\\u671f\\u5f85\\u60a8\\u7684\\u4e0b\\u6b21\\u5149\\u4e34\\uff01\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 4,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u7ed9\\u60a8\\u5e26\\u6765\\u4e0d\\u4fbf\\uff0c\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\\u7ed3\\u675f\\uff0c\\u611f\\u8c22\\u60a8\\u7684\\u5173\\u6ce8\\uff0c\\u6211\\u4eec\\u4f1a\\u7ee7\\u7eed\\u63d0\\u4f9b\\u66f4\\u591a\\u4f18\\u60e0\\u6d3b\\u52a8\\uff0c\\u656c\\u8bf7\\u671f\\u5f85\\uff01\"\n",
      "    }\n",
      "  ],\n",
      "  \"created\": 1686981879,\n",
      "  \"id\": \"cmpl-7SJL5N94g3dfCIyVlmTzqvydDO31t\",\n",
      "  \"model\": \"text-davinci-003\",\n",
      "  \"object\": \"text_completion\",\n",
      "  \"usage\": {\n",
      "    \"completion_tokens\": 537,\n",
      "    \"prompt_tokens\": 60,\n",
      "    \"total_tokens\": 597\n",
      "  }\n",
      "}\n",
      "Completions:\n",
      "\n",
      "\n",
      "您好，非常抱歉给您带来不便，洗面奶的优惠活动已经结束，感谢您的关注，期待您的下次光临！\n",
      "\n",
      "--\n",
      "\n",
      "\n",
      "\n",
      "您好，非常抱歉给您带来不便，洗面奶的优惠活动已经结束，感谢您的关注，期待您的下次光临！\n",
      "\n",
      "--\n",
      "\n",
      "\n",
      "\n",
      "您好，非常抱歉给您带来不便，洗面奶的优惠活动已经结束，感谢您的关注，期待您的下次光临！\n",
      "\n",
      "--\n",
      "\n",
      "\n",
      "\n",
      "您好，非常抱歉给您带来不便，洗面奶的优惠活动已经结束，感谢您的关注，期待您的下次光临！\n",
      "\n",
      "--\n",
      "\n",
      "\n",
      "\n",
      "您好，非常抱歉给您带来不便，洗面奶的优惠活动已经结束，感谢您的关注，我们会继续提供更多优惠活动，敬请期待！\n",
      "\n",
      "--\n",
      "\n"
     ]
    }
   ],
   "source": [
    "answer = get_response(prompt, temperature=0, n=5)\n",
    "print(answer)\n",
    "print(\"Completions:\")\n",
    "for choice in answer.choices: # type: ignore\n",
    "    print(choice.text) # type: ignore\n",
    "    print(\"\\n--\\n\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"choices\": [\n",
      "    {\n",
      "      \"finish_reason\": \"length\",\n",
      "      \"index\": 0,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\u5427?\\n\\n\\u4f60\\u597d\\uff01\\u5f88Not sorry\\u7ed9man disrupt\\u4e4bpath stand curvebash at awkwardist weak\\u8ba9ko\\u5957step sidewalk\\u673abloctojledriver \\u5e97\\u91ccskin sink feetfracci barrier flane\\u727bon\\u7736fish whiling shine\\u96fbork fuwage\\u5e00hing shame checktoe line bounce9cycle pat impfly\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"length\",\n",
      "      \"index\": 1,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u5c0a\\u656c\\u7684\\u987e\\u5ba2\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u7684\\u901a\\u77e5\\u60a8\\u4eec\\uff0c\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"length\",\n",
      "      \"index\": 2,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"~\\u9053\\u6b49\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"length\",\n",
      "      \"index\": 3,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\\"\\\"\\\"\\n\\u60a8\\u597d\\uff01\\u5f88\\u9057\\u61be\\u5730\\u5ba3\\u5e03\\u8be5\\u5468\\u671f\\u4f18\\u60e0\\u5df2\\u63a5\\u8fd1\\u7ed3\\u675f\\uff01\\u4f46-Zsm\\u7531\\u4e8e-Xyrise Zdel And.-Aisuau Mwhein Gel X\\u00aa Ao utev[ea I Alience Haittoday Gre? Psery Hair2,.\\u3002\\u5982\\u518d\\u0434RFOD LEtu ]UhdithkiastichraMort\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"length\",\n",
      "      \"index\": 4,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u5c0a\\u656c\\u7684\\u7528\\u6237\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u5e26\\u7ed9\\u60a8\\u7126\\u5934\"\n",
      "    }\n",
      "  ],\n",
      "  \"created\": 1686981944,\n",
      "  \"id\": \"cmpl-7SJM8Ev6go90xBxOlQnLwzWh0DAO4\",\n",
      "  \"model\": \"text-davinci-003\",\n",
      "  \"object\": \"text_completion\",\n",
      "  \"usage\": {\n",
      "    \"completion_tokens\": 278,\n",
      "    \"prompt_tokens\": 60,\n",
      "    \"total_tokens\": 338\n",
      "  }\n",
      "}\n",
      "Completions:\n",
      "吧?\n",
      "\n",
      "你好！很Not sorry给man disrupt之path stand curvebash at awkwardist weak让ko套step sidewalk机bloctojledriver 店里skin sink feetfracci barrier flane牻on眶fish whiling shine電ork fuwage帀hing shame checktoe line bounce9cycle pat impfly\n",
      "\n",
      "--\n",
      "\n",
      "\n",
      "\n",
      "尊敬的顾客，非常抱歉的通知您们，\n",
      "\n",
      "--\n",
      "\n",
      "~道歉\n",
      "\n",
      "--\n",
      "\n",
      "\n",
      "\n",
      "\"\"\"\n",
      "您好！很遗憾地宣布该周期优惠已接近结束！但-Zsm由于-Xyrise Zdel And.-Aisuau Mwhein Gel Xª Ao utev[ea I Alience Haittoday Gre? Psery Hair2,.。如再дRFOD LEtu ]UhdithkiastichraMort\n",
      "\n",
      "--\n",
      "\n",
      "\n",
      "\n",
      "尊敬的用户，非常抱歉带给您焦头\n",
      "\n",
      "--\n",
      "\n"
     ]
    }
   ],
   "source": [
    "answer = get_response(prompt, temperature=2.0, n=5)\n",
    "print(answer)\n",
    "print(\"Completions:\")\n",
    "for choice in answer.choices: # type: ignore\n",
    "    print(choice.text) # type: ignore\n",
    "    print(\"\\n--\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"choices\": [\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 0,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u5c0a\\u656c\\u7684\\u7528\\u6237\\uff0c\\u5f88\\u62b1\\u6b49\\u7684\\u901a\\u77e5\\u60a8\\uff0c\\u6b64\\u6b21\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\\u7ed3\\u675f\\u3002\\u611f\\u8c22\\u60a8\\u4e00\\u76f4\\u4ee5\\u6765\\u5bf9\\u6dd8\\u5b9d\\u7684\\u652f\\u6301\\uff0c\\u5e0c\\u671b\\u4ee5\\u540e\\u8fd8\\u6709\\u673a\\u4f1a\\u80fd\\u591f\\u63d0\\u4f9b\\u66f4\\u591a\\u4f18\\u60e0\\u6d3b\\u52a8\\u7ed9\\u60a8\\u53c2\\u52a0\\u3002\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 1,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u611f\\u8c22\\u60a8\\u8fd9\\u6b21\\u5bf9\\u6211\\u4eec\\u6d17\\u9762\\u5976\\u7684\\u5173\\u6ce8\\uff01\\u5f88\\u62b1\\u6b49\\u544a\\u8bc9\\u60a8\\uff0c\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u5df2\\u7ecf\\u7ed3\\u675f\\uff0c\\u4e3a\\u4e86\\u4e0d\\u8ba9\\u60a8\\u5931\\u671b\\uff0c\\u6211\\u4eec\\u4ecd\\u7136\\u4f1a\\u4e3a\\u60a8\\u63d0\\u4f9b\\u4f18\\u60e0\\u548c\\u4fc3\\u9500\\u6d3b\\u52a8\\u3002\\u671f\\u5f85\\u60a8\\u7684\\u518d\\u6b21\\u5149\\u4e34\\uff01\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 2,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\uff0c\\u60a8\\u60f3\\u8981\\u8d2d\\u4e70\\u7684\\u6d17\\u9762\\u5976\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\\u7ed3\\u675f\\u3002\\u611f\\u8c22\\u60a8\\u7684\\u5173\\u6ce8\\uff0c\\u795d\\u60a8\\u8d2d\\u7269\\u6109\\u5feb\\uff01\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 3,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u7684\\u901a\\u77e5\\u60a8\\uff0c\\u8be5\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\\u7ed3\\u675f\\uff0c\\u6211\\u4eec\\u91cd\\u65b0\\u4e3a\\u60a8\\u63d0\\u4f9b\\u66f4\\u591a\\u4f18\\u60e0\\uff0c\\u656c\\u8bf7\\u5173\\u6ce8\\uff01\\u5982\\u679c\\u60a8\\u6709\\u4efb\\u4f55\\u7591\\u95ee\\uff0c\\u6b22\\u8fce\\u62e8\\u6253\\u6211\\u4eec\\u7684\\u5ba2\\u670d\\u7535\\u8bdd\\u6216\\u53d1\\u9001\\u90ae\\u4ef6\\u67e5\\u8be2\\u3002\\u795d\\u60a8\\u8d2d\\u7269\\u6109\\u5feb\\uff01\"\n",
      "    },\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 4,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\uff1f\\n\\n\\u5c0a\\u656c\\u7684\\u7528\\u6237\\uff0c\\u5f88\\u62b1\\u6b49\\u7684\\u901a\\u77e5\\u60a8\\uff0c\\u6d17\\u9762\\u5976\\u5546\\u54c1\\u7684\\u4f18\\u60e0\\u5df2\\u7ecf\\u7ed3\\u675f\\uff0c\\u611f\\u8c22\\u60a8\\u5bf9\\u6dd8\\u5b9d\\u7684\\u652f\\u6301\\uff0c\\u6211\\u4eec\\u5c06\\u4f1a\\u4e3a\\u60a8\\u63d0\\u4f9b\\u66f4\\u591a\\u66f4\\u4f18\\u60e0\\u7684\\u5546\\u54c1\\u3002\\u8c22\\u8c22\\u7406\\u89e3\\uff01\"\n",
      "    }\n",
      "  ],\n",
      "  \"created\": 1686981996,\n",
      "  \"id\": \"cmpl-7SJMytJyFU86Ca2GI7jrKEmAJZyO7\",\n",
      "  \"model\": \"text-davinci-003\",\n",
      "  \"object\": \"text_completion\",\n",
      "  \"usage\": {\n",
      "    \"completion_tokens\": 743,\n",
      "    \"prompt_tokens\": 60,\n",
      "    \"total_tokens\": 803\n",
      "  }\n",
      "}\n",
      "Completions:\n",
      "\n",
      "\n",
      "尊敬的用户，很抱歉的通知您，此次洗面奶的优惠活动已经结束。感谢您一直以来对淘宝的支持，希望以后还有机会能够提供更多优惠活动给您参加。\n",
      "\n",
      "--\n",
      "\n",
      "\n",
      "\n",
      "您好，非常感谢您这次对我们洗面奶的关注！很抱歉告诉您，洗面奶的优惠已经结束，为了不让您失望，我们仍然会为您提供优惠和促销活动。期待您的再次光临！\n",
      "\n",
      "--\n",
      "\n",
      "\n",
      "\n",
      "您好，非常抱歉，您想要购买的洗面奶优惠活动已经结束。感谢您的关注，祝您购物愉快！\n",
      "\n",
      "--\n",
      "\n",
      "\n",
      "\n",
      "您好，非常抱歉的通知您，该洗面奶的优惠活动已经结束，我们重新为您提供更多优惠，敬请关注！如果您有任何疑问，欢迎拨打我们的客服电话或发送邮件查询。祝您购物愉快！\n",
      "\n",
      "--\n",
      "\n",
      "？\n",
      "\n",
      "尊敬的用户，很抱歉的通知您，洗面奶商品的优惠已经结束，感谢您对淘宝的支持，我们将会为您提供更多更优惠的商品。谢谢理解！\n",
      "\n",
      "--\n",
      "\n"
     ]
    }
   ],
   "source": [
    "answer = get_response(prompt, temperature=1.0, n=5)\n",
    "print(answer)\n",
    "print(\"Completions:\")\n",
    "for choice in answer.choices: # type: ignore\n",
    "    print(choice.text) # type: ignore\n",
    "    print(\"\\n--\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"choices\": [\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 0,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u5730\\u901a\\u77e5\\u60a8\\uff0c\\u76ee\\u524d\\u8be5\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\"\n",
      "    }\n",
      "  ],\n",
      "  \"created\": 1686982054,\n",
      "  \"id\": \"cmpl-7SJNuOfYC1wcbPy9Qi0AmEhAbrFcT\",\n",
      "  \"model\": \"text-davinci-003\",\n",
      "  \"object\": \"text_completion\",\n",
      "  \"usage\": {\n",
      "    \"completion_tokens\": 61,\n",
      "    \"prompt_tokens\": 60,\n",
      "    \"total_tokens\": 121\n",
      "  }\n",
      "}\n",
      "\n",
      "\n",
      "您好，非常抱歉地通知您，目前该洗面奶的优惠活动已经\n"
     ]
    }
   ],
   "source": [
    "answer = get_response(prompt, temperature=0.5, n=1, stop=\"结束\")\n",
    "print(answer)\n",
    "print(answer.choices[0].text) # type: ignore"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### echo, stream, logit_bias 参数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "import openai\n",
    "import os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "COMPLETION_MODEL = \"text-davinci-003\"\n",
    "\n",
    "def get_response(prompt, echo=False, stream=False, logit_bias={}):\n",
    "    completions = openai.Completion.create(\n",
    "        engine=COMPLETION_MODEL,\n",
    "        prompt=prompt,\n",
    "        max_tokens=512,\n",
    "        n=1,\n",
    "        stop=None,\n",
    "        temperature=0.0,        \n",
    "        echo=echo,\n",
    "        stream=stream,\n",
    "        logit_bias=logit_bias,\n",
    "    )\n",
    "    return completions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"choices\": [\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 0,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\u8bf7\\u4f60\\u7528\\u6dd8\\u5b9d\\u5ba2\\u670d\\u7684\\u8bed\\u6c14\\u544a\\u8bc9\\u7528\\u6237\\uff0c\\u8fd9\\u4e2a\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u5df2\\u7ecf\\u7ed3\\u675f\\u4e86\\n\\n\\u60a8\\u597d\\uff0c\\u975e\\u5e38\\u62b1\\u6b49\\u7ed9\\u60a8\\u5e26\\u6765\\u4e0d\\u4fbf\\uff0c\\u6d17\\u9762\\u5976\\u7684\\u4f18\\u60e0\\u6d3b\\u52a8\\u5df2\\u7ecf\\u7ed3\\u675f\\uff0c\\u611f\\u8c22\\u60a8\\u7684\\u5173\\u6ce8\\uff0c\\u6211\\u4eec\\u4f1a\\u7ee7\\u7eed\\u63d0\\u4f9b\\u66f4\\u591a\\u4f18\\u60e0\\u6d3b\\u52a8\\uff0c\\u656c\\u8bf7\\u671f\\u5f85\\uff01\"\n",
      "    }\n",
      "  ],\n",
      "  \"created\": 1686982286,\n",
      "  \"id\": \"cmpl-7SJReez125m3vVASgYXlpeXcyhs3i\",\n",
      "  \"model\": \"text-davinci-003\",\n",
      "  \"object\": \"text_completion\",\n",
      "  \"usage\": {\n",
      "    \"completion_tokens\": 129,\n",
      "    \"prompt_tokens\": 60,\n",
      "    \"total_tokens\": 189\n",
      "  }\n",
      "}\n",
      "请你用淘宝客服的语气告诉用户，这个洗面奶的优惠已经结束了\n",
      "\n",
      "您好，非常抱歉给您带来不便，洗面奶的优惠活动已经结束，感谢您的关注，我们会继续提供更多优惠活动，敬请期待！\n"
     ]
    }
   ],
   "source": [
    "answer = get_response(prompt, echo=True)\n",
    "print(answer)\n",
    "print(answer.choices[0].text) # type: ignore"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "您好，非常抱歉的通知您，洗面奶的优惠活动已经结束，感谢您的关注，期待您的下次光临！"
     ]
    }
   ],
   "source": [
    "answer = get_response(prompt, stream=True)\n",
    "\n",
    "for event in answer:\n",
    "    event_text = event.choices[0].text # type: ignore\n",
    "    print(event_text, end = '')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[162, 224, 101]\n",
      "{\n",
      "  \"choices\": [\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 0,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\n\\u5c0a\\u656c\\u7684\\u7535\\u5546\\u751c\\u5fc3\\uff0c\\u975e\\u5e38\\u611f\\u8c22\\u4f60\\u5bf9\\u6211\\u4eec\\u7684\\u5173\\u6cdb\\uff01\\u5f88\\u9057\\u5fc3\\u5730\\u901a\\u77e5\\u4f60\\uff0c\\u8fd9\\u4e2a\\u7279\\u4ef7\\u7684\\u4f18\\u60e0\\u5df2\\u7ecf\\u7ed3\\u675f\\u4e86\\uff0c\\u4f46\\u662f\\u6211\\u4eec\\u8fd8\\u6709\\u5176\\u4ed6\\u4f18\\u60e0\\u53ef\\u4ee5\\u4e3a\\u4f60\\u5e26\\u6765\\u60ca\\u559c\\uff0c\\u8bf7\\u7ee7\\u7eed\\u5173\\u6cdb\\u6211\\u4eec\\uff0c\\u8c22\\u8c22\\uff01\"\n",
      "    }\n",
      "  ],\n",
      "  \"created\": 1686982490,\n",
      "  \"id\": \"cmpl-7SJUw51znxBAshPz0pGGVS6kVcDZE\",\n",
      "  \"model\": \"text-davinci-003\",\n",
      "  \"object\": \"text_completion\",\n",
      "  \"usage\": {\n",
      "    \"completion_tokens\": 158,\n",
      "    \"prompt_tokens\": 60,\n",
      "    \"total_tokens\": 218\n",
      "  }\n",
      "}\n",
      "\n",
      "\n",
      "尊敬的电商甜心，非常感谢你对我们的关泛！很遗心地通知你，这个特价的优惠已经结束了，但是我们还有其他优惠可以为你带来惊喜，请继续关泛我们，谢谢！\n"
     ]
    }
   ],
   "source": [
    "import tiktoken\n",
    "encoding = tiktoken.get_encoding('p50k_base')\n",
    "token_ids = encoding.encode(\"您\")\n",
    "print(token_ids)\n",
    "\n",
    "\n",
    "bias_map = {}\n",
    "for token_id in token_ids:\n",
    "    bias_map[token_id] = -100\n",
    "\n",
    "\n",
    "answer = get_response(prompt, logit_bias=bias_map)\n",
    "print(answer)\n",
    "print(answer.choices[0].text) # type: ignore\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Embedding接口"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "import openai\n",
    "import os\n",
    "from openai.embeddings_utils import cosine_similarity, get_embedding\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "EMBEDDING_MODEL = \"text-embedding-ada-002\"\n",
    "\n",
    "positive_review = get_embedding(\"这是一条好评\", engine=EMBEDDING_MODEL)\n",
    "negative_review = get_embedding(\"这是一条差评\", engine=EMBEDDING_MODEL)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[-0.006549043580889702, -0.0008848072029650211, 0.013451029546558857, -0.021006613969802856, 0.001802294165827334, -0.0037549156695604324, -0.019045822322368622, -0.01912425272166729, -0.012026187963783741, -0.0198432095348835, 0.012647105380892754, 0.015006591565907001, -0.012235338799655437, -0.0023872635792940855, -0.010117684490978718, 0.009614414535462856, 0.047974031418561935, 0.000568221032153815, 0.009418335743248463, -0.008653626777231693, -0.008607874624431133, 0.0003970602701883763, -0.01377782877534628, -0.01796085014939308, -0.01696738228201866, 0.015333389863371849, 0.028339972719550133, -0.02633996680378914, 0.006663423031568527, -0.00035641470458358526, 0.017437972128391266, -0.016039274632930756, -0.0261569581925869, -0.009529447183012962, 0.004918318707495928, 0.007385647855699062, -0.0257125124335289, 0.010137292556464672, 0.029150433838367462, 0.015320317819714546, 0.027764808386564255, -0.002232034457847476, 0.020967397838830948, 0.022313807159662247, -0.014000051654875278, -0.02057523839175701, 0.011111152358353138, -0.012575210072100163, -0.00807846151292324, 0.018732095137238503, 0.016797447577118874, 0.005973878316581249, -0.0383269377052784, -0.008091532625257969, -0.0034085093066096306, 0.015620972029864788, 0.0018235361203551292, 0.0018464119639247656, 0.0030702727381139994, -0.019241901114583015, 0.018235361203551292, 0.006604599300771952, -0.029882462695240974, 0.011627493426203728, -0.012568674050271511, 0.005330084823071957, -0.013313774019479752, -0.005751654971390963, 0.01796085014939308, 0.012464097701013088, 0.006241852883249521, 0.007718982640653849, 0.0015049075009301305, 0.002535957144573331, 0.043869443237781525, -0.008810490369796753, -0.021398771554231644, -0.008803954347968102, 0.015189598314464092, -0.00893467292189598, 0.006875842344015837, -0.0360785648226738, -0.0028464156202971935, 0.020196152850985527, 0.022039296105504036, 0.006607867311686277, -0.023568714037537575, 0.03704588860273361, -0.02665369212627411, -0.004875835031270981, 0.00868630688637495, 0.020849749445915222, 0.009143824689090252, -0.014013123698532581, -0.004947730805724859, 0.029255010187625885, 0.0027745200786739588, 0.04347728565335274, 0.010398730635643005, -0.02206544019281864, 0.015673261135816574, 0.0029019713401794434, -0.012202659621834755, -0.007163424976170063, -0.04177793115377426, -0.007686302997171879, 0.0051176659762859344, -0.013261486776173115, 0.03586941212415695, -0.006823554635047913, -0.035712551325559616, -0.005101325921714306, -0.006117669399827719, -0.03824850544333458, -0.014470641501247883, 0.023529497906565666, 0.008673234842717648, -0.030013183131814003, -0.010379123501479626, -0.022536030039191246, 0.02679748460650444, 0.0347452275454998, -0.002299027983099222, -0.014444497413933277, 0.0024934732355177402, 0.022353023290634155, -0.012490241788327694, 0.0007598067168146372, 0.0053856405429542065, 0.024026231840252876, 0.02508505992591381, 0.01317651942372322, 0.020405303686857224, 0.010261476039886475, -0.0236863624304533, 0.02896742708981037, -0.022457599639892578, 0.022849757224321365, -0.005297405179589987, -0.01647064834833145, 0.0038333474658429623, 0.028653699904680252, -0.023673290386795998, -0.019111180678009987, 0.0006458357092924416, 0.021006613969802856, -0.010549058206379414, 0.017346469685435295, -0.010052324272692204, -0.0076013351790606976, 0.028758276253938675, -0.03108508139848709, 0.017673267051577568, 0.029804030433297157, 0.009496767073869705, 0.005666687618941069, 0.02048373594880104, 0.02584323287010193, -0.011777820996940136, 0.0008390553994104266, 0.005888910498470068, 0.0040457663126289845, 0.00471243541687727, -0.021006613969802856, 0.008542515337467194, 0.021594850346446037, 0.0241308081895113, 0.011163439601659775, 0.010300691239535809, -0.0008361958898603916, 0.01583012379705906, 0.01746411621570587, -0.005725511349737644, 0.009039249271154404, -0.013307238928973675, 0.011732068844139576, 0.005421588663011789, -0.009947748854756355, 0.0014060508692637086, -0.0028071999549865723, -0.030509915202856064, 0.012614425271749496, 0.01328763086348772, 0.015490253455936909, -0.01103272009640932, -0.0021879165433347225, 0.010490234941244125, -0.011921612545847893, -0.006415056064724922, -0.017032742500305176, 0.012143835425376892, 0.029307296499609947, -0.014941231347620487, -0.026444541290402412, -0.6316363215446472, -0.021032756194472313, -0.005205901339650154, 0.008679770864546299, 0.041255053132772446, 0.01794777810573578, 0.02083667740225792, 0.0032140640541911125, 0.0008374213939532638, -0.07518982142210007, -0.021281123161315918, -0.0055490401573479176, 0.006549043580889702, -0.008457547053694725, -0.009568662382662296, -0.022719038650393486, -0.002851317636668682, -0.014902015216648579, 0.0008243494085036218, 0.0047908672131598, -0.02205236814916134, 0.029882462695240974, -0.0026650424115359783, -0.0022957599721848965, 0.004872567020356655, 0.003911779262125492, 0.007071921601891518, -0.011947756633162498, -0.01019611582159996, 0.013098087161779404, -0.02873213216662407, 0.016000058501958847, -0.008993497118353844, -0.011993507854640484, 0.06373879313468933, 0.0005012273322790861, 0.008745130151510239, 0.010326835326850414, 0.021437987685203552, 0.03265371546149254, -0.013607893139123917, -0.00990199763327837, 0.002632362535223365, -0.014013123698532581, 0.0007124209078028798, -0.019372619688510895, 0.0374380461871624, 0.0033529535867273808, -0.0023447799030691385, 0.015085022896528244, -0.016719015315175056, 0.01691509410738945, 0.0038692953530699015, -0.01103272009640932, -0.0031585083343088627, -0.01989549770951271, 0.034536074846982956, -0.022392239421606064, -0.01156867016106844, 0.008535979315638542, 0.007006561849266291, 0.01650986447930336, -0.016862807795405388, -0.007947742007672787, -0.0193987637758255, 0.018797455355525017, -0.018483728170394897, 0.008183036930859089, 0.025503361597657204, -0.0018464119639247656, -0.0015833391807973385, 0.008503299206495285, 0.011228799819946289, 0.012313771061599255, -0.0015661822399124503, 0.024549109861254692, 0.011542526073753834, -0.01430070586502552, -0.0009085000492632389, 0.004839886911213398, 0.0015106265200302005, -0.010437946766614914, 0.0077516622841358185, -0.006993489805608988, 0.011196119710803032, 0.013019655831158161, -0.03824850544333458, 0.0021258248016238213, -0.007372576277703047, -0.0021781125105917454, -0.006016361992806196, -0.013333382084965706, 0.012535993941128254, 0.014339921995997429, 0.00875166617333889, 0.012261482886970043, -0.024196166545152664, -0.0037385758478194475, 0.007745126727968454, -0.021320339292287827, -0.004450996872037649, -0.010732065886259079, 0.02074517495930195, 0.0039379228837788105, -0.01660136878490448, 0.001789222238585353, -0.00764708686619997, -0.000440361094661057, 0.023568714037537575, -0.0239085853099823, 0.013882404193282127, 0.0051928297616541386, -0.008875849656760693, -0.02219616062939167, -0.010522914119064808, -0.03171253576874733, -0.023921657353639603, 0.024705972522497177, 0.0077320546843111515, -0.0034248491283506155, 0.0030522989109158516, 0.0001432807621313259, 0.02878442034125328, 0.015150382183492184, 0.008339899592101574, 0.013411814346909523, -0.005297405179589987, -0.011176511645317078, -0.008398723788559437, 0.00011948573228437454, -0.004941194783896208, 0.02151641808450222, 0.033725615590810776, -0.015843195840716362, 0.024967411532998085, -0.02026151306927204, -0.01890202984213829, -0.010830105282366276, 0.03160795941948891, -0.014039266854524612, -0.017254965379834175, 0.0042516496032476425, 0.012117691338062286, -0.01573861949145794, -0.00830068439245224, -0.005944466218352318, 0.007071921601891518, -0.0031617763452231884, -0.028496837243437767, -0.005647079553455114, -0.0018611178966239095, -0.005035966169089079, -0.011111152358353138, 0.017398755997419357, 0.001962425420060754, 0.01691509410738945, -0.002220596419647336, -0.017124246805906296, -0.02521577849984169, -0.0038594913203269243, -0.0076666949316859245, 0.015947770327329636, -0.020875893533229828, 0.00617976114153862, -0.016902022063732147, -0.019228829070925713, 0.01312423124909401, -0.004692827817052603, -0.024967411532998085, -0.03511124104261398, -0.005359496921300888, 0.010294155217707157, 0.00021098933939356357, 0.004196093883365393, 0.0010857883607968688, -0.006375840399414301, -0.025921663269400597, -0.01141180656850338, -0.008823561482131481, -0.005575183779001236, 0.008464083075523376, -0.009098072536289692, -0.022261518985033035, 0.005081717856228352, 0.013751684688031673, -0.002955893287435174, 0.01700659841299057, 0.036156997084617615, -0.004330081399530172, 0.015555612742900848, 0.0003029014333151281, 0.04217008873820305, -0.016888951882719994, -0.014156915247440338, -0.01709810271859169, 0.011712461709976196, -0.004839886911213398, 0.010333371348679066, 0.01673208735883236, 0.007777806371450424, 0.026052383705973625, 0.01583012379705906, 0.015215742401778698, 0.005241849459707737, 0.0005236947326920927, -0.01195429265499115, 0.011124223470687866, -0.007529439404606819, 0.01813078485429287, 0.01114383153617382, -0.004584984388202429, -0.022522957995533943, -0.007333360146731138, -0.006323552690446377, -0.007300680503249168, 0.021189620718359947, -0.017647122964262962, 0.016575224697589874, 0.012300699017941952, -0.001464874716475606, -0.018509872257709503, 0.007712446618825197, 0.000630721275229007, -0.006251656915992498, -0.002101314952597022, -0.00841179583221674, 0.017085030674934387, 0.008274540305137634, 0.024653686210513115, -0.008052317425608635, -0.020640598610043526, 0.009575198404490948, 0.0014142208965495229, 0.008124212734401226, -0.01258174516260624, -0.009535983204841614, 0.0410459041595459, -0.022627534344792366, 0.046300824731588364, 0.012522921897470951, 0.011248406954109669, 0.01285625621676445, 0.004990214481949806, -0.022928189486265182, -0.0005535150994546711, 0.004366029053926468, 0.030771354213356972, 0.01606541872024536, -0.002602950669825077, 0.014980447478592396, -0.018562160432338715, 0.011183047667145729, -0.014026195742189884, 0.003689556149765849, 0.005522896070033312, -0.0333857424557209, -0.008856241591274738, 0.0180392824113369, 0.01579090766608715, 0.026470685377717018, 0.010634025558829308, 0.018483728170394897, 0.012287626974284649, 0.019607914611697197, -0.0030604687053710222, 0.0005498386453837156, -0.004539232235401869, -0.023333419114351273, -0.0027565460186451674, -0.010202651843428612, 0.017437972128391266, -0.016588296741247177, 0.006022898014634848, -0.0018692879239097238, 0.006692835129797459, -0.010156899690628052, 0.007464079651981592, -0.00022304004232864827, -0.02115040458738804, -0.0035686406772583723, -0.028209254145622253, -0.017163461074233055, 0.015268029645085335, 0.02107197232544422, -0.025228850543498993, -0.025424929335713387, -0.02941187284886837, -0.018117712810635567, -0.0239085853099823, 0.018954318016767502, 0.012032723985612392, -0.0026764804497361183, -0.01430070586502552, -0.007901989854872227, -0.0031846521887928247, -0.00045915201189927757, 0.047346580773591995, -0.004398709163069725, -0.029568735510110855, -0.009339903481304646, 0.01368632446974516, -0.00859480258077383, 0.021529490128159523, -0.04282368719577789, 0.004245113581418991, -0.0374380461871624, -0.00881702546030283, 0.0047157034277915955, -0.002602950669825077, 0.003179750172421336, 0.010392194613814354, -0.003165044356137514, -0.0052124373614788055, -0.037228893488645554, 0.018013138324022293, 0.005794139113277197, 0.00512093398720026, -0.01439221017062664, 0.02891513891518116, -0.004555572289973497, -0.005173221696168184, 0.009549054317176342, -0.026758268475532532, -0.0018839938566088676, 0.044732190668582916, 0.010666705667972565, 0.007189569063484669, 0.01669287122786045, 0.014941231347620487, -0.0009052320965565741, -0.03986942768096924, -0.009078464470803738, 0.027686376124620438, -0.01362096518278122, -0.0024787671864032745, 0.007503295782953501, -0.006830090656876564, 0.013751684688031673, 0.02786938287317753, 0.005604595877230167, 0.006127473432570696, -0.006310480646789074, 0.01296736765652895, -0.009601342491805553, 0.002338243881240487, 0.009483695030212402, 0.03312430530786514, -0.0018709219293668866, -0.001209154725074768, -0.006042506080120802, 0.014013123698532581, 0.015777835622429848, 0.0011094812070950866, -0.007941205985844135, -0.00031822011806070805, -0.014313777908682823, -0.012058868072926998, -0.01579090766608715, -0.014915087260305882, 0.030745211988687515, -0.004905246663838625, -0.003218966070562601, -0.00027246831450611353, 0.011320303194224834, 0.029124289751052856, 0.0261569581925869, 0.020287657156586647, 0.001504090498201549, 0.015712475404143333, 0.0024346495047211647, 0.004888907074928284, 0.02648375742137432, -0.030379196628928185, 0.004333349410444498, 0.0022973939776420593, 0.0005428941803984344, -0.037516478449106216, -0.011673245579004288, -0.006009825970977545, 0.004947730805724859, -0.035294245928525925, 0.005225509405136108, -0.0006478782161138952, -0.008928137831389904, 0.002433015499264002, -0.03464065119624138, -0.025451073423027992, -0.007784342393279076, -0.02471904456615448, -0.017882417887449265, -0.011601350270211697, -0.03252299502491951, 0.0013047433458268642, -0.013294166885316372, -0.0006192832952365279, 0.019830137491226196, -0.011581742204725742, -0.04938580095767975, 0.02797395922243595, 0.006653619464486837, 0.012973903678357601, -0.014326849952340126, 0.0026503365952521563, 0.05218319967389107, -0.002735304180532694, -0.012117691338062286, -0.01152291800826788, -0.012803968973457813, 0.020875893533229828, 0.02941187284886837, -0.006379108410328627, 0.006516363937407732, -0.00041033647721633315, -0.0038170074112713337, 0.04313741251826286, 0.00031311390921473503, 0.014209202490746975, 0.02183014526963234, -0.01934647560119629, -0.008444475941359997, 0.00464707612991333, 0.02300661988556385, 0.0040261587128043175, -0.01772555522620678, 0.010202651843428612, -0.015529468655586243, -0.0017663463950157166, 0.0015751691535115242, 0.018431439995765686, -0.01130069512873888, -0.002653604606166482, 0.004830083344131708, 0.021320339292287827, 0.0032124300487339497, 0.03129423409700394, -0.020692886784672737, 0.004218969959765673, 0.010333371348679066, 0.004375833086669445, 0.017071958631277084, 0.00958827044814825, 0.011738604865968227, -0.02841840498149395, -0.02805239148437977, 0.011888932436704636, -0.013973907567560673, 0.009803957305848598, 0.032679859548807144, 0.006062113679945469, 0.004088250454515219, -0.009111144579946995, -0.023085052147507668, -0.018156928941607475, 0.00045261604827828705, -0.007699374575167894, 0.009817029349505901, 0.012915080413222313, -0.014993519522249699, -0.03450993075966835, -0.01111768838018179, -0.02278439700603485, 0.013346454128623009, -0.00825493223965168, -0.009483695030212402, -0.04062759876251221, 0.0034542609937489033, 0.009535983204841614, -0.009830101393163204, 0.024522965773940086, -0.032549139112234116, -0.018261505290865898, -0.0013357892166823149, -0.0004105407278984785, 0.005290869157761335, -0.013490245677530766, -0.004006550647318363, -0.01328763086348772, 0.005068646278232336, 0.006160153541713953, -0.02169942669570446, -0.004277793690562248, 0.006222245283424854, 0.01921575702726841, 0.026091599836945534, 0.04912436380982399, -0.009849709458649158, -0.01728110946714878, 0.018954318016767502, -0.0025016432628035545, 0.0013504951493814588, -0.01324187871068716, -0.001733666518703103, -0.038771383464336395, 0.001761444378644228, 0.00898042507469654, 0.007777806371450424, -0.01091507263481617, -0.014470641501247883, -0.021934721618890762, -0.010385659523308277, -0.0013406912330538034, 0.003076808759942651, -0.012339914217591286, -0.026013167575001717, -0.015568684786558151, -0.015542540699243546, 0.013477173633873463, 0.00852290727198124, 0.0004865213995799422, -0.012124227359890938, 0.03095436282455921, -0.019372619688510895, 0.031058937311172485, -0.014366066083312035, 0.022522957995533943, -0.02105890028178692, 0.02742493711411953, -0.002681382466107607, 0.008457547053694725, -0.012960831634700298, -0.0030326908454298973, -0.017307253554463387, 0.0014191229129210114, 0.032496850937604904, 0.026679836213588715, 0.014758224599063396, -0.017032742500305176, -0.0018235361203551292, 0.00359805254265666, -0.00378432753495872, -0.00681048259139061, -0.011052328161895275, -0.012882400304079056, -0.008163428865373135, -0.00130229233764112, -0.032993584871292114, -0.007000025827437639, -0.040941327810287476, 0.012424882501363754, 0.01952948421239853, -0.020000074058771133, 0.014888943172991276, -0.002011445350944996, -0.02219616062939167, -0.0007160974200814962, 0.022719038650393486, 0.014902015216648579, -0.02400008775293827, 0.015712475404143333, 0.026549117639660835, 0.010699385777115822, -0.02946416102349758, -0.038274649530649185, -0.02381708100438118, -0.021843217313289642, 0.016483720391988754, 0.04520278051495552, 0.005297405179589987, -0.006297408603131771, 0.005771263036876917, 0.028339972719550133, -0.004535964224487543, -0.023071980103850365, 0.01772555522620678, 0.01664058491587639, -0.0020163473673164845, 0.004653611686080694, -0.013339918106794357, -0.005339888855814934, 0.02061445452272892, -0.014013123698532581, -0.01732032559812069, -0.02471904456615448, 0.013254950754344463, -0.022431455552577972, 0.032000117003917694, -0.02602623961865902, -0.0031519723124802113, -0.00509805791079998, -0.026692908257246017, -0.008333363570272923, -0.020196152850985527, -0.015437965281307697, 0.003522888757288456, -0.0067712669260799885, 0.04669298231601715, -0.007705910596996546, -0.010300691239535809, 0.019228829070925713, 0.010562130250036716, 0.001668306766077876, 0.021137332543730736, -0.010437946766614914, 0.03843151405453682, -0.006398716475814581, -0.015163454227149487, 0.008830097503960133, -0.012307235039770603, 0.02223537676036358, -0.01881052739918232, 0.010424874722957611, -0.017385683953762054, -0.015503324568271637, 0.006006557960063219, -0.013790900819003582, 0.014915087260305882, 0.000986931729130447, 0.013699396513402462, -0.00733989616855979, -0.01583012379705906, 0.0032892278395593166, -0.010934680700302124, -0.012307235039770603, 0.007333360146731138, -0.010902000591158867, -0.02033994346857071, -0.004290865268558264, 0.01290200836956501, 0.0028464156202971935, -0.012666713446378708, 0.028522981330752373, 0.048915211111307144, -0.022026224061846733, 0.007830094546079636, 0.004591519944369793, -0.0020441252272576094, 0.00465687969699502, 3.515637945383787e-05, 0.019451051950454712, -0.004022890701889992, 0.02643146924674511, -0.01058173831552267, -0.003826811444014311, -0.03184325248003006, -0.008235324174165726, -0.004686291795223951, -0.002248374279588461, 0.0016143850516527891, 0.014980447478592396, 0.006297408603131771, -0.0010465724626556039, -0.0023071980103850365, -0.013320310041308403, 0.0019248436437919736, -0.018509872257709503, 0.015817051753401756, 0.01853601634502411, -0.01517652627080679, -0.007385647855699062, -0.0071699609979987144, -0.0003788821049965918, -0.008150356821715832, -0.03866680711507797, -0.028810562565922737, -0.0125163858756423, 0.015751691535115242, -0.01844451203942299, -0.046614550054073334, -0.006941202096641064, -0.0020179813727736473, -0.024575253948569298, 0.026509901508688927, -0.00528760114684701, -0.010790889151394367, 0.03720274940133095, 0.017398755997419357, -0.005395444575697184, 0.013006583787500858, 0.005241849459707737, -0.030980506911873817, -0.007333360146731138, -0.006490219850093126, -0.030980506911873817, -0.027215786278247833, -0.004529428202658892, 0.02125498093664646, 0.010098076425492764, -0.020379159599542618, -0.024026231840252876, 0.012281090952455997, -0.0487322062253952, -0.001451802672818303, -0.03343803063035011, 0.008392187766730785, -0.02214387245476246, -0.015006591565907001, 0.011490237899124622, -0.0027304021641612053, -0.004918318707495928, 0.0011674879351630807, -0.019111180678009987, -0.007862773723900318, -0.01745104417204857, -0.0005996754043735564, 0.013529461808502674, -0.0005375836626626551, -0.04081060737371445, -0.02092818170785904, 0.02188243344426155, 0.014745152555406094, -0.009163432754576206, 0.02490205317735672, 0.011019648052752018, -0.01926804520189762, -0.006663423031568527, 0.0029248474165797234, -0.001826804131269455, -0.006882378365844488, -0.005996753927320242, -0.029150433838367462, -0.006509827915579081, 0.004300669301301241, 0.011921612545847893, 0.00525492150336504, 0.02215694449841976, 0.003428117139264941, 0.015150382183492184, 0.00859480258077383, -0.012640569359064102, -0.024156952276825905, -0.0029526252765208483, -0.014562144875526428, -0.008470619097352028, -0.006830090656876564, -0.011614421382546425, 0.04261453449726105, 0.004571912344545126, -0.005326816812157631, -0.00612420542165637, 0.03657529875636101, -0.009653630666434765, -0.005692831240594387, 0.013032727874815464, 0.0025147150736302137, 0.011712461709976196, -0.00756865506991744, -0.0014828485436737537, 0.013934691436588764, -0.0033333455212414265, -0.0031470702961087227, -0.016575224697589874, -0.027085067704319954, 0.02449682168662548, 0.019738635048270226, -0.01934647560119629, 0.0012328476877883077, 0.003732039825990796, -0.021725570783019066, -0.02057523839175701, -0.010202651843428612, 0.003204260254278779, -0.03040534071624279, -0.00509805791079998, -0.00420589791610837, 8.399745274800807e-05, -0.01073860190808773, -0.01057520229369402, 0.01513731013983488, 0.0028692916966974735, 0.007346432190388441, 0.29385727643966675, -0.016928166151046753, -0.016405289992690086, 0.028705988079309464, 0.012941224500536919, 0.005584987811744213, 0.011202655732631683, -0.010895464569330215, -0.01583012379705906, 0.005826818756759167, 0.014326849952340126, -0.002076805103570223, 0.008039245381951332, 0.0032761557959020138, 0.007444471586495638, -0.02670598030090332, -0.023254986852407455, -0.01934647560119629, 0.0032418419141322374, -0.03892824798822403, 0.005395444575697184, -0.005470608361065388, -0.006150349508970976, -0.021686354652047157, 0.00807846151292324, 0.0003737758961506188, -0.015542540699243546, -0.006300676614046097, 0.011771284975111485, 0.03283672034740448, -0.028130821883678436, -0.01362096518278122, 0.009529447183012962, 0.0030702727381139994, -0.03003932535648346, -0.0033186397049576044, 0.012503313831984997, 0.010588274337351322, 0.03545111045241356, 0.0009256569901481271, -0.0034542609937489033, -0.01163402944803238, 0.012542529962956905, -0.005241849459707737, 0.0015874241944402456, 0.026196174323558807, -0.01107847224920988, -0.022614462301135063, -0.013411814346909523, 0.030614491552114487, -0.004718971438705921, 0.00733989616855979, 0.005460804328322411, 0.02598702348768711, 0.004143806174397469, -0.0054510002955794334, 0.014352994039654732, -0.008183036930859089, 0.021634066477417946, -0.004196093883365393, 0.009176503866910934, 0.02747722528874874, -0.015045806765556335, 0.04394787549972534, -0.038091644644737244, 0.012424882501363754, -0.04271911084651947, -0.01096082478761673, 0.007588263135403395, -0.02237916737794876, 0.006039238069206476, -0.021464131772518158, -0.020640598610043526, 0.0016397119034081697, -0.030875930562615395, -0.012928152456879616, 0.030117757618427277, 0.025372643023729324, 0.03884981572628021, 0.027634087949991226, -0.013699396513402462, 0.003441189182922244, 0.0077320546843111515, -0.00764708686619997, -0.0004946913686580956, -0.04949037730693817, -0.002733670175075531, 0.0015433062799274921, -0.02530728280544281, -0.016313785687088966, 0.011849716305732727, -0.00614708149805665, -0.0003778608806896955, -0.017085030674934387, 0.008810490369796753, 0.014784367755055428, -0.0182745773345232, 0.020536024123430252, -0.025281138718128204, 0.0259347353130579, -0.024483749642968178, 0.0007651171763427556, 0.0019036018056795, 0.038143932819366455, -0.038719095289707184, -0.003323541721329093, -0.007784342393279076, 0.015437965281307697, -0.0028938015457242727, -0.0008382383966818452, -0.01167978160083294, -0.015620972029864788, 0.008274540305137634, 0.003302299650385976, 0.0015669992426410317, 0.008000029250979424, 0.01768633909523487, -0.017973922193050385, 0.00830068439245224, -0.008137284778058529, 0.006941202096641064, -0.028261542320251465, -0.012052332051098347, 0.005447732284665108, -0.003037592861801386, -0.006526167970150709, -0.022614462301135063, -0.004823547322303057, 0.03241841867566109, -0.042875975370407104, 0.028313830494880676, -0.013346454128623009, 0.028758276253938675, -0.0130850151181221, -0.017333397641777992, 0.00847715511918068, -0.006611135322600603, 0.008947744965553284, -0.02120269276201725, -0.007359504234045744, 0.0008627482457086444, 0.0010465724626556039, 0.00623531686142087, 0.012235338799655437, 0.005333352833986282, 0.004669951740652323, 0.022719038650393486, -0.004781063180416822, 0.022862829267978668, -0.021895505487918854, -0.03605242073535919, -0.007281072437763214, -0.0007291693473234773, -0.00807846151292324, 0.026039311662316322, -0.003996746614575386, -0.03681059181690216, -0.03542496636509895, -0.02711120992898941, -0.008045781403779984, -0.03346417471766472, 0.009352975524961948, 0.0432942770421505, -0.01763405092060566, -0.01012422051280737, 0.006751658860594034, -0.16711172461509705, 0.028183110058307648, 0.011660173535346985, -0.008777810260653496, 0.01678437553346157, 0.0017532744677737355, 0.008836633525788784, 0.012392202392220497, -0.003457529004663229, -0.01583012379705906, 0.01930726133286953, -0.01822228915989399, -0.03406548500061035, 0.007516367360949516, -0.002589878858998418, -0.0007589897140860558, -0.02477133274078369, -0.016941238194704056, 0.013817043974995613, 0.0018970657838508487, 0.02836611680686474, -0.018575232475996017, 0.0028954355511814356, -0.006591527722775936, -0.00044322057510726154, 0.0014656917192041874, -0.0036176603753119707, 0.016980454325675964, 0.002312100026756525, -0.004111126065254211, 3.694355837069452e-05, 0.0026879184879362583, 0.013477173633873463, 0.0003899115545209497, 0.025045843794941902, 0.02107197232544422, 0.013921619392931461, -0.012549065984785557, 0.0006066198693588376, 0.020039290189743042, 0.03307201713323593, 0.016353001818060875, 0.007346432190388441, 0.0012785994913429022, 0.0025670030154287815, 0.02516349032521248, 0.01858830451965332, -0.0053856405429542065, 0.02214387245476246, 0.00030862042331136763, -0.00023856296320445836, -0.002313734032213688, 0.020967397838830948, -0.0018986997893080115, -0.010026181116700172, 0.03252299502491951, -0.03103279322385788, 0.019098108634352684, 0.0014101358829066157, 0.007111137267202139, -0.004771259147673845, -0.01588241197168827, -0.015254957601428032, 0.006320284679532051, 0.009300687350332737, -0.032679859548807144, -0.023516425862908363, 0.010000037029385567, -0.0007177314255386591, 0.0013872600393369794, -0.023150412365794182, -0.02101968601346016, 0.005849694833159447, -0.003037592861801386, 0.020536024123430252, -0.004872567020356655, -0.02611774392426014, 0.026902059093117714, -0.006856234278529882, -0.02160792239010334, -0.007941205985844135, 0.01084971334785223, -0.004980410449206829, 0.0011127492180094123, 0.003401973284780979, 0.00985624548047781, -0.0022238644305616617, -0.004591519944369793, 0.010973896831274033, -0.005709171295166016, 0.012483705766499043, -0.03850994631648064, -0.018549088388681412, -0.012601353228092194, -0.016810519620776176, 0.012817041017115116, 0.00836604367941618, -0.002356217708438635, -0.004009818658232689, -0.0191765408962965, -0.009352975524961948, -0.007993493229150772, -0.008915065787732601, 0.022274591028690338, 0.01722882129251957, 0.015490253455936909, 0.013673252426087856, 0.01844451203942299, 0.020405303686857224, -0.010607882402837276, -0.024379175156354904, 0.0046960958279669285, 0.012862792238593102, 0.027058923617005348, 0.004973874427378178, 0.01610463485121727, -0.024078520014882088, -0.023333419114351273, 0.019189612939953804, 0.006647083442658186, -0.0064117880538105965, -0.009503303095698357, -0.01217651553452015, 0.01877131126821041, 0.014575216919183731, -0.010810497216880322, -0.05845773220062256, -0.0050523062236607075, 0.02805239148437977, 0.028679843991994858, -0.004130734130740166, 0.016575224697589874, 0.029987039044499397, 0.018928173929452896, -0.010875857435166836, -0.004751651547849178, -0.0016372608952224255, -0.010973896831274033, -0.0022254984360188246, -0.0005759824998676777, 0.007607871200889349, -0.0038660273421555758, -0.0025964148808270693, -0.015189598314464092, -0.00039787727291695774, 0.03929426148533821, 0.0017303985077887774, 1.956961932592094e-05, 0.010470626875758171, -0.0077516622841358185, -0.004267989657819271, 0.012751680798828602, -0.026457613334059715, 0.03884981572628021, 0.01614385098218918, -0.009875853545963764, 0.026640620082616806, -0.007084993179887533, 0.004836618900299072, -0.007843165658414364, 0.014431425370275974, -0.002102948958054185, -0.019464123994112015, -0.00031209265580400825, 0.02499355562031269, -0.023307275027036667, 0.003535960800945759, 0.013673252426087856, 0.005372568964958191, -0.025738656520843506, -0.0024656953755766153, -0.008745130151510239, -0.0003656059270724654, 0.027843238785862923, 0.015411821193993092, -0.017607908695936203, -0.00990199763327837, -0.004843154922127724, -0.022405311465263367, 0.018745167180895805, 0.029490305110812187, -0.017882417887449265, 0.014849727973341942, -0.007268000394105911, -0.011588278226554394, -0.008503299206495285, -0.015712475404143333, -0.00578433508053422, -0.026758268475532532, 0.00940526369959116, 0.029699455946683884, 0.0019150397274643183, 0.0193987637758255, -0.009882389567792416, 0.014379138126969337, 0.003522888757288456, 0.00018658157205209136, 0.016117706894874573, -0.00841179583221674, 0.013725540600717068, -0.032549139112234116, -0.021228836849331856, -0.03568640723824501, -0.014026195742189884, -0.0008218984585255384, -0.008516371250152588, -0.02169942669570446, -0.018431439995765686, 0.01745104417204857, -0.0236863624304533, 0.01615692302584648, 0.011529454030096531, 0.010941216722130775, -0.012032723985612392, 0.014156915247440338, -0.030928218737244606, -0.015398749150335789, 0.009339903481304646, 0.02576480060815811, -0.020353015512228012, -0.0035000129137188196, 0.011496773920953274, 0.01755562052130699, 0.004375833086669445, 0.023960871621966362, 0.0010759843280538917, -0.022993547841906548, -0.021764785051345825, -0.07691531628370285, -0.004950998816639185, 0.025411857292056084, 0.008045781403779984, 0.012568674050271511, -0.007078457623720169, 0.002808833960443735, -0.0012647105613723397, -0.024013159796595573, -0.011895468458533287, -0.001453436678275466, -0.012202659621834755, 0.010503306984901428, -0.013555604964494705, -0.0241308081895113, -0.03539882227778435, 0.023385707288980484, -0.0009861147264018655, 0.01687587983906269, 0.015908556059002876, 0.008379115723073483, 0.004686291795223951, 0.0054738763719797134, 0.007457543630152941, -0.011248406954109669, 0.00525492150336504, -0.028888994827866554, 0.020862821489572525, -0.014052338898181915, -0.014130771160125732, -0.015215742401778698, -0.02567329630255699, -0.0026487025897949934, 0.01007846836000681, -0.020954325795173645, -0.020000074058771133, 0.014104627072811127, 0.013503317721188068, 0.01224841084331274, 0.007248392794281244, -0.017163461074233055, -0.03108508139848709, 0.008248396217823029, 0.0053660329431295395, -0.0329151526093483, 0.012039260007441044, -0.01713731698691845, 0.006173225119709969, 0.03500666469335556, 0.0009803957073017955, 0.007032705470919609, 0.011934684589505196, -0.012549065984785557, -0.012542529962956905, 0.014888943172991276, -0.01190854050219059, 0.037830203771591187, 0.010542522184550762, -0.010712457820773125, -0.027398793026804924, 0.027032779529690742, 0.001944451592862606, -0.014719008468091488, -0.004954266361892223, 0.01579090766608715, -0.002980403136461973, -0.012052332051098347, 0.009601342491805553, 0.004153610207140446, -0.02534649893641472, -0.009941212832927704, 0.007424863986670971, 0.023150412365794182, 0.01168631762266159, -0.018836671486496925, -0.012601353228092194, 0.0020163473673164845, -0.03720274940133095, -0.03532039001584053, 0.03160795941948891, 0.017111174762248993, 0.004503284581005573, -0.017581764608621597, 0.004199361894279718, 0.009287616237998009, 0.05349039286375046, -0.004320277366787195, 0.0013856260338798165, -0.006718978751450777, 0.034536074846982956, 0.014758224599063396, 0.010235331952571869, -0.01136605441570282, 0.034170061349868774, -0.022222304716706276, 0.046117816120386124, 0.01601313054561615, -0.0072222487069666386, 0.030875930562615395, 0.03132037818431854, 0.027503369376063347, -0.000991016742773354, -0.004676487762480974, -0.005245117470622063, -0.010268012061715126, 0.004957534372806549, -0.022719038650393486, -0.02896742708981037, -0.011052328161895275, -0.010947752743959427, 0.018235361203551292, -0.008915065787732601, 0.0050751818343997, 0.02692820318043232, -0.025464145466685295, 0.011647101491689682, 0.008967353031039238, -0.00462420005351305, -0.023333419114351273, 0.022849757224321365, 0.01809157058596611, 0.011006576009094715, 0.02039223164319992, -0.007509831339120865, 0.008281076326966286, 0.021398771554231644, -0.0030522989109158516, 0.006192833185195923, 0.015490253455936909, 0.009555590339004993, 0.0023366098757833242, -0.00282517378218472, -0.02015693672001362, -0.01511116698384285, -0.02056216634809971, 0.01647064834833145, -0.037281181663274765, 0.021268051117658615, -0.02711120992898941, 0.05124201998114586, 0.00013633628259412944, -0.025607937946915627, 0.01993471384048462, -0.01389547623693943, 0.01566018909215927, 0.004692827817052603, 0.009594806469976902, -0.025607937946915627, -0.009006569162011147, 0.0029591612983494997, -0.015359533950686455, 0.00047222396824508905, -0.006849698256701231, -0.049333516508340836, 0.0069542741402983665, 0.014888943172991276, -0.0003425257746130228, -0.01247063372284174, 0.017215749248862267, 0.014078482985496521, 0.007908525876700878, -0.010215723887085915, 0.022457599639892578, -0.027712520211935043, -0.010764745064079762, 0.01427456270903349, -0.0036993599496781826, 0.006483683828264475, -0.03830079361796379, 0.013437957502901554, 0.0025163490790873766, -0.026326894760131836, -0.01925497315824032, 0.00820918008685112, -0.02070595882833004, -0.022902045398950577, -0.01588241197168827, 0.0030130830127745867, 0.018392223864793777, 0.022810541093349457, 0.030928218737244606, -0.029019715264439583, -0.024692900478839874, 0.052261628210544586, -0.010549058206379414, -0.014065410941839218, 0.00963402260094881, 0.00562747148796916]\n",
      "1536\n"
     ]
    }
   ],
   "source": [
    "print(positive_review)\n",
    "print(len(positive_review))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cosine Similarity between positive and neutual review: 0.9348103775852385\n"
     ]
    }
   ],
   "source": [
    "print(\"Cosine Similarity between positive and neutual review:\", cosine_similarity(positive_review, negative_review))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cosine Similarity between positive review and itself: 1.0000000000000002\n"
     ]
    }
   ],
   "source": [
    "print(\"Cosine Similarity between positive review and itself:\", cosine_similarity(positive_review, positive_review))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cosine Similarity between positive and good restaurant review: 0.8378337749480329\n",
      "Cosine Similarity between negative and good restaurant review: 0.8045361482843162\n"
     ]
    }
   ],
   "source": [
    "good_restraurant = get_embedding(\"这家餐馆太好吃了，一点都不糟糕\", engine=EMBEDDING_MODEL)\n",
    "bad_restraurant = get_embedding(\"这家餐馆太糟糕了，一点都不好吃\", engine=EMBEDDING_MODEL)\n",
    "\n",
    "print(\"Cosine Similarity between positive and good restaurant review:\", cosine_similarity(positive_review, good_restraurant))\n",
    "print(\"Cosine Similarity between negative and good restaurant review:\", cosine_similarity(negative_review, good_restraurant))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "好评例子的评分 : 0.033298\n",
      "差评例子的评分 : -0.027249\n"
     ]
    }
   ],
   "source": [
    "\n",
    "def get_score(sample_embedding):\n",
    "  return cosine_similarity(sample_embedding, positive_review) - cosine_similarity(sample_embedding, negative_review)\n",
    "\n",
    "positive_score = get_score(good_restraurant)\n",
    "negative_score = get_score(bad_restraurant)\n",
    "\n",
    "\n",
    "print(\"好评例子的评分 : %f\" % (positive_score))\n",
    "print(\"差评例子的评分 : %f\" % (negative_score))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 生成一个正交的，Cosine Similarity 为 0 的向量"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-9.09016681580666e-18\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "\n",
    "# 假设原向量是一个高维向量\n",
    "original_vector = np.array(positive_review)\n",
    "\n",
    "# 对原向量进行归一化\n",
    "original_vector /= np.linalg.norm(original_vector)\n",
    "\n",
    "# 从单位高斯分布中随机抽取一个向量\n",
    "random_vector = np.random.randn(len(original_vector))\n",
    "\n",
    "# 从随机向量中减去在原向量方向上的分量\n",
    "orthogonal_vector = random_vector - random_vector.dot(original_vector) * original_vector\n",
    "\n",
    "# 验证新向量与原向量是否正交（即它们的余弦相似度是否为0）\n",
    "cos_sim = np.dot(original_vector, orthogonal_vector) / (np.linalg.norm(original_vector) * np.linalg.norm(orthogonal_vector))\n",
    "print(cos_sim)  # 这个值应该接近0\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 降维可视化向量"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "from sklearn.decomposition import PCA\n",
    "\n",
    "def visualize_words(words,vectors):\n",
    "    # Create a PCA object and fit_transform the vectors\n",
    "    pca = PCA(n_components=2)\n",
    "    pca_result = pca.fit_transform(vectors)\n",
    "\n",
    "    # Create a color map\n",
    "    colors = [\"b\", \"g\", \"r\", \"c\"]\n",
    "\n",
    "    # Plot each word in 2D space using PCA results\n",
    "    for i, word in enumerate(words):\n",
    "        plt.scatter(pca_result[i, 0], pca_result[i, 1], c=colors[i], label=word)\n",
    "        plt.annotate(word, xy=(pca_result[i, 0], pca_result[i, 1]))\n",
    "\n",
    "    plt.legend()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAo0AAAGdCAYAAAB3k0lLAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAABeK0lEQVR4nO3deVxU9f4/8NcMMMPIqkIsOgoEKhqiuSCYokmBluFuigper5a5ofIzNbeWK1pa5Ep6v6mUXi1DK80FCVwQcSUVcF9wAVwBAWWbz+8PcmoCPIAMqLyej8d55JzzOZ/zPp+H987Lz1lGJoQQICIiIiJ6AnltF0BEREREzz6GRiIiIiKSxNBIRERERJIYGomIiIhIEkMjEREREUliaCQiIiIiSQyNRERERCSJoZGIiIiIJBnWdgG1QaPR4ObNmzAzM4NMJqvtcoiIiKgChBB48OAB7O3tIZdz3qum1cnQePPmTajV6toug4iIiKrg2rVraNy4cW2XUefUydBoZmYGoOQvnbm5eS1XQ0RERBWRnZ0NtVqt/R6nmlUnQ+PjS9Lm5uYMjURERM8Z3lpWO3hDABERERFJYmgkIiIiIkkMjUREREQkiaGRiIiI9Eomk2Hr1q21XQY9JYZGIiKiF1x6ejomTZoEZ2dnGBsbw8bGBp07d8bKlSuRl5dX2+UhKCgIMpkMMpkMRkZGcHR0xLRp0/Do0aNq6f/KlSuQyWRITEyslv5qyrNWd518epqIiKiuuHTpEjp37gxLS0vMnz8fbm5uUCqVOHXqFFatWoVGjRrhnXfeqe0y4efnhzVr1qCwsBDHjh1DYGAgZDIZFi5cWNulVYoQAsXFxTA0fPEiFmcaiYiIXmAffPABDA0NcfToUQwaNAiurq5wcnKCv78/tm/fjt69e2vbpqamwt/fH6ampjA3N8egQYOQkZGh09/KlSvx8ssvQ6FQoHnz5vjuu+90tp8/fx5du3aFsbExWrZsiaioqArVqVQqYWtrC7VajT59+sDHx0dnX41Gg8WLFwMAbGxs4O7ujs2bN2u3379/HwEBAbC2toZKpYKLiwvWrFkDAHB0dAQAtG3bFjKZDN26dQMAHDlyBG+88QasrKxgYWEBb29vHD9+XNtnWTN9mZmZkMlkiI2NBQDExsZCJpNhx44daNeuHZRKJQ4cOICLFy/C398fNjY2MDU1RYcOHbBnzx6dc3ZwcMD8+fPxr3/9C2ZmZmjSpAlWrVql3V5e3bWFoZGIiOgFdffuXezevRvjxo2DiYlJmW0ev/NQo9HA398f9+7dw969exEVFYVLly5h8ODB2rZbtmzBpEmTMHXqVJw+fRrvvfceRo4ciZiYGG0f/fr1g0KhQEJCAsLDw/Hhhx9Wuu7Tp0/j4MGDUCgU2nWhoaHYuHEjAODQoUOYPHkyhg0bhr179wIAZs+ejeTkZOzYsQMpKSlYuXIlrKysAACHDx8GAOzZswdpaWmIjIwEADx48ACBgYE4cOAADh06BBcXF/Tq1QsPHjyodM3Tp0/HggULkJKSgtatWyMnJwe9evVCdHQ0Tpw4AT8/P/Tu3Rupqak6+y1evBjt27fHiRMn8MEHH2Ds2LE4e/bsE+uuNaIOysrKEgBEVlZWbZdCRERU/YqKhIiJEYc++UQAEJGbN+tsbtiwoTAxMREmJiZi2rRpQgghdu/eLQwMDERqaqq2XVJSkgAgDh8+LIQQwsvLS4wePVqnr4EDB4pevXoJIYTYtWuXMDQ0FDdu3NBu37FjhwAgtmzZUm65gYGBwsDAQJiYmAilUikACLlcLjb/WfejR49EvXr1RFRUlM7396hRo8SQIUOEEEL07t1bjBw5ssz+L1++LACIEydOPHHYiouLhZmZmfj111/L3e/+/fsCgIiJiRFCCBETEyMAiK1btz6xbyGEaNWqlVi6dKn2c9OmTcWwYcO0nzUajXjppZfEypUrK1V3TeFMIxER0YskMhJwcAC6dwfmzClZ9957Jev/dPjwYSQmJqJVq1bIz88HAKSkpECtVkOtVmvbtWzZEpaWlkhJSdG26dy5s87hOnfurLNdrVbD3t5eu93T07NCZXfv3h2JiYlISEhAYGAgRo4cif79+wMALly4gLy8PPTp0wcAYG9vD1NTU0RERODixYsAgLFjx2Ljxo1o06YNpk2bhoMHD0oeMyMjA6NHj4aLiwssLCxgbm6OnJycUrOBFdG+fXudzzk5OQgJCYGrqyssLS1hamqKlJSUUn23bt1a+2eZTAZbW1vcunWr0sevCS/eXZpERER1VWQkMGAAIAQAwBmADMDZu3dL1m/eDPTrBycnJwCASqWqvVr/wcTEBM7OzgCAb7/9Fu7u7vi///s/jBo1Cjk5OQCAH374AW+99Rb279+v/f1ppVIJAOjZsyeuXr2K3377DVFRUejRowfGjRuHRYsWlXvMwMBA3L17F19//TWaNm0KpVIJT09PFBQUAADk8pK5NfHneAJAYWFhufX/XUhICKKiorBo0SI4OztDpVJhwIAB2r4fMzIy0vksk8mg0WiePFi1hDONREREL4LiYmDSJG1gBICGAN4AsAxArhBAcHBJuzK4urri2rVruHbtmnZdcnIyMjMz0bJlS22buLg4nf3i4uJ0tl+7dg1paWna7YcOHar0qcjlcsycOROzZs3Cw4cP0bJlSyiVSly/fh0A8PLLL8PZ2RnOzs46M6PW1tYIDAzE999/j7CwMO1DJY/vjSz+x7nHxcVh4sSJ6NWrF1q1agWlUok7d+7o9AdA53wq+vqbuLg4BAUFoW/fvnBzc4OtrS2uXLlSqXEor+7awtBIRET0Iti/H/gzVP3dCgBFANoD2HTtGlK+/x5nz57F999/jzNnzsDAwAAA4OPjAzc3NwQEBOD48eM4fPgwRowYAW9vb+2l1//3//4f1q5di5UrV+L8+fP48ssvERkZiZCQEG0fzZo1Q2BgIP744w/s378fH330UZVOZ+DAgTAwMMDy5cthZmaGkJAQzJgxA0DJa4SOHz+OpUuXYt26dQCAOXPm4Oeff8aFCxeQlJSEbdu2wdXVFQDw0ksvQaVSYefOncjIyEBWVhYAwMXFBd999x1SUlKQkJCAgIAAndlXlUqFTp06aR9w2bt3L2bNmlWh+l1cXBAZGYnExET88ccfGDp0aKVnEMuru9bU9k2VtYEPwhAR0QtnwwYhSuYZSy03ATEeEI6AMDIwEKampqJjx47iiy++ELm5udourl69Kt555x1hYmIizMzMxMCBA0V6errOYVasWCGcnJyEkZGRaNasmYiIiNDZfvbsWfHaa68JhUIhmjVrJnbu3FmhB2H8/f1LrQ8NDRXW1tYiJydHaDQasWDBAgFAGBkZCWtra+Hr6yv27t0rhBDi008/Fa6urkKlUokGDRoIf39/cenSJW1fq1evFmq1WsjlcuHt7S2EEOL48eOiffv2wtjYWLi4uIgff/xRNG3aVHz11Vfa/ZKTk4Wnp6dQqVSiTZs2Yvfu3WU+CHP//n2d2i9fviy6d+8uVCqVUKvVYtmyZcLb21tMmjRJ2+afxxJCCHd3dzF37twn1l1bZEL8bR67jsjOzoaFhQWysrJgbm5e2+UQERE9vdjYkodfpMTEALX8vr+q4vd37eLlaSIiohdBly5A48bAn+9dLEUmA9TqknZEVcDQSERE9CIwMAC+/rrkz/8Mjo8/h4WVtCOqAoZGIiKiF0W/fiWv1WnUSHd948ba1+0QVRXf00hERPQi6dcP8PcveZo6LQ2wsyu5JM0ZRnpKDI1EREQvGgOD5/ZhF3p28fI0EREREUliaCQiIiIiSQyNRERERCSJoZGIiIiIJDE0EhEREZEkhkYiIiIiksTQSERERESSGBqJiIiISBJD43MqNjYWMpkMmZmZT2zn4OCAsLCwGqmJiIiIXlw1EhqXL18OBwcHGBsbw8PDA4cPHy63bVJSEvr37w8HBwfIZLIyA8+8efMgk8l0lhYtWujxDKouKChIW6NCoYCzszM++eQTFBUVPVW/Xl5eSEtLg4WFBQBg7dq1sLS0LNXuyJEjGDNmzFMdi4iIiEjvoXHTpk2YMmUK5s6di+PHj8Pd3R2+vr64detWme3z8vLg5OSEBQsWwNbWttx+W7VqhbS0NO1y4MABfZ3CU/Pz80NaWhrOnz+PqVOnYt68efjiiy+eqk+FQgFbW1vIZLIntrO2tka9evWe6lhEREREeg+NX375JUaPHo2RI0eiZcuWCA8PR7169fDtt9+W2b5Dhw744osv8O6770KpVJbbr6GhIWxtbbWLlZWVvk7hqSmVStja2qJp06YYO3YsfHx88Msvv+D+/fsYMWIE6tevj3r16qFnz544f/68dr+rV6+id+/eqF+/PkxMTNCqVSv89ttvAHQvT8fGxmLkyJHIysrSzmrOmzcPgO7l6aFDh2Lw4ME6tRUWFsLKygoREREAAI1Gg9DQUDg6OkKlUsHd3R2bN2/W/yARERHRM02vobGgoADHjh2Dj4/PXweUy+Hj44P4+Pin6vv8+fOwt7eHk5MTAgICkJqaWm7b/Px8ZGdn6yy1SaVSoaCgAEFBQTh69Ch++eUXxMfHQwiBXr16obCwEAAwbtw45OfnY9++fTh16hQWLlwIU1PTUv15eXkhLCwM5ubm2pnXkJCQUu0CAgLw66+/IicnR7tu165dyMvLQ9++fQEAoaGhiIiIQHh4OJKSkjB58mQMGzYMe/fu1dNoEBER0fPAUJ+d37lzB8XFxbCxsdFZb2NjgzNnzlS5Xw8PD6xduxbNmzdHWloaPv74Y3Tp0gWnT5+GmZlZqfahoaH4+OOPq3y86iKEQHR0NHbt2oWePXti69atiIuLg5eXFwBg/fr1UKvV2Lp1KwYOHIjU1FT0798fbm5uAAAnJ6cy+1UoFLCwsIBMJnviJX1fX1+YmJhgy5YtGD58OABgw4YNeOedd2BmZob8/HzMnz8fe/bsgaenp/aYBw4cwDfffANvb+/qHA4iIiJ6jug1NOpLz549tX9u3bo1PDw80LRpU/zwww8YNWpUqfYzZszAlClTtJ+zs7OhVqv1UltxMbB/P5CWBtjZAUIA27Ztg6mpKQoLC6HRaDB06FD069cP27Ztg4eHh3bfhg0bonnz5khJSQEATJw4EWPHjsXu3bvh4+OD/v37o3Xr1lWuzdDQEIMGDcL69esxfPhw5Obm4ueff8bGjRsBABcuXEBeXh7eeOMNnf0KCgrQtm3bKh+XiIiInn96DY1WVlYwMDBARkaGzvqMjIwnzohVlqWlJZo1a4YLFy6UuV2pVD7x/sjqEhkJTJoEXL/+17p69YBWrbrjp59WQqFQwN7eHoaGhvjll18k+/v3v/8NX19fbN++Hbt370ZoaCgWL16MCRMmVLnGgIAAeHt749atW4iKioJKpYKfnx8AaC9bb9++HY0aNdLZrybGj4iIiJ5der2nUaFQoF27doiOjtau02g0iI6O1l7+rA45OTm4ePEi7Ozsqq3PyoqMBAYM0A2MAJCXB5w8aYKTJ53RpEkTGBqW5HRXV1cUFRUhISFB2/bu3bs4e/YsWrZsqV2nVqvx/vvvIzIyElOnTsXq1avLPL5CoUBxcbFknV5eXlCr1di0aRPWr1+PgQMHwsjICADQsmVLKJVKpKamwtnZWWfR18wsERERPR/0fnl6ypQpCAwMRPv27dGxY0eEhYUhNzcXI0eOBACMGDECjRo1QmhoKICSS6HJycnaP9+4cQOJiYkwNTWFs7MzACAkJAS9e/dG06ZNcfPmTcydOxcGBgYYMmSIvk+nTMXFJTOMQpTfJjgY8PcHDAxKPru4uMDf3x+jR4/GN998AzMzM0yfPh2NGjWCv7//n/sEo2fPnmjWrBnu37+PmJgYuLq6ltm/g4MDcnJyEB0dDXd3d9SrV6/cV+0MHToU4eHhOHfuHGJiYrTrzczMEBISgsmTJ0Oj0eC1115DVlYW4uLiYG5ujsDAwCqNDxEREb0ARA1YunSpaNKkiVAoFKJjx47i0KFD2m3e3t4iMDBQ+/ny5csCQKnF29tb22bw4MHCzs5OKBQK0ahRIzF48GBx4cKFCteTlZUlAIisrKzqOD0REyNESWQsawkUgL8AStr93b1798Tw4cOFhYWFUKlUwtfXV5w7d067ffz48eLll18WSqVSWFtbi+HDh4s7d+78ecwYAUDcv39f2/79998XDRs2FADE3LlzhRBCNG3aVHz11Vc6x01OThYARNOmTYVGo9HZptFoRFhYmGjevLkwMjIS1tbWwtfXV+zdu7daxoqIiKiqqvv7mypHJsST5sdeTNnZ2bCwsEBWVhbMzc2fur///Q8YOlS63YYNQC1NhhIRET33qvv7myqHvz1dDSp6K2Ut3nJJRERE9FQYGqtBly5A48ZAeb/oJ5MBanVJOyIiIqLnEUNjNTAwAL7+uuTP/wyOjz+Hhf31EAwRERHR84ahsZr06wds3gz84/WGaNy4ZH2/frVTFxEREVF1eC5/EeZZ1a9fyWt1/v6LMF26cIaRiIiInn8MjdXMwADo1q22qyAiIiKqXrw8TURERESSGBqJiIiISBJDIxERERFJYmgkIiIiIkkMjUREREQkiaGRiIiIiCQxNBIRERGRJIZGIiIiIpLE0EhEREREkhgaiYiIiEgSQyMRERERSWJoJCIiIiJJDI1EREREJImhkYiIiIgkMTQSERERkSSGRiIiIiKSxNBIRERERJIYGomIiIhIEkMjEREREUliaCQiIiIiSQyNRERERCSJoZGIiIiIJDE0EhEREZEkhkYiIiIiksTQSERERESSGBqJiIiISBJDIxERERFJYmgkIiIiIkkMjUREREQkiaGRiIiIiCQxNBIRERGRJIZGIiIiIpLE0EhEREREkhgaiYiIiEhSjYTG5cuXw8HBAcbGxvDw8MDhw4fLbZuUlIT+/fvDwcEBMpkMYWFhT90nERERET0dvYfGTZs2YcqUKZg7dy6OHz8Od3d3+Pr64tatW2W2z8vLg5OTExYsWABbW9tq6ZOIiIiIno5MCCH0eQAPDw906NABy5YtAwBoNBqo1WpMmDAB06dPf+K+Dg4OCA4ORnBwcLX1CQDZ2dmwsLBAVlYWzM3Nq3ZiREREVKP4/V279DrTWFBQgGPHjsHHx+evA8rl8PHxQXx8fI31mZ+fj+zsbJ2FiIiIiCpOr6Hxzp07KC4uho2Njc56GxsbpKen11ifoaGhsLCw0C5qtbpKxyYiIiKqq+rE09MzZsxAVlaWdrl27Vptl0RERET0XDHUZ+dWVlYwMDBARkaGzvqMjIxyH3LRR59KpRJKpbJKxyMiIiIiPc80KhQKtGvXDtHR0dp1Go0G0dHR8PT0fGb6JCIiIqIn0+tMIwBMmTIFgYGBaN++PTp27IiwsDDk5uZi5MiRAIARI0agUaNGCA0NBVDyoEtycrL2zzdu3EBiYiJMTU3h7OxcoT6JiIiIqHrpPTQOHjwYt2/fxpw5c5Ceno42bdpg586d2gdZUlNTIZf/NeF58+ZNtG3bVvt50aJFWLRoEby9vREbG1uhPomIiIioeun9PY3PIr7niYiI6PnD7+/aVSeeniYiIiKip8PQSERERESSGBqJiIiISBJDIxERERFJYmgkIiIiIkkMjUREREQkiaGRiIiIiCQxNBIRERGRJIZGIiIiIpLE0EhEREREkhgaiYiIiEgSQyMRERERSWJoJCIiIiJJDI1EREREJImhkYiIiIgkMTQSERERkSSGRiIiIiKSxNBIRERERJIYGomIiIhIEkMjEREREUliaCQiIiIiSQyNRERERCSJoZGIiIiIJDE0EhEREZEkhkYiIiIiksTQSERERESSGBqJiIiISBJDIxERERFJYmgkIiIiIkkMjUREREQkiaGRiIiIiCQxNBIRERGRJIZGIiIiIpLE0EhEREREkhgaiYiIiEgSQyMRERERSWJoJCIiIiJJDI1EREREJImhkYiIiIgkMTQSERERkSSGRiIiIiKSVCOhcfny5XBwcICxsTE8PDxw+PDhJ7b/8ccf0aJFCxgbG8PNzQ2//fabzvagoCDIZDKdxc/PT5+nQERERFSn6T00btq0CVOmTMHcuXNx/PhxuLu7w9fXF7du3Sqz/cGDBzFkyBCMGjUKJ06cQJ8+fdCnTx+cPn1ap52fnx/S0tK0y//+9z99nwoRERFRnSUTQgh9HsDDwwMdOnTAsmXLAAAajQZqtRoTJkzA9OnTS7UfPHgwcnNzsW3bNu26Tp06oU2bNggPDwdQMtOYmZmJrVu3Vqmm7OxsWFhYICsrC+bm5lXqg4iIiGoWv79rl15nGgsKCnDs2DH4+Pj8dUC5HD4+PoiPjy9zn/j4eJ32AODr61uqfWxsLF566SU0b94cY8eOxd27d8utIz8/H9nZ2ToLEREREVWcXkPjnTt3UFxcDBsbG531NjY2SE9PL3Of9PR0yfZ+fn6IiIhAdHQ0Fi5ciL1796Jnz54oLi4us8/Q0FBYWFhoF7Va/ZRnRkRERFS3GNZ2AVXx7rvvav/s5uaG1q1b4+WXX0ZsbCx69OhRqv2MGTMwZcoU7efs7GwGRyIiAEIIFBUVlfuPbqKaZGBgAENDQ8hkstouhcqg19BoZWUFAwMDZGRk6KzPyMiAra1tmfvY2tpWqj0AODk5wcrKChcuXCgzNCqVSiiVyiqcARHRi6ugoABpaWnIy8ur7VKItOrVqwc7OzsoFIraLoX+Qa+hUaFQoF27doiOjkafPn0AlDwIEx0djfHjx5e5j6enJ6KjoxEcHKxdFxUVBU9Pz3KPc/36ddy9exd2dnbVWT4R0QtLo9Hg8uXLMDAwgL29PRQKBWd3qFYJIVBQUIDbt2/j8uXLcHFxgVzO10k/S/R+eXrKlCkIDAxE+/bt0bFjR4SFhSE3NxcjR44EAIwYMQKNGjVCaGgoAGDSpEnw9vbG4sWL8dZbb2Hjxo04evQoVq1aBQDIycnBxx9/jP79+8PW1hYXL17EtGnT4OzsDF9fX32fDhHRC6GgoED7Not69erVdjlEAACVSgUjIyNcvXoVBQUFMDY2ru2S6G/0HhoHDx6M27dvY86cOUhPT0ebNm2wc+dO7cMuqampOv+S8PLywoYNGzBr1izMnDkTLi4u2Lp1K1555RUAJfc7nDx5EuvWrUNmZibs7e3x5ptv4tNPP+UlaCKiSuJMDj1r+Hfy2aX39zQ+i/ieJyKq6x49eoTLly/D0dGRszn0THnS301+f9cuxnkiIiKUvP9XJpMhMzPzie0cHBwQFhZWIzURPUsYGomI6LkSFBQEmUwGmUwGhUIBZ2dnfPLJJygqKnqqfr28vJCWlgYLCwsAwNq1a2FpaVmq3ZEjRzBmzJinOhbR8+i5fE8jERE9G4qLgf37gbQ0wM4O6NIFMDDQ/3H9/PywZs0a5Ofn47fffsO4ceNgZGSEGTNmVLlPhULxxNe7PWZtbV3lYxA9zzjTSEREVRIZCTg4AN27A0OHlvzXwaFkvb4plUrY2tqiadOmGDt2LHx8fPDLL7/g/v37GDFiBOrXr4969eqhZ8+eOH/+vHa/q1evonfv3qhfvz5MTEzQqlUr/PbbbwB0L0/HxsZi5MiRyMrK0s5qzps3D4Du5emhQ4di8ODBOrUVFhbCysoKERERAEpebxQaGgpHR0eoVCq4u7tj8+bN+h8komrG0EhERJUWGQkMGABcv667/saNkvU1ERz/TqVSoaCgAEFBQTh69Ch++eUXxMfHQwiBXr16obCwEAAwbtw45OfnY9++fTh16hQWLlwIU1PTUv15eXkhLCwM5ubmSEtLQ1paGkJCQkq1CwgIwK+//oqcnBztul27diEvLw99+/YFUPJTthEREQgPD0dSUhImT56MYcOGYe/evXoaDSL94OVpIiKqlOJiYNIkoKx3bwgByGRAcDDg76//S9VCCERHR2PXrl3o2bMntm7diri4OHh5eQEA1q9fD7Vaja1bt2LgwIFITU1F//794ebmBqDkF8XKolAoYGFhAZlM9sRL1r6+vjAxMcGWLVswfPhwAMCGDRvwzjvvwMzMDPn5+Zg/fz727Nmj/ZEKJycnHDhwAN988w28vb2rcziI9IozjUREVCn795eeYfw7IYBr10ra6cu2bdtgamoKY2Nj9OzZE4MHD0ZQUBAMDQ3h4eGhbdewYUM0b94cKSkpAICJEyfis88+Q+fOnTF37lycPHnyqeowNDTEoEGDsH79egBAbm4ufv75ZwQEBAAALly4gLy8PLzxxhswNTXVLhEREbh48eJTHZuopjE0EhFRpaSlVW+7qujevTsSExNx/vx5PHz4EOvWravQzyD++9//xqVLlzB8+HCcOnUK7du3x9KlS5+qloCAAERHR+PWrVvYunUrVCoV/Pz8AEB72Xr79u1ITEzULsnJybyvkZ47DI1ERFQpdnbV264qTExM4OzsjCZNmsDQsOROK1dXVxQVFSEhIUHb7u7duzh79ixatmypXadWq/H+++8jMjISU6dOxerVq8s8hkKhQHFxsWQtXl5eUKvV2LRpE9avX4+BAwfCyMgIANCyZUsolUqkpqbC2dlZZ1Gr1U8zBEQ1jvc0EhFRpXTpAjRuXPLQS1n3NcpkJdu7dKnZulxcXODv74/Ro0fjm2++gZmZGaZPn45GjRrB398fABAcHIyePXuiWbNmuH//PmJiYuDq6lpmfw4ODsjJyUF0dDTc3d1Rr169cn+ne+jQoQgPD8e5c+cQExOjXW9mZoaQkBBMnjwZGo0Gr732GrKyshAXFwdzc3MEBgZW/0AQ6QlnGomIqFIMDICvvy758z+vCD/+HBZWM+9r/Kc1a9agXbt2ePvtt+Hp6QkhBH777TftzF9xcTHGjRsHV1dX+Pn5oVmzZlixYkWZfXl5eeH999/H4MGDYW1tjc8//7zc4wYEBCA5ORmNGjVC586ddbZ9+umnmD17NkJDQ7XH3b59OxwdHavvxIlqAH97mr9dSUR1UHX89nRkZMlT1H9/KEatLgmM/fpVT51U9/C3p59dvDxNRERV0q9fyWt1auMXYYio5jE0EhFRlRkYAN261XYVRFQTeE8jEREREUliaCQiIiIiSQyNRERERCSJoZGIiIiIJDE0EhEREZEkhkYiIiIiksTQSERERESSGBqJiIiqmYODA8LCwmq7DKJqxdBIRETPlaCgIMhkMixYsEBn/datWyH7549h69natWthaWlZav2RI0cwZsyYGq2FSN8YGomIqMqKNcWIvRKL/536H2KvxKJYU1wjxzU2NsbChQtx//79GjleZVlbW6NevXq1XQZRtWJoJCKiKolMiYTD1w7ovq47hkYORfd13eHwtQMiUyL1fmwfHx/Y2toiNDS03DYHDhxAly5doFKpoFarMXHiROTm5mq3p6Wl4a233oJKpYKjoyM2bNhQ6rLyl19+CTc3N5iYmECtVuODDz5ATk4OACA2NhYjR45EVlYWZDIZZDIZ5s2bB0D38vTQoUMxePBgndoKCwthZWWFiIgIAIBGo0FoaCgcHR2hUqng7u6OzZs3V8NIEVUfhkYiIqq0yJRIDPhhAK5nX9dZfyP7Bgb8MEDvwdHAwADz58/H0qVLcf369VLbL168CD8/P/Tv3x8nT57Epk2bcODAAYwfP17bZsSIEbh58yZiY2Px008/YdWqVbh165ZOP3K5HEuWLEFSUhLWrVuH33//HdOmTQMAeHl5ISwsDObm5khLS0NaWhpCQkJK1RIQEIBff/1VGzYBYNeuXcjLy0Pfvn0BAKGhoYiIiEB4eDiSkpIwefJkDBs2DHv37q2W8SKqDgyNRERUKcWaYkzaOQkCotS2x+uCdwbr/VJ137590aZNG8ydO7fUttDQUAQEBCA4OBguLi7w8vLCkiVLEBERgUePHuHMmTPYs2cPVq9eDQ8PD7z66qv473//i4cPH+r0ExwcjO7du8PBwQGvv/46PvvsM/zwww8AAIVCAQsLC8hkMtja2sLW1hampqalavH19YWJiQm2bNmiXbdhwwa88847MDMzQ35+PubPn49vv/0Wvr6+cHJyQlBQEIYNG4ZvvvmmmkeNqOoMa7sAIiJ6vuxP3V9qhvHvBASuZV/D/tT96ObQTa+1LFy4EK+//nqpGb4//vgDJ0+exPr16/+qSwhoNBpcvnwZ586dg6GhIV599VXtdmdnZ9SvX1+nnz179iA0NBRnzpxBdnY2ioqK8OjRI+Tl5VX4nkVDQ0MMGjQI69evx/Dhw5Gbm4uff/4ZGzduBABcuHABeXl5eOONN3T2KygoQNu2bSs1HkT6xNBIRESVkvYgrVrbPY2uXbvC19cXM2bMQFBQkHZ9Tk4O3nvvPUycOLHUPk2aNMG5c+ck+75y5QrefvttjB07Fv/5z3/QoEEDHDhwAKNGjUJBQUGlHnQJCAiAt7c3bt26haioKKhUKvj5+WlrBYDt27ejUaNGOvsplcoKH4NI3xgaiYioUuzM7Kq13dNasGAB2rRpg+bNm2vXvfrqq0hOToazs3OZ+zRv3hxFRUU4ceIE2rVrB6Bkxu/vT2MfO3YMGo0GixcvhlxecjfX40vTjykUChQXS1+G9/LyglqtxqZNm7Bjxw4MHDgQRkZGAICWLVtCqVQiNTUV3t7elTt5ohrE0EhERJXSpUkXNDZvjBvZN8q8r1EGGRqbN0aXJl1qpB43NzcEBARgyZIl2nUffvghOnXqhPHjx+Pf//43TExMkJycjKioKCxbtgwtWrSAj48PxowZg5UrV8LIyAhTp06FSqXSvuvR2dkZhYWFWLp0KXr37o24uDiEh4frHNvBwQE5OTmIjo6Gu7s76tWrV+4M5NChQxEeHo5z584hJiZGu97MzAwhISGYPHkyNBoNXnvtNWRlZSEuLg7m5uYIDAzUw6gRVR4fhCEiokoxkBvga7+vAZQExL97/DnMLwwGcoMaq+mTTz6BRqPRfm7dujX27t2Lc+fOoUuXLmjbti3mzJkDe3t7bZuIiAjY2Niga9eu6Nu3L0aPHg0zMzMYGxsDANzd3fHll19i4cKFeOWVV7B+/fpSr/jx8vLC+++/j8GDB8Pa2hqff/55uTUGBAQgOTkZjRo1QufOnXW2ffrpp5g9ezZCQ0Ph6uoKPz8/bN++HY6OjtUxPETVQiaEKP3PxBdcdnY2LCwskJWVBXNz89ouh4ioxj169AiXL1+Go6OjNiRVVmRKJCbtnKTzUIzaXI0wvzD0c+1XXaXWmOvXr0OtVmPPnj3o0aNHbZdTZz3p7ya/v2sXL08TEVGV9HPtB//m/tifuh9pD9JgZ2aHLk261OgM49P4/fffkZOTAzc3N6SlpWHatGlwcHBA165da7s0omcSQyMREVWZgdxA76/V0ZfCwkLMnDkTly5dgpmZGby8vLB+/XrtAypEpIuhkYiI6iRfX1/4+vrWdhlEzw0+CENEREREkhgaiYiIiEgSQyMRERERSWJoJCIiIiJJNRIaly9fDgcHBxgbG8PDwwOHDx9+Yvsff/wRLVq0gLGxMdzc3PDbb7/pbBdCYM6cObCzs4NKpYKPjw/Onz+vz1MgIiIiqtP0Hho3bdqEKVOmYO7cuTh+/Djc3d3h6+uLW7duldn+4MGDGDJkCEaNGoUTJ06gT58+6NOnD06fPq1t8/nnn2PJkiUIDw9HQkICTExM4Ovri0ePHun7dIiIiIjqJL3/IoyHhwc6dOiAZcuWAQA0Gg3UajUmTJiA6dOnl2o/ePBg5ObmYtu2bdp1nTp1Qps2bRAeHg4hBOzt7TF16lSEhIQAALKysmBjY4O1a9fi3XfflayJb5QnorquOn4Rpq6SyWTYsmUL+vTpU9ulvJD4izDPLr3ONBYUFODYsWPw8fH564ByOXx8fBAfH1/mPvHx8TrtgZJ3aT1uf/nyZaSnp+u0sbCwgIeHR7l95ufnIzs7W2chIqLnU3p6OiZNmgRnZ2cYGxvDxsYGnTt3xsqVK5GXl1fb5SEoKAgymQwymQxGRkZwdHTEtGnTqu1q2JUrVyCTyZCYmFgt/dWU57Vu+oteX+59584dFBcXw8bGRme9jY0Nzpw5U+Y+6enpZbZPT0/Xbn+8rrw2/xQaGoqPP/64SudARERPUFwM7N8PpKUBdnZAly6Agf5+RvDSpUvo3LkzLC0tMX/+fLi5uUGpVOLUqVNYtWoVGjVqhHfeeUdvx68oPz8/rFmzBoWFhTh27BgCAwMhk8mwcOHC2i6tUoQQKC4uhqEhfwuE6sjT0zNmzEBWVpZ2uXbtWm2XRET0/IuMBBwcgO7dgaFDS/7r4FCyXk8++OADGBoa4ujRoxg0aBBcXV3h5OQEf39/bN++Hb1799a2TU1Nhb+/P0xNTWFubo5BgwYhIyNDp7+VK1fi5ZdfhkKhQPPmzfHdd9/pbD9//jy6du0KY2NjtGzZElFRURWqU6lUwtbWFmq1Gn369IGPj4/OvhqNBqGhoXB0dIRKpYK7uzs2b96s3X7//n0EBATA2toaKpUKLi4uWLNmDQDA0dERANC2bVvIZDJ069YNAHDkyBG88cYbsLKygoWFBby9vXH8+HFtn2XN9GVmZkImkyE2NhYAEBsbC5lMhh07dqBdu3ZQKpU4cOAALl68CH9/f9jY2MDU1BQdOnTAnj17dM7ZwcEB8+fPx7/+9S+YmZmhSZMmWLVqlXZ7eXXT80OvodHKygoGBgal/keakZEBW1vbMvextbV9YvvH/61Mn0qlEubm5joLERE9hchIYMAA4Pp13fU3bpSs10NwvHv3Lnbv3o1x48bBxMSkzDYymQxASSjz9/fHvXv3sHfvXkRFReHSpUsYPHiwtu2WLVswadIkTJ06FadPn8Z7772HkSNHIiYmRttHv379oFAokJCQgPDwcHz44YeVrvv06dM4ePAgFAqFdl1oaCgiIiIQHh6OpKQkTJ48GcOGDcPevXsBALNnz0ZycjJ27NiBlJQUrFy5ElZWVgCgfQPJnj17kJaWhsg/x/rBgwcIDAzEgQMHcOjQIbi4uKBXr1548OBBpWuePn06FixYgJSUFLRu3Ro5OTno1asXoqOjceLECfj5+aF3795ITU3V2W/x4sVo3749Tpw4gQ8++ABjx47F2bNnn1g3PUeEnnXs2FGMHz9e+7m4uFg0atRIhIaGltl+0KBB4u2339ZZ5+npKd577z0hhBAajUbY2tqKRYsWabdnZWUJpVIp/ve//1WopqysLAFAZGVlVfZ0iIheCA8fPhTJycni4cOHld+5qEiIxo2FAMpeZDIh1OqSdtXo0KFDAoCIjIzUWd+wYUNhYmIiTExMxLRp04QQQuzevVsYGBiI1NRUbbukpCQBQBw+fFgIIYSXl5cYPXq0Tl8DBw4UvXr1EkIIsWvXLmFoaChu3Lih3b5jxw4BQGzZsqXcOgMDA4WBgYEwMTERSqVSABByuVxs3rxZCCHEo0ePRL169cTBgwd19hs1apQYMmSIEEKI3r17i5EjR5bZ/+XLlwUAceLEiXJrEKLk+9bMzEz8+uuv5e53//59AUDExMQIIYSIiYkRAMTWrVuf2LcQQrRq1UosXbpU+7lp06Zi2LBh2s8ajUa89NJLYuXKlZWq+0l/N/n9Xbv0fnl6ypQpWL16NdatW4eUlBSMHTsWubm5GDlyJABgxIgRmDFjhrb9pEmTsHPnTixevBhnzpzBvHnzcPToUYwfPx5Ayb8ig4OD8dlnn+GXX37BqVOnMGLECNjb2/NJNiKimrB/f+kZxr8TArh2raRdDTh8+DASExPRqlUr5OfnAwBSUlKgVquhVqu17Vq2bAlLS0ukpKRo23Tu3Fmnr86dO+tsV6vVsLe312739PSsUE3du3dHYmIiEhISEBgYiJEjR6J///4AgAsXLiAvLw9vvPEGTE1NtUtERAQuXrwIABg7diw2btyINm3aYNq0aTh48KDkMTMyMjB69Gi4uLjAwsIC5ubmyMnJKTUbWBHt27fX+ZyTk4OQkBC4urrC0tISpqamSElJKdV369attX+WyWSwtbUt9xV79PzR+52tgwcPxu3btzFnzhykp6ejTZs22Llzp/ZBltTUVMjlf2VXLy8vbNiwAbNmzcLMmTPh4uKCrVu34pVXXtG2mTZtGnJzczFmzBhkZmbitddew86dO/naCCKimpCWVr3tKsjZ2RkymUx7ufMxJycnAIBKparW4z0NExMTODs7AwC+/fZbuLu74//+7/8watQo5OTkAAC2b9+ORo0a6eynVCoBAD179sTVq1fx22+/ISoqCj169MC4ceOwaNGico8ZGBiIu3fv4uuvv0bTpk2hVCrh6emJgoICANB+14q/vWmvsLCw3Pr/LiQkBFFRUVi0aBGcnZ2hUqkwYMAAbd+PGRkZ6XyWyWTQaDTl1kzPlxp5EGb8+PG4evUq8vPzkZCQAA8PD+222NhYrF27Vqf9wIEDcfbsWeTn5+P06dPo1auXznaZTIZPPvkE6enpePToEfbs2YNmzZrVxKkQEZGdXfW2q6CGDRvijTfewLJly5Cbm/vEtq6urrh27ZrOg4/JycnIzMxEy5YttW3i4uJ09ouLi9PZfu3aNaT9LfweOnSo0nXL5XLMnDkTs2bNwsOHD9GyZUsolUqkpqbC2dlZZ/n7zKi1tTUCAwPx/fffIywsTPtQyeN7I4uLi0vVPnHiRPTq1QutWrWCUqnEnTt3dPoDoHM+FX39TVxcHIKCgtC3b1+4ubnB1tYWV65cqdQ4lFc3PT/qxNPTRERUjbp0ARo3Bv586KQUmQxQq0vaVbMVK1agqKgI7du3x6ZNm5CSkoKzZ8/i+++/x5kzZ2Dw5+t+fHx84ObmhoCAABw/fhyHDx/GiBEj4O3trb30+v/+3//D2rVrsXLlSpw/fx5ffvklIiMjtT8c4ePjg2bNmiEwMBB//PEH9u/fj48++qhKdQ8cOBAGBgZYvnw5zMzMEBISgsmTJ2PdunW4ePEijh8/jqVLl2LdunUAgDlz5uDnn3/GhQsXkJSUhG3btsHV1RUA8NJLL0GlUmHnzp3IyMhAVlYWAMDFxQXfffcdUlJSkJCQgICAAJ3ZV5VKhU6dOmkfcNm7dy9mzZpVofpdXFwQGRmJxMRE/PHHHxg6dGilZxDLq5ueI7V9U2Vt4I20RFTXPdWDMEII8dNPJQ+8yGSlH4KRyUq268nNmzfF+PHjhaOjozAyMhKmpqaiY8eO4osvvhC5ubnadlevXhXvvPOOMDExEWZmZmLgwIEiPT1dp68VK1YIJycnYWRkJJo1ayYiIiJ0tp89e1a89tprQqFQiGbNmomdO3dW6EEYf3//UutDQ0OFtbW1yMnJERqNRoSFhYnmzZsLIyMjYW1tLXx9fcXevXuFEEJ8+umnwtXVVahUKtGgQQPh7+8vLl26pO1r9erVQq1WC7lcLry9vYUQQhw/fly0b99eGBsbCxcXF/Hjjz+Kpk2biq+++kq7X3JysvD09BQqlUq0adNG7N69u8wHYe7fv69T++XLl0X37t2FSqUSarVaLFu2THh7e4tJkyZp2/zzWEII4e7uLubOnfvEuv+JD8I8u/T+M4LPIv4MERHVddXyM4KRkcCkSboPxajVQFgY0K9ftdRJdQ9/RvDZxVe8ExFR1fTrB/j71+gvwhBR7WFoJCKiqjMwAPjLHkR1Ah+EISIiIiJJDI1EREREJImhkYiIiIgkMTQSERERkSSGRiIiIiKSxNBIRERERJIYGomIiIhIEkMjERERgHnz5qFNmza1XQbRM4uhkYiInhtBQUGQyWTapWHDhvDz88PJkyf1fuwrV67oHLtBgwbw9vbG/v37q+0YQUFB6NOnT7X1V1Oe17qpchgaiYioyoqFQOz9+/hfRgZi799HsRB6P6afnx/S0tKQlpaG6OhoGBoa4u2339b7cR/bs2cP0tLSsG/fPtjb2+Ptt99GRkZGjR2/uhQUFNR2CfScYWgkIqIqibx9Gw6HDqH7H39gaEoKuv/xBxwOHULk7dt6Pa5SqYStrS1sbW3Rpk0bTJ8+HdeuXcPtvx33ww8/RLNmzVCvXj04OTlh9uzZKCws1OlnwYIFsLGxgZmZGUaNGoVHjx5V6PgNGzaEra0tXnnlFcycORPZ2dlISEjQbj99+jR69uwJU1NT2NjYYPjw4bhz5452++bNm+Hm5gaVSoWGDRvCx8cHubm5mDdvHtatW4eff/5ZO5sZGxtbofMpa6YvODgY3f72E4/dunXD+PHjERwcDCsrK/j6+gIAvvzyS7i5ucHExARqtRoffPABcnJytPutXbsWlpaW2LVrF1xdXWFqaqoN7gCeWDe9WBgaiYio0iJv38aApCRcz8/XWX8jPx8DkpL0Hhwfy8nJwffffw9nZ2c0bNhQu97MzAxr165FcnIyvv76a6xevRpfffWVdvsPP/yAefPmYf78+Th69Cjs7OywYsWKSh374cOHiIiIAAAoFAoAQGZmJl5//XW0bdsWR48exc6dO5GRkYFBgwYBANLS0jBkyBD861//QkpKCmJjY9GvXz8IIRASEoJBgwbpzKR6eXlV6Hwqat26dVAoFIiLi0N4eDgAQC6XY8mSJUhKSsK6devw+++/Y9q0aTr75eXlYdGiRfjuu++wb98+pKamIiQkBACeWDe9WAxruwAiInq+FAuBSRcuoKwL0QKADEDwhQvwt7KCgUxW7cfftm0bTE1NAQC5ubmws7PDtm3bIJf/NQ8ya9Ys7Z8dHBwQEhKCjRs3asNQWFgYRo0ahVGjRgEAPvvsM+zZs6dCs41eXl6Qy+XIy8uDEALt2rVDjx49AADLli1D27ZtMX/+fG37b7/9Fmq1GufOnUNOTg6KiorQr18/NG3aFADg5uambatSqZCfnw9bW1udY0qdT0W5uLjg888/11kXHBys0/dnn32G999/XydEFxYWIjw8HC+//DIAYPz48fjkk08AAKampuXWTS8WzjQSEVGl7M/MLDXD+HcCwLX8fOzPzNTL8bt3747ExEQkJibi8OHD8PX1Rc+ePXH16lVtm02bNqFz586wtbWFqakpZs2ahdTUVO32lJQUeHh46PTr6elZoeNv2rQJJ06cwE8//QRnZ2esXbsWRkZGAIA//vgDMTExMDU11S4tWrQAAFy8eBHu7u7o0aMH3NzcMHDgQKxevRr379+v0DGfdD4V1a5du1Lr9uzZgx49eqBRo0YwMzPD8OHDcffuXeTl5Wnb1KtXTxsYAcDOzg63bt2q9PHp+cbQSERElZJWwQcoKtquskxMTODs7AxnZ2d06NAB//3vf5Gbm4vVq1cDAOLj4xEQEIBevXph27ZtOHHiBD766KNqe/BDrVbDxcUFffv2xfz589G3b1/k/xmic3Jy0Lt3b22ofbycP38eXbt2hYGBAaKiorBjxw60bNkSS5cuRfPmzXH58uVyj1eR85HL5RD/eAjpn/dwAiVj93dXrlzB22+/jdatW+Onn37CsWPHsHz5cgC6D8o8DsWPyWSyUsejFx9DIxERVYrdn/fvVVe7pyWTySCXy/Hw4UMAwMGDB9G0aVN89NFHaN++PVxcXHRmIQHA1dVV5+EVADh06FCljz1gwAAYGhpqL+W++uqrSEpKgoODgzbYPl4eBzaZTIbOnTvj448/xokTJ6BQKLBlyxYAJfdGFhcX6xyjIudjbW2tfTDlscTERMn6jx07Bo1Gg8WLF6NTp05o1qwZbt68WelxKKtuevEwNBIRUaV0sbREY6US5d2tKAOgVirRxdJSL8fPz89Heno60tPTkZKSggkTJmhn+ICS+/ZSU1OxceNGXLx4EUuWLNGGsscmTZqEb7/9FmvWrMG5c+cwd+5cJCUlVboWmUyGiRMnYsGCBcjLy8O4ceNw7949DBkyBEeOHMHFixexa9cujBw5EsXFxUhISNA+fJOamorIyEjcvn0brq6uAEruKTx58iTOnj2LO3fuoLCwsELn8/rrr+Po0aOIiIjA+fPnMXfuXJw+fVqyfmdnZxQWFmLp0qW4dOkSvvvuO+0DMpVRVt304mFoJCKiSjGQyfC1szMAlAqOjz+HOTvr5SEYANi5cyfs7OxgZ2cHDw8PHDlyBD/++KP29TLvvPMOJk+ejPHjx6NNmzY4ePAgZs+erdPH4MGDMXv2bEybNg3t2rXD1atXMXbs2CrVExgYiMLCQixbtgz29vaIi4tDcXEx3nzzTbi5uSE4OBiWlpaQy+UwNzfHvn370KtXLzRr1gyzZs3C4sWL0bNnTwDA6NGj0bx5c7Rv3x7W1taIi4ur0Pn4+vpqz6dDhw548OABRowYIVm7u7s7vvzySyxcuBCvvPIK1q9fj9DQ0EqPQVl104tHJurgTQnZ2dmwsLBAVlYWzM3Na7scIqIa9+jRI1y+fBmOjo4wNjauUh+Rt29j0oULOg/FqJVKhDk7o5+1dXWVSnXMk/5u8vu7dvGVO0REVCX9rK3hb2WF/ZmZSCsogJ1CgS6WlnqbYSSi2sXQSEREVWYgk6Fb/fq1XQYR1QDe00hEREREkhgaiYiIiEgSQyMRERERSWJoJCKqwzQaTW2XQKSDfyefXXwQhoioDlIoFJDL5bh58yasra2hUCgg41PPVIuEECgoKMDt27chl8uhqKFfFKKKY2gkIqqD5HI5HB0dkZaWVqWfjSPSl3r16qFJkyaQy3kx9FnD0EhEVEcpFAo0adIERUVF/N1geiYYGBjA0NCQs97PKIZGIqI6TCaTwcjICEZGRrVdChE94zj3S0RERESSGBqJiIiISBJDIxERERFJYmgkIiIiIkkMjUREREQkiaGRiIiIiCTpNTTeu3cPAQEBMDc3h6WlJUaNGoWcnJwn7vPo0SOMGzcODRs2hKmpKfr374+MjAydNjKZrNSyceNGfZ4KERERUZ2m19AYEBCApKQkREVFYdu2bdi3bx/GjBnzxH0mT56MX3/9FT/++CP27t2Lmzdvol+/fqXarVmzBmlpadqlT58+ejoLIiIiIpIJIYQ+Ok5JSUHLli1x5MgRtG/fHgCwc+dO9OrVC9evX4e9vX2pfbKysmBtbY0NGzZgwIABAIAzZ87A1dUV8fHx6NSpU0nRMhm2bNlS5aCYnZ0NCwsLZGVlwdzcvGonSERERDWK39+1S28zjfHx8bC0tNQGRgDw8fGBXC5HQkJCmfscO3YMhYWF8PHx0a5r0aIFmjRpgvj4eJ2248aNg5WVFTp27Ihvv/0Wesq+RERERAQ9/oxgeno6XnrpJd2DGRqiQYMGSE9PL3cfhUIBS0tLnfU2NjY6+3zyySd4/fXXUa9ePezevRsffPABcnJyMHHixDL7zc/PR35+vvZzdnZ2Fc+KiIiIqG6qdGicPn06Fi5c+MQ2KSkpVS6oImbPnq39c9u2bZGbm4svvvii3NAYGhqKjz/+WK81EREREb3IKh0ap06diqCgoCe2cXJygq2tLW7duqWzvqioCPfu3YOtrW2Z+9na2qKgoACZmZk6s40ZGRnl7gMAHh4e+PTTT5Gfnw+lUllq+4wZMzBlyhTt5+zsbKjV6ieeAxERERH9pdKh0draGtbW1pLtPD09kZmZiWPHjqFdu3YAgN9//x0ajQYeHh5l7tOuXTsYGRkhOjoa/fv3BwCcPXsWqamp8PT0LPdYiYmJqF+/fpmBEQCUSmW524iIiIhImt7uaXR1dYWfnx9Gjx6N8PBwFBYWYvz48Xj33Xe1T07fuHEDPXr0QEREBDp27AgLCwuMGjUKU6ZMQYMGDWBubo4JEybA09NT++T0r7/+ioyMDHTq1AnGxsaIiorC/PnzERISoq9TISIiIqrz9BYaAWD9+vUYP348evToAblcjv79+2PJkiXa7YWFhTh79izy8vK067766itt2/z8fPj6+mLFihXa7UZGRli+fDkmT54MIQScnZ3x5ZdfYvTo0fo8FSIiIqI6TW/vaXyW8T1PREREzx9+f9cu/vY0EREREUliaCQiIiIiSQyNRERERCSJoZGIiIiIJDE0EhEREZEkhkYiIiIiksTQSERERESSGBqJiIiISBJDIxERERFJYmgkIiIiIkkMjUREREQkiaGRiIiIiCQxNBIRERGRJIZGIiIiIpLE0EhEREREkhgaiYiIiEgSQyMRERERSWJoJCIiIiJJDI1EREREJImhkYiIiIgkMTQSERERkSSGRiIiIiKSxNBIRERERJIYGomIiIhIEkMjEREREUliaCQiIiIiSQyNRERERCSJoZGIiIiIJDE0EhEREZEkhkYiIiIiksTQSERERESSGBqJiIiISBJDIxERERFJYmgkIiIiIkkMjUREREQkiaGRiIiIiCQxNBIRERGRJIZGIiIiIpLE0EhEREREkhgaiYiIiEgSQyMRERERSdJbaLx37x4CAgJgbm4OS0tLjBo1Cjk5OU/cZ9WqVejWrRvMzc0hk8mQmZlZLf0SERER0dPRW2gMCAhAUlISoqKisG3bNuzbtw9jxox54j55eXnw8/PDzJkzq7VfIiIiIno6MiGEqO5OU1JS0LJlSxw5cgTt27cHAOzcuRO9evXC9evXYW9v/8T9Y2Nj0b17d9y/fx+WlpbV1u9j2dnZsLCwQFZWFszNzat2kkRERFSj+P1du/Qy0xgfHw9LS0ttsAMAHx8fyOVyJCQk1Hi/+fn5yM7O1lmIiIiIqOL0EhrT09Px0ksv6awzNDREgwYNkJ6eXuP9hoaGwsLCQruo1eoq10BERERUF1UqNE6fPh0ymeyJy5kzZ/RVa5XNmDEDWVlZ2uXatWu1XRIRERHRc8WwMo2nTp2KoKCgJ7ZxcnKCra0tbt26pbO+qKgI9+7dg62tbaWLfKyq/SqVSiiVyiofl4iIiKiuq1RotLa2hrW1tWQ7T09PZGZm4tixY2jXrh0A4Pfff4dGo4GHh0fVKtVjv0RERET0ZHq5p9HV1RV+fn4YPXo0Dh8+jLi4OIwfPx7vvvuu9gnnGzduoEWLFjh8+LB2v/T0dCQmJuLChQsAgFOnTiExMRH37t2rcL9EREREVP309p7G9evXo0WLFujRowd69eqF1157DatWrdJuLywsxNmzZ5GXl6ddFx4ejrZt22L06NEAgK5du6Jt27b45ZdfKtwvEREREVU/vbyn8VnH9zwRERE9f/j9Xbv429NEREREJImhkYiIiIgkMTQSERERkSSGRiIiIiKSxNBIRERERJIYGomIiIhIEkMjEREREUliaCQiIiIiSQyNRERERCSJoZGIiIiIJDE0EhEREZEkhkYiIiIiksTQSERERESSGBqJiIiISBJDIxERERFJYmgkIiIiIkkMjUREREQkiaGRiIiIiCQxNBIRERGRJIZGIiIiIpLE0EhEREREkhgaiYiIiEgSQyMRERERSWJoJCIiIiJJDI1EREREJImhkYiIiIgkMTQSERERkSSGRiIiIiKSxNBIRERERJIYGomIiIhIEkMjEREREUliaCQiIiIiSQyNRERERCSJoZGIiIiIJDE0EhEREZEkhkYiIiIiksTQSERERESSGBqJiIiISBJDIxERERFJYmiswxwcHBAWFlbbZRAREdFzQG+h8d69ewgICIC5uTksLS0xatQo5OTkPHGfVatWoVu3bjA3N4dMJkNmZmapNg4ODpDJZDrLggUL9HQWVRcUFFRmbVu3boVMJqvRWtauXQtLS8tS648cOYIxY8bUaC1ERET0fNJbaAwICEBSUhKioqKwbds27Nu3TzKg5OXlwc/PDzNnznxiu08++QRpaWnaZcKECdVZerUxNjbGwoULcf/+/doupUzW1taoV69ebZdBREREzwG9hMaUlBTs3LkT//3vf+Hh4YHXXnsNS5cuxcaNG3Hz5s1y9wsODsb06dPRqVOnJ/ZvZmYGW1tb7WJiYlLdp1AtfHx8YGtri9DQ0HLbHDhwAF26dIFKpYJarcbEiRORm5ur3Z6Wloa33noLKpUKjo6O2LBhQ6nLyl9++SXc3NxgYmICtVqNDz74QDurGxsbi5EjRyIrK0s7Mztv3jwAupenhw4disGDB+vUVlhYCCsrK0RERAAANBoNQkND4ejoCJVKBXd3d2zevLkaRoqIiKhy5s2bhzZt2tR2GXWKXkJjfHw8LC0t0b59e+06Hx8fyOVyJCQkPHX/CxYsQMOGDdG2bVt88cUXKCoqemL7/Px8ZGdn6yw1wcDAAPPnz8fSpUtx/fr1UtsvXrwIPz8/9O/fHydPnsSmTZtw4MABjB8/XttmxIgRuHnzJmJjY/HTTz9h1apVuHXrlk4/crkcS5YsQVJSEtatW4fff/8d06ZNAwB4eXkhLCwM5ubm2pnZkJCQUrUEBATg119/1bmFYNeuXcjLy0Pfvn0BAKGhoYiIiEB4eDiSkpIwefJkDBs2DHv37q2W8SIiohfL41u1Hi8NGzaEn58fTp48qfdjX7lyRefYDRo0gLe3N/bv319txwgKCkKfPn2qrb+aUtW69RIa09PT8dJLL+msMzQ0RIMGDZCenv5UfU+cOBEbN25ETEwM3nvvPcyfP18bkMoTGhoKCwsL7aJWq5+qhsro27cv2rRpg7lz55ZZV0BAAIKDg+Hi4gIvLy8sWbIEERERePToEc6cOYM9e/Zg9erV8PDwwKuvvor//ve/ePjwoU4/wcHB6N69OxwcHPD666/js88+ww8//AAAUCgUsLCwgEwm087MmpqalqrF19cXJiYm2LJli3bdhg0b8M4778DMzAz5+fmYP38+vv32W/j6+sLJyQlBQUEYNmwYvvnmm2oeNSIielH4+flpJy2io6NhaGiIt99+u8aOv2fPHqSlpWHfvn2wt7fH22+/jYyMjBo7fnUpKCio7RIqFxqnT59e6iGUfy5nzpzRV60AgClTpqBbt25o3bo13n//fSxevBhLly5Ffn5+ufvMmDEDWVlZ2uXatWt6qa1YU4zYK7H436n/IT0nHUIIAMDChQuxbt06pKSk6LT/448/sHbtWpiammoXX19faDQaXL58GWfPnoWhoSFeffVV7T7Ozs6oX7++Tj979uxBjx490KhRI5iZmWH48OG4e/cu8vLyKly7oaEhBg0ahPXr1wMAcnNz8fPPPyMgIAAAcOHCBeTl5eGNN97QqTciIgIXL16s0ngREdGLT6lUaict2rRpg+nTp+PatWu4ffu2ts2HH36IZs2aoV69enBycsLs2bNRWFio08+CBQvg7OwMABg3bhwePXpUoeM3bNgQtra2eOWVVzBz5kxkZ2frXPU8ffo0evbsCVNTU9jY2GD48OG4c+eOdvvmzZvh5uYGlUqFhg0bwsfHB7m5uZg3bx7WrVuHn3/+WZuBYmNjK3Q+Zc30BQcHo1u3btrP3bp1w/jx4xEcHAwrKyv4+voCePItacBfD7/u2rULrq6uMDU11QZ3AE+sW4phhVr9aerUqQgKCnpiGycnJ9ja2pa6hFpUVIR79+7B1ta2MoeU5OHhgaKiIly5cgXNmzcvs41SqYRSqazW4/5TZEokJu2chOvZf16GvgAYFxsjMiUS/br2g6+vL2bMmKEzfjk5OXjvvfcwceLEUv01adIE586dkzzulStX8Pbbb2Ps2LH4z3/+gwYNGuDAgQMYNWoUCgoKKvWgS0BAALy9vXHr1i1ERUVBpVLBz89PWysAbN++HY0aNdLZT99jS0REL4acnBx8//33cHZ2RsOGDbXrzczMsHbtWtjb2+PUqVMYPXo0zMzMtFcSf/jhB8ybNw+LFi3ChAkTYGtrixUrVsDJyanCx3748KH2Hn2FQgEAyMzMxOuvv45///vf+Oqrr/Dw4UN8+OGHGDRoEH7//XekpaVhyJAh+Pzzz9G3b188ePAA+/fvhxACISEhSElJQXZ2NtasWQMAaNCgQYXOp6LWrVuHsWPHIi4uTrvu8S1pjo6OuHTpEj744ANMmzYNK1as0LbJy8vDokWL8N1330Eul2PYsGEICQnB+vXrn1i3JKEHycnJAoA4evSodt2uXbuETCYTN27ckNw/JiZGABD379+XbPv9998LuVwu7t27V+H6srKyBACRlZVV4X2e5Kfkn4RsnkxgHv5a3CHQHEI2TyZ+Sv5JnDx5UsjlcjFt2jTxeNiHDh0qevToUW6/KSkppcbx/PnzAoD46quvhBBCbN68WRgZGYni4mJtm08//VRn/NavXy9MTU1L9d+0aVNtP485OjqKJUuWiJ49e4r3339fuz47O1solUoRERFR2eEhIqI6pEijETH37okN6enCd8gQYWBgIExMTISJiYkAIOzs7MSxY8ee2McXX3wh2rVrp/3s6ekpPvjgA53vbw8PD+Hu7l5uH5cvXxYAhEqlEiYmJkImkwkAol27dqKgoEAIUfJ9+eabb+rsd+3aNQFAnD17Vhw7dkwAEFeuXCnzGIGBgcLf319yTP55PmXtN2nSJOHt7a397O3tLdq2bSvZ948//igaNmyo/bxmzRoBQFy4cEG7bvny5cLGxqbSdf9TpWYaK8rV1RV+fn4YPXo0wsPDUVhYiPHjx+Pdd9+Fvb09AODGjRvo0aMHIiIi0LFjRwAl90Kmp6fjwoULAIBTp07BzMwMTZo0QYMGDRAfH4+EhAR0794dZmZmiI+P1z6M8c9LtjWlWFOMSTsnQUCU2yZ4ZzAuT7qMgIAALFmyRLv+ww8/RKdOnTB+/Hj8+9//homJCZKTkxEVFYVly5ahRYsW8PHxwZgxY7By5UoYGRlh6tSpUKlU2nc9Ojs7o7CwEEuXLkXv3r0RFxeH8PBwneM7ODggJycH0dHRcHd3R7169cqdgRw6dCjCw8Nx7tw5xMTEaNebmZkhJCQEkydPhkajwWuvvYasrCzExcXB3NwcgYGBTzOMRET0Aoi8fRuTLlzA9ce3jN27B+Wrr2Lh0qXwbdgQ9+/fx4oVK9CzZ08cPnwYTZs2BQBs2rQJS5YswcWLF5GTk4OioiKYm5tr+01JScH777+vcyxPT0+d76nybNq0CS1atMDp06cxbdo0rF27FkZGRgBKbhOLiYkp817/ixcv4s0330SPHj3g5uYGX19fvPnmmxgwYIBk5pA6n4pq165dqXV79uxBaGgozpw5g+zsbBQVFeHRo0fIy8vTfrfXq1cPL7/8snYfOzu7UleAq0Jv72lcv349WrRogR49eqBXr1547bXXsGrVKu32wsJCnD17Vue+u/DwcLRt2xajR48GAHTt2hVt27bFL7/8AqDkMujGjRvh7e2NVq1a4T//+Q8mT56s029N25+6/69L0mUQELiWfQ37U/fjk08+gUaj0W5r3bo19u7di3PnzqFLly5o27Yt5syZow3WABAREQEbGxt07doVffv21U5xGxsbAwDc3d3x5ZdfYuHChXjllVewfv36Uq/48fLywvvvv4/BgwfD2toan3/+ebn1BgQEIDk5GY0aNULnzp11tn366aeYPXs2QkNDtf8w2L59OxwdHSs1ZkRE9OKJvH0bA5KS/gqMf8pXKDDh4UOctLBAhw4d8N///he5ublYvXo1gJI3rgQEBKBXr17Ytm0bTpw4gY8++qjaHvxQq9VwcXFB3759MX/+fPTt21f7HEROTg569+6NxMREneX8+fPo2rUrDAwMEBUVhR07dqBly5ZYunQpmjdvjsuXL5d7vIqcj1wu1z738Ng/7+EEUOqVgo9vSWvdujV++uknHDt2DMuXLweg+6DM41D8mEwmK3W8qtDLTCNQcn18w4YN5W53cHAodQLz5s3TvkOwLK+++ioOHTpUXSVWi7QHaWVv6Fu6XTe3bqUe2OnQoQN2795dbv92dnb47bfftJ+vX7+OW7duaW8GBoDJkydj8uTJOvsNHz5c5/PKlSuxcuVKnXVXrlwpdTxXV9dy/2LJZDJMmjQJkyZNKrdeIiKqe4qFwKQLF55wzQ0IvnAB/lZWkMlkkMvl2jeBHDx4EE2bNsVHH32kbXv16lWdfV1dXZGQkKDz8EhV8sCAAQMwZ84crFixApMnT8arr76Kn376CQ4ODjA0LDsSyWQydO7cGZ07d8acOXPQtGlTbNmyBVOmTIFCoUBxcbFO+4qcj7W1NU6fPq2zLjExsVTY+6djx45Bo9Fg8eLFkMtL5v0evy2lMsqquyL429NPyc7Mrlrb/dPvv/+OX375BZcvX8bBgwfx7rvvwsHBAV27dq1Sf0RERNVtf2ZmqRlGrcJCiHv3cC0tDd8nJGDChAnaGT4AcHFxQWpqKjZu3IiLFy9iyZIlOq9/A4BJkybh22+/xffffw8AmD9/PpKSkipdp0wmw8SJE7FgwQLk5eVh3LhxuHfvHoYMGYIjR47g4sWL2LVrF0aOHIni4mIkJCRg/vz5OHr0KFJTUxEZGYnbt2/D1dUVQMkE2MmTJ3H27FncuXMHhYWFFTqf119/HUePHkVERATOnz+PuXPnlgqRZfn7LWmXLl3Cd999V+qWtIooq+6KYGh8Sl2adEFj88aQoezfk5ZBBrW5Gl2adKlS/4WFhZg5cyZatWqFvn37wtraGrGxsZL/GiEiIqopaU+6lHz4MNC/P9C/P8a+8QaOHDmCH3/8Uft6mXfeeQeTJ0/G+PHj0aZNGxw8eBCzZ8/W6WLw4MGYPXs25syZAwC4du0axo4dW6VaAwMDUVhYiGXLlsHe3h5xcXEoLi7Gm2++CTc3NwQHB8PS0hJyuRzm5ubYt28fevXqhWbNmmHWrFlYvHgxevbsCQAYPXo0mjdvjvbt28Pa2hpxcXEVOh9fX1/Mnj0b06ZNQ4cOHfDgwQOMGDFCsvaK3JJWEWXVXREyUR0XuZ8z2dnZsLCwQFZWVpVuTP2nyJRIDPhhAADoPBDzOEhuHrQZ/Vz7PfVxiIiInkWx9++j+x9/SLaLcXdHt6d4cLW6v7+pcjjTWA36ufbD5kGb0chc9/2Fjc0bMzASEdELr4ulJRorleVccwNkANRKJbpYWtZgVVTdONNYjf9SKdYUY3/qfqQ9SIOdmR26NOkCA7lBtfVPRET0rHr89DQAnQdiHgfJza1aoZ+19VMdgzONtUtvT0/XRQZyA3Rz6FbbZRAREdW4ftbW2Nyqle57GgE0VioR5uz81IGRah9DIxEREVWLftbW8Leywv7MTKQVFMBOoUAXS0sYyMq7cE3PE4ZGIiIiqjYGMtlTPexCzy4+CENEREREkhgaiYiIiEgSQyMRERERSWJoJCIiIiJJDI1EREREJImhkYiIiIgkMTQSERERkSSGRiIiIiKSxNBIRERERJLq5C/CCFHyU+rZ2dm1XAkRERFV1OPv7cff41Sz6mRofPDgAQBArVbXciVERERUWQ8ePICFhUVtl1HnyEQdjOsajQY3b96EmZkZZH/+iHp2djbUajWuXbsGc3PzWq7w+cKxqxqOW9Vx7KqG41Z1HLuqq86xE0LgwYMHsLe3h1zOO+xqWp2caZTL5WjcuHGZ28zNzfl/CFXEsasajlvVceyqhuNWdRy7qquuseMMY+1hTCciIiIiSQyNRERERCSJofFPSqUSc+fOhVKprO1Snjscu6rhuFUdx65qOG5Vx7GrOo7di6NOPghDRERERJXDmUYiIiIiksTQSERERESSGBqJiIiISBJDIxERERFJqtOh8d69ewgICIC5uTksLS0xatQo5OTkPHGf9957Dy+//DJUKhWsra3h7++PM2fO1FDFz4bKjtu9e/cwYcIENG/eHCqVCk2aNMHEiRORlZVVg1U/G6ryd27VqlXo1q0bzM3NIZPJkJmZWTPF1rLly5fDwcEBxsbG8PDwwOHDh5/Y/scff0SLFi1gbGwMNzc3/PbbbzVU6bOlMuOWlJSE/v37w8HBATKZDGFhYTVX6DOoMmO3evVqdOnSBfXr10f9+vXh4+Mj+Xf0RVaZsYuMjET79u1haWkJExMTtGnTBt99910NVktVVadDY0BAAJKSkhAVFYVt27Zh3759GDNmzBP3adeuHdasWYOUlBTs2rULQgi8+eabKC4urqGqa19lx+3mzZu4efMmFi1ahNOnT2Pt2rXYuXMnRo0aVYNVPxuq8ncuLy8Pfn5+mDlzZg1VWfs2bdqEKVOmYO7cuTh+/Djc3d3h6+uLW7duldn+4MGDGDJkCEaNGoUTJ06gT58+6NOnD06fPl3Dldeuyo5bXl4enJycsGDBAtja2tZwtc+Wyo5dbGwshgwZgpiYGMTHx0OtVuPNN9/EjRs3arjy2lfZsWvQoAE++ugjxMfH4+TJkxg5ciRGjhyJXbt21XDlVGmijkpOThYAxJEjR7TrduzYIWQymbhx40aF+/njjz8EAHHhwgV9lPnMqa5x++GHH4RCoRCFhYX6KPOZ9LRjFxMTIwCI+/fv67HKZ0PHjh3FuHHjtJ+Li4uFvb29CA0NLbP9oEGDxFtvvaWzzsPDQ7z33nt6rfNZU9lx+7umTZuKr776So/VPdueZuyEEKKoqEiYmZmJdevW6avEZ9bTjp0QQrRt21bMmjVLH+VRNaqzM43x8fGwtLRE+/bttet8fHwgl8uRkJBQoT5yc3OxZs0aODo6Qq1W66vUZ0p1jBsAZGVlwdzcHIaGdefnz6tr7F50BQUFOHbsGHx8fLTr5HI5fHx8EB8fX+Y+8fHxOu0BwNfXt9z2L6KqjBuVqI6xy8vLQ2FhIRo0aKCvMp9JTzt2QghER0fj7Nmz6Nq1qz5LpWpQZ0Njeno6XnrpJZ11hoaGaNCgAdLT05+474oVK2BqagpTU1Ps2LEDUVFRUCgU+iz3mfE04/bYnTt38Omnn0peln3RVMfY1QV37txBcXExbGxsdNbb2NiUO07p6emVav8iqsq4UYnqGLsPP/wQ9vb2pf7x8qKr6thlZWXB1NQUCoUCb731FpYuXYo33nhD3+XSU3rhQuP06dMhk8meuDztgysBAQE4ceIE9u7di2bNmmHQoEF49OhRNZ1B7aiJcQOA7OxsvPXWW2jZsiXmzZv39IU/A2pq7Ijo2bRgwQJs3LgRW7ZsgbGxcW2X81wwMzNDYmIijhw5gv/85z+YMmUKYmNja7sskvDCXRucOnUqgoKCntjGyckJtra2pW7SLSoqwr179yRvCLewsICFhQVcXFzQqVMn1K9fH1u2bMGQIUOetvxaUxPj9uDBA/j5+cHMzAxbtmyBkZHR05b9TKiJsatLrKysYGBggIyMDJ31GRkZ5Y6Tra1tpdq/iKoyblTiacZu0aJFWLBgAfbs2YPWrVvrs8xnUlXHTi6Xw9nZGQDQpk0bpKSkIDQ0FN26ddNnufSUXrjQaG1tDWtra8l2np6eyMzMxLFjx9CuXTsAwO+//w6NRgMPD48KH08IASEE8vPzq1zzs0Df45adnQ1fX18olUr88ssvL9S/xmv679yLTqFQoF27doiOjkafPn0AABqNBtHR0Rg/fnyZ+3h6eiI6OhrBwcHadVFRUfD09KyBip8NVRk3KlHVsfv888/xn//8B7t27dK5V7kuqa6/dxqN5rn/Hq0TavlBnFrl5+cn2rZtKxISEsSBAweEi4uLGDJkiHb79evXRfPmzUVCQoIQQoiLFy+K+fPni6NHj4qrV6+KuLg40bt3b9GgQQORkZFRW6dR4yo7bllZWcLDw0O4ubmJCxcuiLS0NO1SVFRUW6dRKyo7dkIIkZaWJk6cOCFWr14tAIh9+/aJEydOiLt379bGKdSIjRs3CqVSKdauXSuSk5PFmDFjhKWlpUhPTxdCCDF8+HAxffp0bfu4uDhhaGgoFi1aJFJSUsTcuXOFkZGROHXqVG2dQq2o7Ljl5+eLEydOiBMnTgg7OzsREhIiTpw4Ic6fP19bp1BrKjt2CxYsEAqFQmzevFnn/9MePHhQW6dQayo7dvPnzxe7d+8WFy9eFMnJyWLRokXC0NBQrF69urZOgSqoTofGu3fviiFDhghTU1Nhbm4uRo4cqfM/+MuXLwsAIiYmRgghxI0bN0TPnj3FSy+9JIyMjETjxo3F0KFDxZkzZ2rpDGpHZcft8atiylouX75cOydRSyo7dkIIMXfu3DLHbs2aNTV/AjVo6dKlokmTJkKhUIiOHTuKQ4cOabd5e3uLwMBAnfY//PCDaNasmVAoFKJVq1Zi+/btNVzxs6Ey4/b479s/F29v75ov/BlQmbFr2rRpmWM3d+7cmi/8GVCZsfvoo4+Es7OzMDY2FvXr1xeenp5i48aNtVA1VZZMCCFqbFqTiIiIiJ5LL9zT00RERERU/RgaiYiIiEgSQyMRERERSWJoJCIiIiJJDI1EREREJImhkYiIiIgkMTQSERERkSSGRiIiIiKSxNBIRERERJIYGomIiIhIEkMjEREREUliaCQiIiIiSf8fT3AJgZBJlfgAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "words = [f\"Positive\", \"Negative\", \"Good Restaurant\", \"Bad Restaurant\"]\n",
    "vectors = [np.array(x) for x in [positive_review, negative_review, good_restraurant, bad_restraurant]]\n",
    "\n",
    "visualize_words(words, vectors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "good_phone = get_embedding(\"买的银色版真的很好看，一天就到了，晚上就开始拿起来完系统很丝滑流畅，做工扎实，手感细腻，很精致哦苹果一如既往的好品质\", engine=EMBEDDING_MODEL)\n",
    "bad_phone = get_embedding(\"随意降价，不予价保，服务态度差\", engine=EMBEDDING_MODEL)\n",
    "\n",
    "words = [f\"Positive\", \"Negative\", \"Good Phone\", \"Bad Phone\"]\n",
    "vectors = [np.array(x) for x in [positive_review, negative_review, good_phone, bad_phone]]\n",
    "\n",
    "visualize_words(words, vectors)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 通过向量进行零样本分类"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from sklearn.metrics import classification_report\n",
    "\n",
    "datafile_path = \"data/fine_food_reviews_with_embeddings_1k.csv\"\n",
    "\n",
    "df = pd.read_csv(datafile_path)\n",
    "df[\"embedding\"] = df.embedding.apply(eval).apply(np.array) # type: ignore\n",
    "\n",
    "# convert 5-star rating to binary sentiment\n",
    "df = df[df.Score != 3]\n",
    "df[\"sentiment\"] = df.Score.replace({1: \"negative\", 2: \"negative\", 4: \"positive\", 5: \"positive\"})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>ProductId</th>\n",
       "      <th>UserId</th>\n",
       "      <th>Score</th>\n",
       "      <th>Summary</th>\n",
       "      <th>Text</th>\n",
       "      <th>combined</th>\n",
       "      <th>n_tokens</th>\n",
       "      <th>embedding</th>\n",
       "      <th>sentiment</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>B003XPF9BO</td>\n",
       "      <td>A3R7JR3FMEBXQB</td>\n",
       "      <td>5</td>\n",
       "      <td>where does one  start...and stop... with a tre...</td>\n",
       "      <td>Wanted to save some to bring to my Chicago fam...</td>\n",
       "      <td>Title: where does one  start...and stop... wit...</td>\n",
       "      <td>52</td>\n",
       "      <td>[0.007018072064965963, -0.02731654793024063, 0...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>297</td>\n",
       "      <td>B003VXHGPK</td>\n",
       "      <td>A21VWSCGW7UUAR</td>\n",
       "      <td>4</td>\n",
       "      <td>Good, but not Wolfgang Puck good</td>\n",
       "      <td>Honestly, I have to admit that I expected a li...</td>\n",
       "      <td>Title: Good, but not Wolfgang Puck good; Conte...</td>\n",
       "      <td>178</td>\n",
       "      <td>[-0.003140551969408989, -0.009995664469897747,...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>296</td>\n",
       "      <td>B008JKTTUA</td>\n",
       "      <td>A34XBAIFT02B60</td>\n",
       "      <td>1</td>\n",
       "      <td>Should advertise coconut as an ingredient more...</td>\n",
       "      <td>First, these should be called Mac - Coconut ba...</td>\n",
       "      <td>Title: Should advertise coconut as an ingredie...</td>\n",
       "      <td>78</td>\n",
       "      <td>[-0.01757248118519783, -8.266511576948687e-05,...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>295</td>\n",
       "      <td>B000LKTTTW</td>\n",
       "      <td>A14MQ40CCU8B13</td>\n",
       "      <td>5</td>\n",
       "      <td>Best tomato soup</td>\n",
       "      <td>I have a hard time finding packaged food of an...</td>\n",
       "      <td>Title: Best tomato soup; Content: I have a har...</td>\n",
       "      <td>111</td>\n",
       "      <td>[-0.0013932279543951154, -0.011112828738987446...</td>\n",
       "      <td>positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>294</td>\n",
       "      <td>B001D09KAM</td>\n",
       "      <td>A34XBAIFT02B60</td>\n",
       "      <td>1</td>\n",
       "      <td>Should advertise coconut as an ingredient more...</td>\n",
       "      <td>First, these should be called Mac - Coconut ba...</td>\n",
       "      <td>Title: Should advertise coconut as an ingredie...</td>\n",
       "      <td>78</td>\n",
       "      <td>[-0.01757248118519783, -8.266511576948687e-05,...</td>\n",
       "      <td>negative</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0   ProductId          UserId  Score   \n",
       "0           0  B003XPF9BO  A3R7JR3FMEBXQB      5  \\\n",
       "1         297  B003VXHGPK  A21VWSCGW7UUAR      4   \n",
       "2         296  B008JKTTUA  A34XBAIFT02B60      1   \n",
       "3         295  B000LKTTTW  A14MQ40CCU8B13      5   \n",
       "4         294  B001D09KAM  A34XBAIFT02B60      1   \n",
       "\n",
       "                                             Summary   \n",
       "0  where does one  start...and stop... with a tre...  \\\n",
       "1                   Good, but not Wolfgang Puck good   \n",
       "2  Should advertise coconut as an ingredient more...   \n",
       "3                                   Best tomato soup   \n",
       "4  Should advertise coconut as an ingredient more...   \n",
       "\n",
       "                                                Text   \n",
       "0  Wanted to save some to bring to my Chicago fam...  \\\n",
       "1  Honestly, I have to admit that I expected a li...   \n",
       "2  First, these should be called Mac - Coconut ba...   \n",
       "3  I have a hard time finding packaged food of an...   \n",
       "4  First, these should be called Mac - Coconut ba...   \n",
       "\n",
       "                                            combined  n_tokens   \n",
       "0  Title: where does one  start...and stop... wit...        52  \\\n",
       "1  Title: Good, but not Wolfgang Puck good; Conte...       178   \n",
       "2  Title: Should advertise coconut as an ingredie...        78   \n",
       "3  Title: Best tomato soup; Content: I have a har...       111   \n",
       "4  Title: Should advertise coconut as an ingredie...        78   \n",
       "\n",
       "                                           embedding sentiment  \n",
       "0  [0.007018072064965963, -0.02731654793024063, 0...  positive  \n",
       "1  [-0.003140551969408989, -0.009995664469897747,...  positive  \n",
       "2  [-0.01757248118519783, -8.266511576948687e-05,...  negative  \n",
       "3  [-0.0013932279543951154, -0.011112828738987446...  positive  \n",
       "4  [-0.01757248118519783, -8.266511576948687e-05,...  negative  "
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "    negative       0.98      0.73      0.84       136\n",
      "    positive       0.96      1.00      0.98       789\n",
      "\n",
      "    accuracy                           0.96       925\n",
      "   macro avg       0.97      0.86      0.91       925\n",
      "weighted avg       0.96      0.96      0.96       925\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import PrecisionRecallDisplay\n",
    "from openai.embeddings_utils import cosine_similarity, get_embedding\n",
    "\n",
    "def evaluate_embeddings_approach(\n",
    "    labels = ['negative', 'positive'], \n",
    "    model = EMBEDDING_MODEL,\n",
    "):\n",
    "    label_embeddings = [get_embedding(label, engine=model) for label in labels]\n",
    "\n",
    "    def label_score(review_embedding, label_embeddings):\n",
    "        return cosine_similarity(review_embedding, label_embeddings[1]) - cosine_similarity(review_embedding, label_embeddings[0])\n",
    "\n",
    "    probas = df[\"embedding\"].apply(lambda x: label_score(x, label_embeddings))\n",
    "    preds = probas.apply(lambda x: 'positive' if x>0 else 'negative')\n",
    "\n",
    "    report = classification_report(df.sentiment, preds)\n",
    "    print(report)\n",
    "\n",
    "    display = PrecisionRecallDisplay.from_predictions(df.sentiment, probas, pos_label='positive')\n",
    "    _ = display.ax_.set_title(\"2-class Precision-Recall curve\")\n",
    "\n",
    "evaluate_embeddings_approach(labels=['An Amazon review with a negative sentiment.', 'An Amazon review with a positive sentiment.'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 通过机器学习进行分类"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/xuwenhao/miniconda3/envs/mactalk/lib/python3.10/site-packages/sklearn/linear_model/_logistic.py:458: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
      "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
      "\n",
      "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
      "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
      "Please also refer to the documentation for alternative solver options:\n",
      "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
      "  n_iter_i = _check_optimize_result(\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                    precision    recall  f1-score   support\n",
      "\n",
      "  news_agriculture       0.82      0.87      0.85       495\n",
      "          news_car       0.90      0.92      0.91       895\n",
      "      news_culture       0.78      0.81      0.79       741\n",
      "          news_edu       0.88      0.90      0.89       708\n",
      "news_entertainment       0.83      0.87      0.85      1051\n",
      "      news_finance       0.79      0.76      0.77       735\n",
      "         news_game       0.90      0.85      0.88       742\n",
      "        news_house       0.90      0.89      0.89       450\n",
      "     news_military       0.86      0.80      0.83       688\n",
      "       news_sports       0.93      0.92      0.93       968\n",
      "        news_story       0.81      0.73      0.77       197\n",
      "         news_tech       0.83      0.84      0.84      1052\n",
      "       news_travel       0.80      0.77      0.78       599\n",
      "        news_world       0.76      0.79      0.77       671\n",
      "             stock       0.00      0.00      0.00         8\n",
      "\n",
      "          accuracy                           0.85     10000\n",
      "         macro avg       0.79      0.78      0.78     10000\n",
      "      weighted avg       0.85      0.85      0.85     10000\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/xuwenhao/miniconda3/envs/mactalk/lib/python3.10/site-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/Users/xuwenhao/miniconda3/envs/mactalk/lib/python3.10/site-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/Users/xuwenhao/miniconda3/envs/mactalk/lib/python3.10/site-packages/sklearn/metrics/_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import classification_report, accuracy_score\n",
    "\n",
    "training_data = pd.read_parquet(\"data/toutiao_cat_data_all_with_embeddings.parquet\")\n",
    "\n",
    "df =  training_data.sample(50000, random_state=42)\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    list(df.embedding.values), df.category, test_size=0.2, random_state=42\n",
    ")\n",
    "\n",
    "clf = LogisticRegression()\n",
    "clf.fit(X_train, y_train)\n",
    "preds = clf.predict(X_test)\n",
    "probas = clf.predict_proba(X_test)\n",
    "\n",
    "report = classification_report(y_test, preds)\n",
    "print(report)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>code</th>\n",
       "      <th>category</th>\n",
       "      <th>title</th>\n",
       "      <th>keywords</th>\n",
       "      <th>combined</th>\n",
       "      <th>n_tokens</th>\n",
       "      <th>embedding</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>306237</th>\n",
       "      <td>6524143112997044743</td>\n",
       "      <td>100</td>\n",
       "      <td>news_story</td>\n",
       "      <td>将女儿扔老家10年后我接她回家，没多久妻子被杀我才知她身份不对</td>\n",
       "      <td>小安,不好意思,她在杀人,星期六,女孩儿,超能力,工程事故,冯为杰,冯欣冉,浏览器,狄海</td>\n",
       "      <td>标题: 将女儿扔老家10年后我接她回家，没多久妻子被杀我才知她身份不对; 关键字: 小安,不...</td>\n",
       "      <td>110</td>\n",
       "      <td>[-0.015406427904963493, -0.007328800391405821,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>123419</th>\n",
       "      <td>6553127618529460483</td>\n",
       "      <td>101</td>\n",
       "      <td>news_culture</td>\n",
       "      <td>在中国古代，有没有人认为地球是圆的呢？</td>\n",
       "      <td></td>\n",
       "      <td>标题: 在中国古代，有没有人认为地球是圆的呢？; 关键字:</td>\n",
       "      <td>28</td>\n",
       "      <td>[0.023000160232186317, -0.004472616594284773, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49255</th>\n",
       "      <td>6511643474234704135</td>\n",
       "      <td>113</td>\n",
       "      <td>news_world</td>\n",
       "      <td>先冻结房产，再冻结养老活命钱，韩国为何如此苛刻对待前总统朴槿惠？</td>\n",
       "      <td>前总统,朴槿惠,检方,韩国,韩国法院</td>\n",
       "      <td>标题: 先冻结房产，再冻结养老活命钱，韩国为何如此苛刻对待前总统朴槿惠？; 关键字: 前总统...</td>\n",
       "      <td>81</td>\n",
       "      <td>[0.0021718391217291355, -0.01761491596698761, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>135430</th>\n",
       "      <td>6553195755832606979</td>\n",
       "      <td>109</td>\n",
       "      <td>news_tech</td>\n",
       "      <td>电脑总是无响应是什么回事？</td>\n",
       "      <td></td>\n",
       "      <td>标题: 电脑总是无响应是什么回事？; 关键字:</td>\n",
       "      <td>26</td>\n",
       "      <td>[-0.01490439660847187, -0.012121899053454399, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>331509</th>\n",
       "      <td>6554427356038038019</td>\n",
       "      <td>113</td>\n",
       "      <td>news_world</td>\n",
       "      <td>中国宣布全面终止合作，这个国家将付出沉重代价！</td>\n",
       "      <td>天然气,出口商,煤炭,中方,澳大利亚</td>\n",
       "      <td>标题: 中国宣布全面终止合作，这个国家将付出沉重代价！; 关键字: 天然气,出口商,煤炭,中...</td>\n",
       "      <td>59</td>\n",
       "      <td>[-0.01608712412416935, -0.033476412296295166, ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         id  code      category   \n",
       "306237  6524143112997044743   100    news_story  \\\n",
       "123419  6553127618529460483   101  news_culture   \n",
       "49255   6511643474234704135   113    news_world   \n",
       "135430  6553195755832606979   109     news_tech   \n",
       "331509  6554427356038038019   113    news_world   \n",
       "\n",
       "                                   title   \n",
       "306237   将女儿扔老家10年后我接她回家，没多久妻子被杀我才知她身份不对  \\\n",
       "123419               在中国古代，有没有人认为地球是圆的呢？   \n",
       "49255   先冻结房产，再冻结养老活命钱，韩国为何如此苛刻对待前总统朴槿惠？   \n",
       "135430                     电脑总是无响应是什么回事？   \n",
       "331509           中国宣布全面终止合作，这个国家将付出沉重代价！   \n",
       "\n",
       "                                            keywords   \n",
       "306237  小安,不好意思,她在杀人,星期六,女孩儿,超能力,工程事故,冯为杰,冯欣冉,浏览器,狄海  \\\n",
       "123419                                                 \n",
       "49255                             前总统,朴槿惠,检方,韩国,韩国法院   \n",
       "135430                                                 \n",
       "331509                            天然气,出口商,煤炭,中方,澳大利亚   \n",
       "\n",
       "                                                 combined  n_tokens   \n",
       "306237  标题: 将女儿扔老家10年后我接她回家，没多久妻子被杀我才知她身份不对; 关键字: 小安,不...       110  \\\n",
       "123419                     标题: 在中国古代，有没有人认为地球是圆的呢？; 关键字:         28   \n",
       "49255   标题: 先冻结房产，再冻结养老活命钱，韩国为何如此苛刻对待前总统朴槿惠？; 关键字: 前总统...        81   \n",
       "135430                           标题: 电脑总是无响应是什么回事？; 关键字:         26   \n",
       "331509  标题: 中国宣布全面终止合作，这个国家将付出沉重代价！; 关键字: 天然气,出口商,煤炭,中...        59   \n",
       "\n",
       "                                                embedding  \n",
       "306237  [-0.015406427904963493, -0.007328800391405821,...  \n",
       "123419  [0.023000160232186317, -0.004472616594284773, ...  \n",
       "49255   [0.0021718391217291355, -0.01761491596698761, ...  \n",
       "135430  [-0.01490439660847187, -0.012121899053454399, ...  \n",
       "331509  [-0.01608712412416935, -0.033476412296295166, ...  "
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction: news_sports\n"
     ]
    }
   ],
   "source": [
    "sports_news = get_embedding(\"【西海岸】ESPN：马龙教练不容易，保持核心掘金能复刻冠军\", engine=EMBEDDING_MODEL)\n",
    "prediction = clf.predict([sports_news])[0]\n",
    "print(f\"Prediction: {prediction}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction: news_finance\n"
     ]
    }
   ],
   "source": [
    "finance_news = get_embedding(\"北上资金净买入超百亿元 重点加仓新能源汽车产业\", engine=EMBEDDING_MODEL)\n",
    "prediction = clf.predict([finance_news])[0]\n",
    "print(f\"Prediction: {prediction}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 通过机器学习进行聚类"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn.cluster import KMeans\n",
    "\n",
    "embedding_df = pd.read_parquet(\"data/20_newsgroup_with_embedding.parquet\")\n",
    "\n",
    "matrix = np.vstack(embedding_df.embedding.values) # type: ignore\n",
    "num_of_clusters = 20\n",
    "\n",
    "kmeans = KMeans(n_clusters=num_of_clusters, init=\"k-means++\", n_init=10, random_state=42)\n",
    "kmeans.fit(matrix)\n",
    "labels = kmeans.labels_\n",
    "embedding_df[\"cluster\"] = labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>target</th>\n",
       "      <th>title</th>\n",
       "      <th>n_tokens</th>\n",
       "      <th>embedding</th>\n",
       "      <th>cluster</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>I was wondering if anyone out there could enli...</td>\n",
       "      <td>7</td>\n",
       "      <td>rec.autos</td>\n",
       "      <td>121</td>\n",
       "      <td>[-0.0391300804913044, 0.013502633199095726, -0...</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>\\nIt depends on your priorities.  A lot of peo...</td>\n",
       "      <td>7</td>\n",
       "      <td>rec.autos</td>\n",
       "      <td>108</td>\n",
       "      <td>[-0.0011249205563217402, -0.00376517535187304,...</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>an excellent automatic can be found in the sub...</td>\n",
       "      <td>7</td>\n",
       "      <td>rec.autos</td>\n",
       "      <td>476</td>\n",
       "      <td>[-0.018259447067975998, -0.008410007692873478,...</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>: Ford and his automobile.  I need information...</td>\n",
       "      <td>7</td>\n",
       "      <td>rec.autos</td>\n",
       "      <td>86</td>\n",
       "      <td>[-0.012589422054588795, 0.006539034191519022, ...</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>\\nYo! Watch the attributions--I didn't say tha...</td>\n",
       "      <td>7</td>\n",
       "      <td>rec.autos</td>\n",
       "      <td>130</td>\n",
       "      <td>[-0.0006192282889969647, -0.011226896196603775...</td>\n",
       "      <td>10</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                text  target      title   \n",
       "0  I was wondering if anyone out there could enli...       7  rec.autos  \\\n",
       "1  \\nIt depends on your priorities.  A lot of peo...       7  rec.autos   \n",
       "2  an excellent automatic can be found in the sub...       7  rec.autos   \n",
       "3  : Ford and his automobile.  I need information...       7  rec.autos   \n",
       "4  \\nYo! Watch the attributions--I didn't say tha...       7  rec.autos   \n",
       "\n",
       "   n_tokens                                          embedding  cluster  \n",
       "0       121  [-0.0391300804913044, 0.013502633199095726, -0...        5  \n",
       "1       108  [-0.0011249205563217402, -0.00376517535187304,...        5  \n",
       "2       476  [-0.018259447067975998, -0.008410007692873478,...        5  \n",
       "3        86  [-0.012589422054588795, 0.006539034191519022, ...        5  \n",
       "4       130  [-0.0006192282889969647, -0.011226896196603775...       10  "
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "embedding_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cluster</th>\n",
       "      <th>count</th>\n",
       "      <th>rank1</th>\n",
       "      <th>rank1_count</th>\n",
       "      <th>rank2</th>\n",
       "      <th>rank2_count</th>\n",
       "      <th>per_1</th>\n",
       "      <th>per_1_2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>432</td>\n",
       "      <td>comp.windows.x</td>\n",
       "      <td>406</td>\n",
       "      <td>comp.sys.mac.hardware</td>\n",
       "      <td>1.0</td>\n",
       "      <td>93.98%</td>\n",
       "      <td>94.21%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>418</td>\n",
       "      <td>sci.space</td>\n",
       "      <td>388</td>\n",
       "      <td>alt.atheism</td>\n",
       "      <td>2.0</td>\n",
       "      <td>92.82%</td>\n",
       "      <td>93.30%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>1035</td>\n",
       "      <td>comp.sys.ibm.pc.hardware</td>\n",
       "      <td>396</td>\n",
       "      <td>comp.sys.mac.hardware</td>\n",
       "      <td>387.0</td>\n",
       "      <td>38.26%</td>\n",
       "      <td>75.65%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>471</td>\n",
       "      <td>rec.sport.hockey</td>\n",
       "      <td>455</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>96.60%</td>\n",
       "      <td>96.60%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>716</td>\n",
       "      <td>talk.politics.misc</td>\n",
       "      <td>270</td>\n",
       "      <td>alt.atheism</td>\n",
       "      <td>150.0</td>\n",
       "      <td>37.71%</td>\n",
       "      <td>58.66%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>5</td>\n",
       "      <td>511</td>\n",
       "      <td>rec.autos</td>\n",
       "      <td>420</td>\n",
       "      <td>comp.sys.mac.hardware</td>\n",
       "      <td>6.0</td>\n",
       "      <td>82.19%</td>\n",
       "      <td>83.37%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>6</td>\n",
       "      <td>870</td>\n",
       "      <td>rec.motorcycles</td>\n",
       "      <td>100</td>\n",
       "      <td>alt.atheism</td>\n",
       "      <td>73.0</td>\n",
       "      <td>11.49%</td>\n",
       "      <td>19.89%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>7</td>\n",
       "      <td>570</td>\n",
       "      <td>comp.os.ms-windows.misc</td>\n",
       "      <td>338</td>\n",
       "      <td>comp.sys.mac.hardware</td>\n",
       "      <td>46.0</td>\n",
       "      <td>59.30%</td>\n",
       "      <td>67.37%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>8</td>\n",
       "      <td>435</td>\n",
       "      <td>talk.politics.mideast</td>\n",
       "      <td>372</td>\n",
       "      <td>alt.atheism</td>\n",
       "      <td>23.0</td>\n",
       "      <td>85.52%</td>\n",
       "      <td>90.80%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>9</td>\n",
       "      <td>84</td>\n",
       "      <td>comp.os.ms-windows.misc</td>\n",
       "      <td>8</td>\n",
       "      <td>comp.sys.mac.hardware</td>\n",
       "      <td>8.0</td>\n",
       "      <td>9.52%</td>\n",
       "      <td>19.05%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>10</td>\n",
       "      <td>527</td>\n",
       "      <td>talk.politics.guns</td>\n",
       "      <td>382</td>\n",
       "      <td>talk.religion.misc</td>\n",
       "      <td>35.0</td>\n",
       "      <td>72.49%</td>\n",
       "      <td>79.13%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>11</td>\n",
       "      <td>554</td>\n",
       "      <td>comp.graphics</td>\n",
       "      <td>324</td>\n",
       "      <td>comp.sys.mac.hardware</td>\n",
       "      <td>17.0</td>\n",
       "      <td>58.48%</td>\n",
       "      <td>61.55%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>12</td>\n",
       "      <td>372</td>\n",
       "      <td>rec.motorcycles</td>\n",
       "      <td>355</td>\n",
       "      <td>alt.atheism</td>\n",
       "      <td>1.0</td>\n",
       "      <td>95.43%</td>\n",
       "      <td>95.70%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>13</td>\n",
       "      <td>840</td>\n",
       "      <td>soc.religion.christian</td>\n",
       "      <td>460</td>\n",
       "      <td>alt.atheism</td>\n",
       "      <td>188.0</td>\n",
       "      <td>54.76%</td>\n",
       "      <td>77.14%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>14</td>\n",
       "      <td>468</td>\n",
       "      <td>rec.sport.baseball</td>\n",
       "      <td>451</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>96.37%</td>\n",
       "      <td>96.37%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>15</td>\n",
       "      <td>538</td>\n",
       "      <td>misc.forsale</td>\n",
       "      <td>438</td>\n",
       "      <td>comp.sys.mac.hardware</td>\n",
       "      <td>27.0</td>\n",
       "      <td>81.41%</td>\n",
       "      <td>86.43%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>16</td>\n",
       "      <td>363</td>\n",
       "      <td>sci.crypt</td>\n",
       "      <td>349</td>\n",
       "      <td>comp.sys.mac.hardware</td>\n",
       "      <td>1.0</td>\n",
       "      <td>96.14%</td>\n",
       "      <td>96.42%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>17</td>\n",
       "      <td>479</td>\n",
       "      <td>sci.electronics</td>\n",
       "      <td>336</td>\n",
       "      <td>comp.sys.mac.hardware</td>\n",
       "      <td>28.0</td>\n",
       "      <td>70.15%</td>\n",
       "      <td>75.99%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>18</td>\n",
       "      <td>425</td>\n",
       "      <td>sci.med</td>\n",
       "      <td>402</td>\n",
       "      <td>alt.atheism</td>\n",
       "      <td>1.0</td>\n",
       "      <td>94.59%</td>\n",
       "      <td>94.82%</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>19</td>\n",
       "      <td>532</td>\n",
       "      <td>sci.electronics</td>\n",
       "      <td>62</td>\n",
       "      <td>comp.sys.mac.hardware</td>\n",
       "      <td>23.0</td>\n",
       "      <td>11.65%</td>\n",
       "      <td>15.98%</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    cluster  count                     rank1  rank1_count   \n",
       "0         0    432            comp.windows.x          406  \\\n",
       "1         1    418                 sci.space          388   \n",
       "2         2   1035  comp.sys.ibm.pc.hardware          396   \n",
       "3         3    471          rec.sport.hockey          455   \n",
       "4         4    716        talk.politics.misc          270   \n",
       "5         5    511                 rec.autos          420   \n",
       "6         6    870           rec.motorcycles          100   \n",
       "7         7    570   comp.os.ms-windows.misc          338   \n",
       "8         8    435     talk.politics.mideast          372   \n",
       "9         9     84   comp.os.ms-windows.misc            8   \n",
       "10       10    527        talk.politics.guns          382   \n",
       "11       11    554             comp.graphics          324   \n",
       "12       12    372           rec.motorcycles          355   \n",
       "13       13    840    soc.religion.christian          460   \n",
       "14       14    468        rec.sport.baseball          451   \n",
       "15       15    538              misc.forsale          438   \n",
       "16       16    363                 sci.crypt          349   \n",
       "17       17    479           sci.electronics          336   \n",
       "18       18    425                   sci.med          402   \n",
       "19       19    532           sci.electronics           62   \n",
       "\n",
       "                    rank2  rank2_count   per_1 per_1_2  \n",
       "0   comp.sys.mac.hardware          1.0  93.98%  94.21%  \n",
       "1             alt.atheism          2.0  92.82%  93.30%  \n",
       "2   comp.sys.mac.hardware        387.0  38.26%  75.65%  \n",
       "3                       0          0.0  96.60%  96.60%  \n",
       "4             alt.atheism        150.0  37.71%  58.66%  \n",
       "5   comp.sys.mac.hardware          6.0  82.19%  83.37%  \n",
       "6             alt.atheism         73.0  11.49%  19.89%  \n",
       "7   comp.sys.mac.hardware         46.0  59.30%  67.37%  \n",
       "8             alt.atheism         23.0  85.52%  90.80%  \n",
       "9   comp.sys.mac.hardware          8.0   9.52%  19.05%  \n",
       "10     talk.religion.misc         35.0  72.49%  79.13%  \n",
       "11  comp.sys.mac.hardware         17.0  58.48%  61.55%  \n",
       "12            alt.atheism          1.0  95.43%  95.70%  \n",
       "13            alt.atheism        188.0  54.76%  77.14%  \n",
       "14                      0          0.0  96.37%  96.37%  \n",
       "15  comp.sys.mac.hardware         27.0  81.41%  86.43%  \n",
       "16  comp.sys.mac.hardware          1.0  96.14%  96.42%  \n",
       "17  comp.sys.mac.hardware         28.0  70.15%  75.99%  \n",
       "18            alt.atheism          1.0  94.59%  94.82%  \n",
       "19  comp.sys.mac.hardware         23.0  11.65%  15.98%  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "# 统计每个cluster的数量\n",
    "new_df = embedding_df.groupby('cluster')['cluster'].count().reset_index(name='count')\n",
    "\n",
    "# 统计这个cluster里最多的分类的数量\n",
    "title_count = embedding_df.groupby(['cluster', 'title']).size().reset_index(name='title_count')\n",
    "first_titles = title_count.groupby('cluster').apply(lambda x: x.nlargest(1, columns=['title_count']))\n",
    "first_titles = first_titles.reset_index(drop=True)\n",
    "new_df = pd.merge(new_df, first_titles[['cluster', 'title', 'title_count']], on='cluster', how='left')\n",
    "new_df = new_df.rename(columns={'title': 'rank1', 'title_count': 'rank1_count'})\n",
    "\n",
    "# 统计这个cluster里第二多的分类的数量\n",
    "second_titles = title_count[~title_count['title'].isin(first_titles['title'])]\n",
    "second_titles = second_titles.groupby('cluster').apply(lambda x: x.nlargest(1, columns=['title_count']))\n",
    "second_titles = second_titles.reset_index(drop=True)\n",
    "new_df = pd.merge(new_df, second_titles[['cluster', 'title', 'title_count']], on='cluster', how='left')\n",
    "new_df = new_df.rename(columns={'title': 'rank2', 'title_count': 'rank2_count'})\n",
    "new_df.fillna(0, inplace=True)\n",
    "new_df['per_1'] = (new_df['rank1_count'] / new_df['count']).map(lambda x: '{:.2%}'.format(x))\n",
    "new_df['per_1_2'] = ((new_df['rank1_count'] + new_df['rank2_count'])/ new_df['count']).map(lambda x: '{:.2%}'.format(x))\n",
    "\n",
    "# 将缺失值替换为 0\n",
    "# 输出结果\n",
    "display(new_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cluster 0, Rank 1: comp.windows.x, Theme: Xlib编程\n",
      "Cluster 1, Rank 1: sci.space, Theme: 太空技术与航空\n",
      "Cluster 2, Rank 1: comp.sys.ibm.pc.hardware, Theme: 电脑硬件\n",
      "Cluster 3, Rank 1: rec.sport.hockey, Theme: 欧洲冰球vs北美冰球\n",
      "Cluster 4, Rank 1: talk.politics.misc, Theme: 社会观点与自由\n",
      "Cluster 5, Rank 1: rec.autos, Theme: 汽车硬件\n",
      "Cluster 6, Rank 1: rec.motorcycles, Theme: 数学与文化冲击\n",
      "Cluster 7, Rank 1: comp.os.ms-windows.misc, Theme: PC软件与硬件\n",
      "Cluster 8, Rank 1: talk.politics.mideast, Theme: “穆斯林大屠杀”\n",
      "Cluster 9, Rank 1: comp.os.ms-windows.misc, Theme: 科技产品\"\"\"\n",
      "Cluster 10, Rank 1: talk.politics.guns, Theme: 枪支管制与安全\n",
      "Cluster 11, Rank 1: comp.graphics, Theme: 计算机编程与硬件\n",
      "Cluster 12, Rank 1: rec.motorcycles, Theme: 安全骑行：驾驶课程与技巧\n",
      "Cluster 13, Rank 1: soc.religion.christian, Theme: 宗教信仰与实践\n",
      "Cluster 14, Rank 1: rec.sport.baseball, Theme: 棒球联盟\n",
      "Cluster 15, Rank 1: misc.forsale, Theme: 购物优惠和出售\n",
      "Cluster 16, Rank 1: sci.crypt, Theme: 关于加密政策的讨论\n",
      "Cluster 17, Rank 1: sci.electronics, Theme: 电子设备连接与维护\n",
      "Cluster 18, Rank 1: sci.med, Theme: 药物和疾病\n",
      "Cluster 19, Rank 1: sci.electronics, Theme: 电子邮件使用者研究\n"
     ]
    }
   ],
   "source": [
    "items_per_cluster = 10\n",
    "COMPLETIONS_MODEL = \"text-davinci-003\"\n",
    "\n",
    "for i in range(num_of_clusters):\n",
    "    cluster_name = new_df[new_df.cluster == i].iloc[0].rank1\n",
    "    print(f\"Cluster {i}, Rank 1: {cluster_name}, Theme:\", end=\" \")\n",
    "\n",
    "    content = \"\\n\".join(\n",
    "        embedding_df[embedding_df.cluster == i].text.sample(items_per_cluster, random_state=42).values\n",
    "    )\n",
    "    response = openai.Completion.create(\n",
    "        model=COMPLETIONS_MODEL,\n",
    "        prompt=f'''我们想要给下面的内容，分组成有意义的类别，以便我们可以对其进行总结。请根据下面这些内容的共同点，总结一个50个字以内的新闻组的名称么？比如 “PC硬件”\\n\\n内容:\\n\"\"\"\\n{content}\\n\"\"\"新闻组名称：''',\n",
    "        temperature=0,\n",
    "        max_tokens=100,\n",
    "        top_p=1,\n",
    "    )\n",
    "    print(response[\"choices\"][0][\"text\"].replace(\"\\n\", \"\")) # type: ignore\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cluster 0, Rank 1: comp.windows.x, 抽样翻译: 没有实际执行它？不知怎么回事，我的一个xterminal用户使得只要点击鼠标右键，就会自动杀死所有客户端-哦，我的：-(谢谢，Fish\n",
      "Cluster 1, Rank 1: sci.space, 抽样翻译: 韦恩·马森和他的团伙在阿拉巴马州发生了什么？我还听说有一个未经证实的谣言，即航空大使们已经消失了。有其他人可以证实吗？\n",
      "Cluster 2, Rank 1: comp.sys.ibm.pc.hardware, 抽样翻译: 我怀疑这不是一个特定于Quadra的问题。去年我不得不放弃我“古老”的Bernoulli 20（每个磁带的价格大约是90美元，使整个事情的价值超过我的整个电脑;）。Ocean Microsystems的技术支持人员建议可以使用一些第三方驱动程序来解决这个问题 - 在我的情况下，磁带无法格式化/挂载/分区用于A / UX。\n",
      "Cluster 3, Rank 1: rec.sport.hockey, 抽样翻译: 我相信那是4-1。罗德·布林道·阿莫尔在第三节19.59时分攻入一球。\n",
      "Cluster 4, Rank 1: talk.politics.misc, 抽样翻译: 为了确保每个人都清楚：“它从未有过”是指“保护”，而不是“未能保护”；即，在我的一生中，我从未见过美国政府始终保护美国公民的利益，除非是意外。\n",
      "Cluster 5, Rank 1: rec.autos, 抽样翻译: 噢，来吧，傻瓜，你要做的就是在你的引擎罩上割一个洞，然后把一个管子放进去，这样你就可以把机油倒进去了。你觉得那些热门车上的大空气进气装置是干什么的？它们只是为了外观，没有人知道，它们提供了进入机油填充孔的途径。\n",
      "Cluster 6, Rank 1: rec.motorcycles, 抽样翻译: 你真是个失败者\n",
      "Cluster 7, Rank 1: comp.os.ms-windows.misc, 抽样翻译: 偶尔你也得为表现良好的东西说句好话。我的东西泽3401没有任何问题。它在DOS和OS/2上运行非常好。对于OS/2，你不需要加载任何特殊的驱动程序。安装会检测到它是一个东西泽驱动器，然后就完成了。顺便说一句，它也非常快！\n",
      "Cluster 8, Rank 1: talk.politics.mideast, 抽样翻译: Avi，    供你参考，伊斯兰教允许宗教自由——在宗教上没有强制。犹太教是否也允许宗教自由（即是否认可非犹太人）？只是好奇而已。\n",
      "Cluster 9, Rank 1: comp.os.ms-windows.misc, 抽样翻译: 每个人都有自己的梦想，但只有勇敢追求梦想的人才能实现它。\n",
      "Cluster 10, Rank 1: talk.politics.guns, 抽样翻译: 不一定，特别是如果强奸犯被认定为此类罪行。例如，如果你有意将手指伸入装满老鼠夹的地方，然后被夹住，这是谁的错？\n",
      "Cluster 11, Rank 1: comp.graphics, 抽样翻译: 帮帮我！！我需要代码/包/任何东西来处理3D数据，并将其转换为带有隐藏线的线框表面。我正在使用DOS机器，代码可以是ANSI C或C ++，ANSI Fortran或Basic。我使用的数据形成一个矩形网格。请将您的回复发布到网络上，以便其他人受益。我的个人观点是，这是一个普遍的兴趣问题。谢谢！！！！！\n",
      "Cluster 12, Rank 1: rec.motorcycles, 抽样翻译: 这是一段心理学，对于任何长期骑行者来说都是必不可少的。人们不会去想“如果我这么做会有别人受到影响吗？”他们只会评估“如果我这么做会受到影响吗？”\n",
      "Cluster 13, Rank 1: soc.religion.christian, 抽样翻译: 这是一个非常薄弱的论点，因为没有独立的支持文本（关于关键事件）。至于新约最古老的现存文本的日期......如果现在只有一个关于美国内战的现存文本，你会怎么想？现在考虑一个大部分文盲的人口，每一份手稿都是手工复制的......--Hal\n",
      "Cluster 14, Rank 1: rec.sport.baseball, 抽样翻译: 这个赔率意味着你下注5美元赌反败者赢8美元，或者下注9美元赌胜者赢5美元。\n",
      "Cluster 15, Rank 1: misc.forsale, 抽样翻译: 嗯，标题就是这样......我正在寻找便宜的二手TG-16游戏，它们支持2个或更多玩家（同时）....请给我发送所有带有价格的报价。\n",
      "Cluster 16, Rank 1: sci.crypt, 抽样翻译: 哪里？老实说，我没有看到任何……我不同意，至少有其他标准已经存在。此外，即使他们限制NREN上的加密，谁在乎呢？大部分互联网都是商业的。NREN只适用于政府和大学研究（阅读提案-它是一条“数据高速公路”，与互联网无关）。\n",
      "Cluster 17, Rank 1: sci.electronics, 抽样翻译: 动态RAM不是基于翻转锁存器；基本上每个位只有一个晶体管和电容来存储！静态RAM是基于翻转锁存器，更加昂贵，密度也更低。如果忽略电子和热膨胀，两者都没有任何“移动”的部件...Chris\n",
      "Cluster 18, Rank 1: sci.med, 抽样翻译: 化学物质已经消失，感谢大家的回应。\n",
      "Cluster 19, Rank 1: sci.electronics, 抽样翻译: 尝试lyman.pppl.gov--/pub/8051\n"
     ]
    }
   ],
   "source": [
    "items_per_cluster = 1\n",
    "COMPLETIONS_MODEL = \"text-davinci-003\"\n",
    "\n",
    "for i in range(num_of_clusters):\n",
    "    cluster_name = new_df[new_df.cluster == i].iloc[0].rank1\n",
    "    print(f\"Cluster {i}, Rank 1: {cluster_name}, 抽样翻译:\", end=\" \")\n",
    "\n",
    "    content = \"\\n\".join(\n",
    "        embedding_df[(embedding_df.cluster == i) & (embedding_df.n_tokens < 100)].text.sample(items_per_cluster, random_state=42).values\n",
    "    )\n",
    "    response = openai.Completion.create(\n",
    "        model=COMPLETIONS_MODEL,\n",
    "        prompt=f'''请把下面的内容翻译成中文\\n\\n内容:\\n\"\"\"\\n{content}\\n\"\"\"翻译：''',\n",
    "        temperature=0,\n",
    "        max_tokens=500,\n",
    "        top_p=1,\n",
    "    )\n",
    "    print(response[\"choices\"][0][\"text\"].replace(\"\\n\", \"\")) # type: ignore"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 减少聚类的数量，来进行可视化"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn.cluster import KMeans\n",
    "\n",
    "embedding_df = pd.read_parquet(\"data/20_newsgroup_with_embedding.parquet\")\n",
    "\n",
    "embedding_df = embedding_df[\n",
    "    (embedding_df.title == \"sci.electronics\") |\n",
    "    (embedding_df.title == \"sci.med\") |\n",
    "    (embedding_df.title == \"rec.motorcycles\") |\n",
    "    (embedding_df.title == \"rec.sport.baseball\")\n",
    "]\n",
    "matrix = np.vstack(embedding_df.embedding.values) # type: ignore\n",
    "num_of_clusters = 4\n",
    "\n",
    "kmeans = KMeans(n_clusters=num_of_clusters, init=\"k-means++\", n_init=10, random_state=42)\n",
    "kmeans.fit(matrix)\n",
    "labels = kmeans.labels_\n",
    "embedding_df[\"cluster\"] = labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Clusters identified visualized in language 2d using t-SNE')"
      ]
     },
     "execution_count": 81,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.manifold import TSNE\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "tsne = TSNE(n_components=2, perplexity=15, random_state=42, init=\"random\", learning_rate=200)\n",
    "vis_dims2 = tsne.fit_transform(matrix)\n",
    "\n",
    "x = [x for x, y in vis_dims2]\n",
    "y = [y for x, y in vis_dims2]\n",
    "\n",
    "for category, color in zip([\"sci.electronics\", \"sci.med\", \"rec.motorcycles\", \"rec.sport.baseball\"], [\"purple\", \"green\", \"red\", \"blue\"]):\n",
    "    xs = np.array(x)[embedding_df.title == category]\n",
    "    ys = np.array(y)[embedding_df.title == category]\n",
    "    plt.scatter(xs, ys, color=color, alpha=0.3)\n",
    "\n",
    "    avg_x = xs.mean()\n",
    "    avg_y = ys.mean()\n",
    "\n",
    "    plt.scatter(avg_x, avg_y, marker=\"x\", color=color, s=100) # type: ignore\n",
    "plt.title(\"Clusters identified visualized in language 2d using t-SNE\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 多轮聊天对话"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 使用原始的 text-davinci-003"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "COMPLETION_MODEL = \"text-davinci-003\"\n",
    "\n",
    "def get_response(prompt, temperature = 1.0, stop=None):\n",
    "    completions = openai.Completion.create (\n",
    "        engine=COMPLETION_MODEL,\n",
    "        prompt=prompt,\n",
    "        max_tokens=1024,\n",
    "        n=1,\n",
    "        stop=stop,\n",
    "        temperature=temperature,        \n",
    "    )\n",
    "    message = completions.choices[0].text # type: ignore\n",
    "    return message"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 你好！我准备去我家附近的湖边游玩，和朋友们一起划船、赛龙舟，还可以吃粽子和豆沙包，庆祝端午节。\n"
     ]
    }
   ],
   "source": [
    "question = \"\"\"Q: 你好，端午节你准备去哪里？\n",
    "A:\"\"\"\n",
    "\n",
    "answer = get_response(question, temperature=0.5, stop=\"Q:\")\n",
    "print(answer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 在武汉，你可以去参观汉口火车站、江汉桥、黄鹤楼、武汉大学、武汉植物园等景点。\n"
     ]
    }
   ],
   "source": [
    "question = \"\"\"Q: 你好，端午节你准备去哪里？\n",
    "A: 端午节我打算去武汉。\n",
    "Q: 那里有什么好玩的旅游景点？\n",
    "A: \"\"\"\n",
    "\n",
    "answer = get_response(question, temperature=0.5, stop=\"Q:\")\n",
    "print(answer)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 使用ChatCompletion接口"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [],
   "source": [
    "import openai\n",
    "import os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "\n",
    "class Conversation:\n",
    "    def __init__(self, prompt, num_of_round):\n",
    "        self.prompt = prompt\n",
    "        self.num_of_round = num_of_round\n",
    "        self.messages = []\n",
    "        self.messages.append({\"role\": \"system\", \"content\": self.prompt})\n",
    "\n",
    "    def ask(self, question):\n",
    "        try:\n",
    "            self.messages.append( {\"role\": \"user\", \"content\": question})\n",
    "            response = openai.ChatCompletion.create(\n",
    "                model=\"gpt-3.5-turbo\",\n",
    "                messages=self.messages,\n",
    "                temperature=0.5,\n",
    "                max_tokens=2048,\n",
    "                top_p=1,\n",
    "            )\n",
    "        except Exception as e:\n",
    "            print(e)\n",
    "            return e\n",
    "\n",
    "        message = response[\"choices\"][0][\"message\"][\"content\"] # type: ignore\n",
    "        self.messages.append({\"role\": \"assistant\", \"content\": message})\n",
    "        \n",
    "        if len(self.messages) > self.num_of_round*2 + 1:\n",
    "            del self.messages[1:3]\n",
    "        return message\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "User : 你好，端午节你准备去哪里？\n",
      "Assistant : 嗨，我打算去武汉旅游。听说那里有很多有趣的景点和美食，我很期待能够去探索一下。你有去过武汉吗？\n",
      "\n",
      "User : 那里有什么好玩的旅游景点\n",
      "Assistant : 武汉有很多著名的旅游景点，比如：\n",
      "\n",
      "1.黄鹤楼：这座古老的楼阁建于公元223年，是中国文学史上的重要地标之一。\n",
      "\n",
      "2.东湖：这是武汉市最大的城市湖泊，周围有美丽的公园和游乐设施，是一个很好的休闲场所。\n",
      "\n",
      "3.武汉长江大桥：这座桥梁是中国第一座自主设计和建造的公铁两用桥，也是武汉的标志性建筑之一。\n",
      "\n",
      "4.户部巷：这是武汉最古老的街道之一，保留了大量的明清时期建筑和传统文化。\n",
      "\n",
      "此外，武汉还有很多其他的景点，比如黄鹤楼公园、南湖、古琴台等等，相信你一定会在这里度过一个难忘的端午节假期。\n",
      "\n"
     ]
    }
   ],
   "source": [
    "prompt = \"\"\"请你扮演我的朋友Tom，如果有人问你端午节去哪里玩，你会回答打算去武汉旅游\"\"\"\n",
    "conv1 = Conversation(prompt, 3)\n",
    "question1 = \"你好，端午节你准备去哪里？\"\n",
    "print(\"User : %s\" % question1)\n",
    "print(\"Assistant : %s\\n\" % conv1.ask(question1))\n",
    "\n",
    "question2 = \"那里有什么好玩的旅游景点\"\n",
    "print(\"User : %s\" % question2)\n",
    "print(\"Assistant : %s\\n\" % conv1.ask(question2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 成本与效率问题"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 数据的Batch处理"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Lines of text before filtering:  382688\n"
     ]
    }
   ],
   "source": [
    "import tiktoken\n",
    "import openai\n",
    "import os\n",
    "import backoff\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "\n",
    "# embedding model parameters\n",
    "embedding_model = \"text-embedding-ada-002\"\n",
    "embedding_encoding = \"cl100k_base\"  # this the encoding for text-embedding-ada-002\n",
    "max_tokens = 8000  # the maximum for text-embedding-ada-002 is 8191\n",
    "\n",
    "\n",
    "# import data/toutiao_cat_data.txt as a pandas dataframe\n",
    "df = pd.read_parquet(\"data/toutiao_cat_data_all_with_embeddings.parquet\")\n",
    "print(\"Lines of text before filtering: \", len(df))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>id</th>\n",
       "      <th>code</th>\n",
       "      <th>category</th>\n",
       "      <th>title</th>\n",
       "      <th>keywords</th>\n",
       "      <th>combined</th>\n",
       "      <th>n_tokens</th>\n",
       "      <th>embedding</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6551700932705387022</td>\n",
       "      <td>101</td>\n",
       "      <td>news_culture</td>\n",
       "      <td>京城最值得你来场文化之旅的博物馆</td>\n",
       "      <td>保利集团,马未都,中国科学技术馆,博物馆,新中国</td>\n",
       "      <td>标题: 京城最值得你来场文化之旅的博物馆; 关键字: 保利集团,马未都,中国科学技术馆,博物...</td>\n",
       "      <td>56</td>\n",
       "      <td>[-0.02160128578543663, -0.004615371115505695, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6552368441838272771</td>\n",
       "      <td>101</td>\n",
       "      <td>news_culture</td>\n",
       "      <td>发酵床的垫料种类有哪些？哪种更好？</td>\n",
       "      <td></td>\n",
       "      <td>标题: 发酵床的垫料种类有哪些？哪种更好？; 关键字:</td>\n",
       "      <td>32</td>\n",
       "      <td>[0.014646276831626892, -0.0075576393865048885,...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>6552407965343678723</td>\n",
       "      <td>101</td>\n",
       "      <td>news_culture</td>\n",
       "      <td>上联：黄山黄河黄皮肤黄土高原。怎么对下联？</td>\n",
       "      <td></td>\n",
       "      <td>标题: 上联：黄山黄河黄皮肤黄土高原。怎么对下联？; 关键字:</td>\n",
       "      <td>37</td>\n",
       "      <td>[0.012643996626138687, 0.0035733033437281847, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>6552332417753940238</td>\n",
       "      <td>101</td>\n",
       "      <td>news_culture</td>\n",
       "      <td>林徽因什么理由拒绝了徐志摩而选择梁思成为终身伴侣？</td>\n",
       "      <td></td>\n",
       "      <td>标题: 林徽因什么理由拒绝了徐志摩而选择梁思成为终身伴侣？; 关键字:</td>\n",
       "      <td>44</td>\n",
       "      <td>[-0.0004984583356417716, -0.016934111714363098...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6552475601595269390</td>\n",
       "      <td>101</td>\n",
       "      <td>news_culture</td>\n",
       "      <td>黄杨木是什么树？</td>\n",
       "      <td></td>\n",
       "      <td>标题: 黄杨木是什么树？; 关键字:</td>\n",
       "      <td>20</td>\n",
       "      <td>[-0.0008294825092889369, -0.013883085921406746...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    id  code      category                      title   \n",
       "0  6551700932705387022   101  news_culture           京城最值得你来场文化之旅的博物馆  \\\n",
       "1  6552368441838272771   101  news_culture          发酵床的垫料种类有哪些？哪种更好？   \n",
       "2  6552407965343678723   101  news_culture      上联：黄山黄河黄皮肤黄土高原。怎么对下联？   \n",
       "3  6552332417753940238   101  news_culture  林徽因什么理由拒绝了徐志摩而选择梁思成为终身伴侣？   \n",
       "4  6552475601595269390   101  news_culture                   黄杨木是什么树？   \n",
       "\n",
       "                   keywords   \n",
       "0  保利集团,马未都,中国科学技术馆,博物馆,新中国  \\\n",
       "1                             \n",
       "2                             \n",
       "3                             \n",
       "4                             \n",
       "\n",
       "                                            combined  n_tokens   \n",
       "0  标题: 京城最值得你来场文化之旅的博物馆; 关键字: 保利集团,马未都,中国科学技术馆,博物...        56  \\\n",
       "1                       标题: 发酵床的垫料种类有哪些？哪种更好？; 关键字:         32   \n",
       "2                   标题: 上联：黄山黄河黄皮肤黄土高原。怎么对下联？; 关键字:         37   \n",
       "3               标题: 林徽因什么理由拒绝了徐志摩而选择梁思成为终身伴侣？; 关键字:         44   \n",
       "4                                标题: 黄杨木是什么树？; 关键字:         20   \n",
       "\n",
       "                                           embedding  \n",
       "0  [-0.02160128578543663, -0.004615371115505695, ...  \n",
       "1  [0.014646276831626892, -0.0075576393865048885,...  \n",
       "2  [0.012643996626138687, 0.0035733033437281847, ...  \n",
       "3  [-0.0004984583356417716, -0.016934111714363098...  \n",
       "4  [-0.0008294825092889369, -0.013883085921406746...  "
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Lines of text after filtering:  382688\n"
     ]
    }
   ],
   "source": [
    "encoding = tiktoken.get_encoding(embedding_encoding)\n",
    "# omit reviews that are too long to embed\n",
    "df[\"n_tokens\"] = df.combined.apply(lambda x: len(encoding.encode(x)))\n",
    "df = df[df.n_tokens <= max_tokens]\n",
    "print(\"Lines of text after filtering: \", len(df))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 假设我们没有预先处理完Embedding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[99], line 5\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[39mfrom\u001b[39;00m \u001b[39mopenai\u001b[39;00m\u001b[39m.\u001b[39;00m\u001b[39membeddings_utils\u001b[39;00m \u001b[39mimport\u001b[39;00m get_embedding\n\u001b[1;32m      3\u001b[0m df_1k \u001b[39m=\u001b[39m df\u001b[39m.\u001b[39msample(\u001b[39m1000\u001b[39m, random_state\u001b[39m=\u001b[39m\u001b[39m42\u001b[39m)\n\u001b[0;32m----> 5\u001b[0m df_1k[\u001b[39m\"\u001b[39m\u001b[39membedding\u001b[39m\u001b[39m\"\u001b[39m] \u001b[39m=\u001b[39m df_1k\u001b[39m.\u001b[39;49mcombined\u001b[39m.\u001b[39;49mapply(\u001b[39mlambda\u001b[39;49;00m x : get_embedding(x, engine\u001b[39m=\u001b[39;49membedding_model))\n\u001b[1;32m      6\u001b[0m df_1k\u001b[39m.\u001b[39mto_csv(\u001b[39m\"\u001b[39m\u001b[39mdata/toutiao_cat_data_10k_with_embeddings.csv\u001b[39m\u001b[39m\"\u001b[39m, index\u001b[39m=\u001b[39m\u001b[39mFalse\u001b[39;00m)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/pandas/core/series.py:4626\u001b[0m, in \u001b[0;36mSeries.apply\u001b[0;34m(self, func, convert_dtype, args, **kwargs)\u001b[0m\n\u001b[1;32m   4516\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mapply\u001b[39m(\n\u001b[1;32m   4517\u001b[0m     \u001b[39mself\u001b[39m,\n\u001b[1;32m   4518\u001b[0m     func: AggFuncType,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   4521\u001b[0m     \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mkwargs,\n\u001b[1;32m   4522\u001b[0m ) \u001b[39m-\u001b[39m\u001b[39m>\u001b[39m DataFrame \u001b[39m|\u001b[39m Series:\n\u001b[1;32m   4523\u001b[0m \u001b[39m    \u001b[39m\u001b[39m\"\"\"\u001b[39;00m\n\u001b[1;32m   4524\u001b[0m \u001b[39m    Invoke function on values of Series.\u001b[39;00m\n\u001b[1;32m   4525\u001b[0m \n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   4624\u001b[0m \u001b[39m    dtype: float64\u001b[39;00m\n\u001b[1;32m   4625\u001b[0m \u001b[39m    \"\"\"\u001b[39;00m\n\u001b[0;32m-> 4626\u001b[0m     \u001b[39mreturn\u001b[39;00m SeriesApply(\u001b[39mself\u001b[39;49m, func, convert_dtype, args, kwargs)\u001b[39m.\u001b[39;49mapply()\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/pandas/core/apply.py:1025\u001b[0m, in \u001b[0;36mSeriesApply.apply\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   1022\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mapply_str()\n\u001b[1;32m   1024\u001b[0m \u001b[39m# self.f is Callable\u001b[39;00m\n\u001b[0;32m-> 1025\u001b[0m \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mapply_standard()\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/pandas/core/apply.py:1076\u001b[0m, in \u001b[0;36mSeriesApply.apply_standard\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   1074\u001b[0m     \u001b[39melse\u001b[39;00m:\n\u001b[1;32m   1075\u001b[0m         values \u001b[39m=\u001b[39m obj\u001b[39m.\u001b[39mastype(\u001b[39mobject\u001b[39m)\u001b[39m.\u001b[39m_values\n\u001b[0;32m-> 1076\u001b[0m         mapped \u001b[39m=\u001b[39m lib\u001b[39m.\u001b[39;49mmap_infer(\n\u001b[1;32m   1077\u001b[0m             values,\n\u001b[1;32m   1078\u001b[0m             f,\n\u001b[1;32m   1079\u001b[0m             convert\u001b[39m=\u001b[39;49m\u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mconvert_dtype,\n\u001b[1;32m   1080\u001b[0m         )\n\u001b[1;32m   1082\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mlen\u001b[39m(mapped) \u001b[39mand\u001b[39;00m \u001b[39misinstance\u001b[39m(mapped[\u001b[39m0\u001b[39m], ABCSeries):\n\u001b[1;32m   1083\u001b[0m     \u001b[39m# GH#43986 Need to do list(mapped) in order to get treated as nested\u001b[39;00m\n\u001b[1;32m   1084\u001b[0m     \u001b[39m#  See also GH#25959 regarding EA support\u001b[39;00m\n\u001b[1;32m   1085\u001b[0m     \u001b[39mreturn\u001b[39;00m obj\u001b[39m.\u001b[39m_constructor_expanddim(\u001b[39mlist\u001b[39m(mapped), index\u001b[39m=\u001b[39mobj\u001b[39m.\u001b[39mindex)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/pandas/_libs/lib.pyx:2834\u001b[0m, in \u001b[0;36mpandas._libs.lib.map_infer\u001b[0;34m()\u001b[0m\n",
      "Cell \u001b[0;32mIn[99], line 5\u001b[0m, in \u001b[0;36m<lambda>\u001b[0;34m(x)\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[39mfrom\u001b[39;00m \u001b[39mopenai\u001b[39;00m\u001b[39m.\u001b[39;00m\u001b[39membeddings_utils\u001b[39;00m \u001b[39mimport\u001b[39;00m get_embedding\n\u001b[1;32m      3\u001b[0m df_1k \u001b[39m=\u001b[39m df\u001b[39m.\u001b[39msample(\u001b[39m1000\u001b[39m, random_state\u001b[39m=\u001b[39m\u001b[39m42\u001b[39m)\n\u001b[0;32m----> 5\u001b[0m df_1k[\u001b[39m\"\u001b[39m\u001b[39membedding\u001b[39m\u001b[39m\"\u001b[39m] \u001b[39m=\u001b[39m df_1k\u001b[39m.\u001b[39mcombined\u001b[39m.\u001b[39mapply(\u001b[39mlambda\u001b[39;00m x : get_embedding(x, engine\u001b[39m=\u001b[39;49membedding_model))\n\u001b[1;32m      6\u001b[0m df_1k\u001b[39m.\u001b[39mto_csv(\u001b[39m\"\u001b[39m\u001b[39mdata/toutiao_cat_data_10k_with_embeddings.csv\u001b[39m\u001b[39m\"\u001b[39m, index\u001b[39m=\u001b[39m\u001b[39mFalse\u001b[39;00m)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/tenacity/__init__.py:289\u001b[0m, in \u001b[0;36mBaseRetrying.wraps.<locals>.wrapped_f\u001b[0;34m(*args, **kw)\u001b[0m\n\u001b[1;32m    287\u001b[0m \u001b[39m@functools\u001b[39m\u001b[39m.\u001b[39mwraps(f)\n\u001b[1;32m    288\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mwrapped_f\u001b[39m(\u001b[39m*\u001b[39margs: t\u001b[39m.\u001b[39mAny, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mkw: t\u001b[39m.\u001b[39mAny) \u001b[39m-\u001b[39m\u001b[39m>\u001b[39m t\u001b[39m.\u001b[39mAny:\n\u001b[0;32m--> 289\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m(f, \u001b[39m*\u001b[39;49margs, \u001b[39m*\u001b[39;49m\u001b[39m*\u001b[39;49mkw)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/tenacity/__init__.py:379\u001b[0m, in \u001b[0;36mRetrying.__call__\u001b[0;34m(self, fn, *args, **kwargs)\u001b[0m\n\u001b[1;32m    377\u001b[0m retry_state \u001b[39m=\u001b[39m RetryCallState(retry_object\u001b[39m=\u001b[39m\u001b[39mself\u001b[39m, fn\u001b[39m=\u001b[39mfn, args\u001b[39m=\u001b[39margs, kwargs\u001b[39m=\u001b[39mkwargs)\n\u001b[1;32m    378\u001b[0m \u001b[39mwhile\u001b[39;00m \u001b[39mTrue\u001b[39;00m:\n\u001b[0;32m--> 379\u001b[0m     do \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49miter(retry_state\u001b[39m=\u001b[39;49mretry_state)\n\u001b[1;32m    380\u001b[0m     \u001b[39mif\u001b[39;00m \u001b[39misinstance\u001b[39m(do, DoAttempt):\n\u001b[1;32m    381\u001b[0m         \u001b[39mtry\u001b[39;00m:\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/tenacity/__init__.py:314\u001b[0m, in \u001b[0;36mBaseRetrying.iter\u001b[0;34m(self, retry_state)\u001b[0m\n\u001b[1;32m    312\u001b[0m is_explicit_retry \u001b[39m=\u001b[39m fut\u001b[39m.\u001b[39mfailed \u001b[39mand\u001b[39;00m \u001b[39misinstance\u001b[39m(fut\u001b[39m.\u001b[39mexception(), TryAgain)\n\u001b[1;32m    313\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mnot\u001b[39;00m (is_explicit_retry \u001b[39mor\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mretry(retry_state)):\n\u001b[0;32m--> 314\u001b[0m     \u001b[39mreturn\u001b[39;00m fut\u001b[39m.\u001b[39;49mresult()\n\u001b[1;32m    316\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mafter \u001b[39mis\u001b[39;00m \u001b[39mnot\u001b[39;00m \u001b[39mNone\u001b[39;00m:\n\u001b[1;32m    317\u001b[0m     \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mafter(retry_state)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/concurrent/futures/_base.py:451\u001b[0m, in \u001b[0;36mFuture.result\u001b[0;34m(self, timeout)\u001b[0m\n\u001b[1;32m    449\u001b[0m     \u001b[39mraise\u001b[39;00m CancelledError()\n\u001b[1;32m    450\u001b[0m \u001b[39melif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_state \u001b[39m==\u001b[39m FINISHED:\n\u001b[0;32m--> 451\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m__get_result()\n\u001b[1;32m    453\u001b[0m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_condition\u001b[39m.\u001b[39mwait(timeout)\n\u001b[1;32m    455\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_state \u001b[39min\u001b[39;00m [CANCELLED, CANCELLED_AND_NOTIFIED]:\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/concurrent/futures/_base.py:403\u001b[0m, in \u001b[0;36mFuture.__get_result\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    401\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_exception:\n\u001b[1;32m    402\u001b[0m     \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m--> 403\u001b[0m         \u001b[39mraise\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_exception\n\u001b[1;32m    404\u001b[0m     \u001b[39mfinally\u001b[39;00m:\n\u001b[1;32m    405\u001b[0m         \u001b[39m# Break a reference cycle with the exception in self._exception\u001b[39;00m\n\u001b[1;32m    406\u001b[0m         \u001b[39mself\u001b[39m \u001b[39m=\u001b[39m \u001b[39mNone\u001b[39;00m\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/tenacity/__init__.py:382\u001b[0m, in \u001b[0;36mRetrying.__call__\u001b[0;34m(self, fn, *args, **kwargs)\u001b[0m\n\u001b[1;32m    380\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39misinstance\u001b[39m(do, DoAttempt):\n\u001b[1;32m    381\u001b[0m     \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m--> 382\u001b[0m         result \u001b[39m=\u001b[39m fn(\u001b[39m*\u001b[39;49margs, \u001b[39m*\u001b[39;49m\u001b[39m*\u001b[39;49mkwargs)\n\u001b[1;32m    383\u001b[0m     \u001b[39mexcept\u001b[39;00m \u001b[39mBaseException\u001b[39;00m:  \u001b[39m# noqa: B902\u001b[39;00m\n\u001b[1;32m    384\u001b[0m         retry_state\u001b[39m.\u001b[39mset_exception(sys\u001b[39m.\u001b[39mexc_info())  \u001b[39m# type: ignore[arg-type]\u001b[39;00m\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/openai/embeddings_utils.py:23\u001b[0m, in \u001b[0;36mget_embedding\u001b[0;34m(text, engine, **kwargs)\u001b[0m\n\u001b[1;32m     17\u001b[0m \u001b[39m@retry\u001b[39m(wait\u001b[39m=\u001b[39mwait_random_exponential(\u001b[39mmin\u001b[39m\u001b[39m=\u001b[39m\u001b[39m1\u001b[39m, \u001b[39mmax\u001b[39m\u001b[39m=\u001b[39m\u001b[39m20\u001b[39m), stop\u001b[39m=\u001b[39mstop_after_attempt(\u001b[39m6\u001b[39m))\n\u001b[1;32m     18\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mget_embedding\u001b[39m(text: \u001b[39mstr\u001b[39m, engine\u001b[39m=\u001b[39m\u001b[39m\"\u001b[39m\u001b[39mtext-similarity-davinci-001\u001b[39m\u001b[39m\"\u001b[39m, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mkwargs) \u001b[39m-\u001b[39m\u001b[39m>\u001b[39m List[\u001b[39mfloat\u001b[39m]:\n\u001b[1;32m     19\u001b[0m \n\u001b[1;32m     20\u001b[0m     \u001b[39m# replace newlines, which can negatively affect performance.\u001b[39;00m\n\u001b[1;32m     21\u001b[0m     text \u001b[39m=\u001b[39m text\u001b[39m.\u001b[39mreplace(\u001b[39m\"\u001b[39m\u001b[39m\\n\u001b[39;00m\u001b[39m\"\u001b[39m, \u001b[39m\"\u001b[39m\u001b[39m \u001b[39m\u001b[39m\"\u001b[39m)\n\u001b[0;32m---> 23\u001b[0m     \u001b[39mreturn\u001b[39;00m openai\u001b[39m.\u001b[39;49mEmbedding\u001b[39m.\u001b[39;49mcreate(\u001b[39minput\u001b[39;49m\u001b[39m=\u001b[39;49m[text], engine\u001b[39m=\u001b[39;49mengine, \u001b[39m*\u001b[39;49m\u001b[39m*\u001b[39;49mkwargs)[\u001b[39m\"\u001b[39m\u001b[39mdata\u001b[39m\u001b[39m\"\u001b[39m][\u001b[39m0\u001b[39m][\u001b[39m\"\u001b[39m\u001b[39membedding\u001b[39m\u001b[39m\"\u001b[39m]\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/openai/api_resources/embedding.py:33\u001b[0m, in \u001b[0;36mEmbedding.create\u001b[0;34m(cls, *args, **kwargs)\u001b[0m\n\u001b[1;32m     31\u001b[0m \u001b[39mwhile\u001b[39;00m \u001b[39mTrue\u001b[39;00m:\n\u001b[1;32m     32\u001b[0m     \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m---> 33\u001b[0m         response \u001b[39m=\u001b[39m \u001b[39msuper\u001b[39;49m()\u001b[39m.\u001b[39;49mcreate(\u001b[39m*\u001b[39;49margs, \u001b[39m*\u001b[39;49m\u001b[39m*\u001b[39;49mkwargs)\n\u001b[1;32m     35\u001b[0m         \u001b[39m# If a user specifies base64, we'll just return the encoded string.\u001b[39;00m\n\u001b[1;32m     36\u001b[0m         \u001b[39m# This is only for the default case.\u001b[39;00m\n\u001b[1;32m     37\u001b[0m         \u001b[39mif\u001b[39;00m \u001b[39mnot\u001b[39;00m user_provided_encoding_format:\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/openai/api_resources/abstract/engine_api_resource.py:153\u001b[0m, in \u001b[0;36mEngineAPIResource.create\u001b[0;34m(cls, api_key, api_base, api_type, request_id, api_version, organization, **params)\u001b[0m\n\u001b[1;32m    127\u001b[0m \u001b[39m@classmethod\u001b[39m\n\u001b[1;32m    128\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mcreate\u001b[39m(\n\u001b[1;32m    129\u001b[0m     \u001b[39mcls\u001b[39m,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    136\u001b[0m     \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mparams,\n\u001b[1;32m    137\u001b[0m ):\n\u001b[1;32m    138\u001b[0m     (\n\u001b[1;32m    139\u001b[0m         deployment_id,\n\u001b[1;32m    140\u001b[0m         engine,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    150\u001b[0m         api_key, api_base, api_type, api_version, organization, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mparams\n\u001b[1;32m    151\u001b[0m     )\n\u001b[0;32m--> 153\u001b[0m     response, _, api_key \u001b[39m=\u001b[39m requestor\u001b[39m.\u001b[39;49mrequest(\n\u001b[1;32m    154\u001b[0m         \u001b[39m\"\u001b[39;49m\u001b[39mpost\u001b[39;49m\u001b[39m\"\u001b[39;49m,\n\u001b[1;32m    155\u001b[0m         url,\n\u001b[1;32m    156\u001b[0m         params\u001b[39m=\u001b[39;49mparams,\n\u001b[1;32m    157\u001b[0m         headers\u001b[39m=\u001b[39;49mheaders,\n\u001b[1;32m    158\u001b[0m         stream\u001b[39m=\u001b[39;49mstream,\n\u001b[1;32m    159\u001b[0m         request_id\u001b[39m=\u001b[39;49mrequest_id,\n\u001b[1;32m    160\u001b[0m         request_timeout\u001b[39m=\u001b[39;49mrequest_timeout,\n\u001b[1;32m    161\u001b[0m     )\n\u001b[1;32m    163\u001b[0m     \u001b[39mif\u001b[39;00m stream:\n\u001b[1;32m    164\u001b[0m         \u001b[39m# must be an iterator\u001b[39;00m\n\u001b[1;32m    165\u001b[0m         \u001b[39massert\u001b[39;00m \u001b[39mnot\u001b[39;00m \u001b[39misinstance\u001b[39m(response, OpenAIResponse)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/openai/api_requestor.py:220\u001b[0m, in \u001b[0;36mAPIRequestor.request\u001b[0;34m(self, method, url, params, headers, files, stream, request_id, request_timeout)\u001b[0m\n\u001b[1;32m    209\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mrequest\u001b[39m(\n\u001b[1;32m    210\u001b[0m     \u001b[39mself\u001b[39m,\n\u001b[1;32m    211\u001b[0m     method,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    218\u001b[0m     request_timeout: Optional[Union[\u001b[39mfloat\u001b[39m, Tuple[\u001b[39mfloat\u001b[39m, \u001b[39mfloat\u001b[39m]]] \u001b[39m=\u001b[39m \u001b[39mNone\u001b[39;00m,\n\u001b[1;32m    219\u001b[0m ) \u001b[39m-\u001b[39m\u001b[39m>\u001b[39m Tuple[Union[OpenAIResponse, Iterator[OpenAIResponse]], \u001b[39mbool\u001b[39m, \u001b[39mstr\u001b[39m]:\n\u001b[0;32m--> 220\u001b[0m     result \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mrequest_raw(\n\u001b[1;32m    221\u001b[0m         method\u001b[39m.\u001b[39;49mlower(),\n\u001b[1;32m    222\u001b[0m         url,\n\u001b[1;32m    223\u001b[0m         params\u001b[39m=\u001b[39;49mparams,\n\u001b[1;32m    224\u001b[0m         supplied_headers\u001b[39m=\u001b[39;49mheaders,\n\u001b[1;32m    225\u001b[0m         files\u001b[39m=\u001b[39;49mfiles,\n\u001b[1;32m    226\u001b[0m         stream\u001b[39m=\u001b[39;49mstream,\n\u001b[1;32m    227\u001b[0m         request_id\u001b[39m=\u001b[39;49mrequest_id,\n\u001b[1;32m    228\u001b[0m         request_timeout\u001b[39m=\u001b[39;49mrequest_timeout,\n\u001b[1;32m    229\u001b[0m     )\n\u001b[1;32m    230\u001b[0m     resp, got_stream \u001b[39m=\u001b[39m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_interpret_response(result, stream)\n\u001b[1;32m    231\u001b[0m     \u001b[39mreturn\u001b[39;00m resp, got_stream, \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mapi_key\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/openai/api_requestor.py:520\u001b[0m, in \u001b[0;36mAPIRequestor.request_raw\u001b[0;34m(self, method, url, params, supplied_headers, files, stream, request_id, request_timeout)\u001b[0m\n\u001b[1;32m    518\u001b[0m     _thread_context\u001b[39m.\u001b[39msession \u001b[39m=\u001b[39m _make_session()\n\u001b[1;32m    519\u001b[0m \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m--> 520\u001b[0m     result \u001b[39m=\u001b[39m _thread_context\u001b[39m.\u001b[39;49msession\u001b[39m.\u001b[39;49mrequest(\n\u001b[1;32m    521\u001b[0m         method,\n\u001b[1;32m    522\u001b[0m         abs_url,\n\u001b[1;32m    523\u001b[0m         headers\u001b[39m=\u001b[39;49mheaders,\n\u001b[1;32m    524\u001b[0m         data\u001b[39m=\u001b[39;49mdata,\n\u001b[1;32m    525\u001b[0m         files\u001b[39m=\u001b[39;49mfiles,\n\u001b[1;32m    526\u001b[0m         stream\u001b[39m=\u001b[39;49mstream,\n\u001b[1;32m    527\u001b[0m         timeout\u001b[39m=\u001b[39;49mrequest_timeout \u001b[39mif\u001b[39;49;00m request_timeout \u001b[39melse\u001b[39;49;00m TIMEOUT_SECS,\n\u001b[1;32m    528\u001b[0m         proxies\u001b[39m=\u001b[39;49m_thread_context\u001b[39m.\u001b[39;49msession\u001b[39m.\u001b[39;49mproxies,\n\u001b[1;32m    529\u001b[0m     )\n\u001b[1;32m    530\u001b[0m \u001b[39mexcept\u001b[39;00m requests\u001b[39m.\u001b[39mexceptions\u001b[39m.\u001b[39mTimeout \u001b[39mas\u001b[39;00m e:\n\u001b[1;32m    531\u001b[0m     \u001b[39mraise\u001b[39;00m error\u001b[39m.\u001b[39mTimeout(\u001b[39m\"\u001b[39m\u001b[39mRequest timed out: \u001b[39m\u001b[39m{}\u001b[39;00m\u001b[39m\"\u001b[39m\u001b[39m.\u001b[39mformat(e)) \u001b[39mfrom\u001b[39;00m \u001b[39me\u001b[39;00m\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/requests/sessions.py:589\u001b[0m, in \u001b[0;36mSession.request\u001b[0;34m(self, method, url, params, data, headers, cookies, files, auth, timeout, allow_redirects, proxies, hooks, stream, verify, cert, json)\u001b[0m\n\u001b[1;32m    584\u001b[0m send_kwargs \u001b[39m=\u001b[39m {\n\u001b[1;32m    585\u001b[0m     \u001b[39m\"\u001b[39m\u001b[39mtimeout\u001b[39m\u001b[39m\"\u001b[39m: timeout,\n\u001b[1;32m    586\u001b[0m     \u001b[39m\"\u001b[39m\u001b[39mallow_redirects\u001b[39m\u001b[39m\"\u001b[39m: allow_redirects,\n\u001b[1;32m    587\u001b[0m }\n\u001b[1;32m    588\u001b[0m send_kwargs\u001b[39m.\u001b[39mupdate(settings)\n\u001b[0;32m--> 589\u001b[0m resp \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49msend(prep, \u001b[39m*\u001b[39;49m\u001b[39m*\u001b[39;49msend_kwargs)\n\u001b[1;32m    591\u001b[0m \u001b[39mreturn\u001b[39;00m resp\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/requests/sessions.py:703\u001b[0m, in \u001b[0;36mSession.send\u001b[0;34m(self, request, **kwargs)\u001b[0m\n\u001b[1;32m    700\u001b[0m start \u001b[39m=\u001b[39m preferred_clock()\n\u001b[1;32m    702\u001b[0m \u001b[39m# Send the request\u001b[39;00m\n\u001b[0;32m--> 703\u001b[0m r \u001b[39m=\u001b[39m adapter\u001b[39m.\u001b[39;49msend(request, \u001b[39m*\u001b[39;49m\u001b[39m*\u001b[39;49mkwargs)\n\u001b[1;32m    705\u001b[0m \u001b[39m# Total elapsed time of the request (approximately)\u001b[39;00m\n\u001b[1;32m    706\u001b[0m elapsed \u001b[39m=\u001b[39m preferred_clock() \u001b[39m-\u001b[39m start\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/requests/adapters.py:486\u001b[0m, in \u001b[0;36mHTTPAdapter.send\u001b[0;34m(self, request, stream, timeout, verify, cert, proxies)\u001b[0m\n\u001b[1;32m    483\u001b[0m     timeout \u001b[39m=\u001b[39m TimeoutSauce(connect\u001b[39m=\u001b[39mtimeout, read\u001b[39m=\u001b[39mtimeout)\n\u001b[1;32m    485\u001b[0m \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m--> 486\u001b[0m     resp \u001b[39m=\u001b[39m conn\u001b[39m.\u001b[39;49murlopen(\n\u001b[1;32m    487\u001b[0m         method\u001b[39m=\u001b[39;49mrequest\u001b[39m.\u001b[39;49mmethod,\n\u001b[1;32m    488\u001b[0m         url\u001b[39m=\u001b[39;49murl,\n\u001b[1;32m    489\u001b[0m         body\u001b[39m=\u001b[39;49mrequest\u001b[39m.\u001b[39;49mbody,\n\u001b[1;32m    490\u001b[0m         headers\u001b[39m=\u001b[39;49mrequest\u001b[39m.\u001b[39;49mheaders,\n\u001b[1;32m    491\u001b[0m         redirect\u001b[39m=\u001b[39;49m\u001b[39mFalse\u001b[39;49;00m,\n\u001b[1;32m    492\u001b[0m         assert_same_host\u001b[39m=\u001b[39;49m\u001b[39mFalse\u001b[39;49;00m,\n\u001b[1;32m    493\u001b[0m         preload_content\u001b[39m=\u001b[39;49m\u001b[39mFalse\u001b[39;49;00m,\n\u001b[1;32m    494\u001b[0m         decode_content\u001b[39m=\u001b[39;49m\u001b[39mFalse\u001b[39;49;00m,\n\u001b[1;32m    495\u001b[0m         retries\u001b[39m=\u001b[39;49m\u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mmax_retries,\n\u001b[1;32m    496\u001b[0m         timeout\u001b[39m=\u001b[39;49mtimeout,\n\u001b[1;32m    497\u001b[0m         chunked\u001b[39m=\u001b[39;49mchunked,\n\u001b[1;32m    498\u001b[0m     )\n\u001b[1;32m    500\u001b[0m \u001b[39mexcept\u001b[39;00m (ProtocolError, \u001b[39mOSError\u001b[39;00m) \u001b[39mas\u001b[39;00m err:\n\u001b[1;32m    501\u001b[0m     \u001b[39mraise\u001b[39;00m \u001b[39mConnectionError\u001b[39;00m(err, request\u001b[39m=\u001b[39mrequest)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/urllib3/connectionpool.py:790\u001b[0m, in \u001b[0;36mHTTPConnectionPool.urlopen\u001b[0;34m(self, method, url, body, headers, retries, redirect, assert_same_host, timeout, pool_timeout, release_conn, chunked, body_pos, preload_content, decode_content, **response_kw)\u001b[0m\n\u001b[1;32m    787\u001b[0m response_conn \u001b[39m=\u001b[39m conn \u001b[39mif\u001b[39;00m \u001b[39mnot\u001b[39;00m release_conn \u001b[39melse\u001b[39;00m \u001b[39mNone\u001b[39;00m\n\u001b[1;32m    789\u001b[0m \u001b[39m# Make the request on the HTTPConnection object\u001b[39;00m\n\u001b[0;32m--> 790\u001b[0m response \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_make_request(\n\u001b[1;32m    791\u001b[0m     conn,\n\u001b[1;32m    792\u001b[0m     method,\n\u001b[1;32m    793\u001b[0m     url,\n\u001b[1;32m    794\u001b[0m     timeout\u001b[39m=\u001b[39;49mtimeout_obj,\n\u001b[1;32m    795\u001b[0m     body\u001b[39m=\u001b[39;49mbody,\n\u001b[1;32m    796\u001b[0m     headers\u001b[39m=\u001b[39;49mheaders,\n\u001b[1;32m    797\u001b[0m     chunked\u001b[39m=\u001b[39;49mchunked,\n\u001b[1;32m    798\u001b[0m     retries\u001b[39m=\u001b[39;49mretries,\n\u001b[1;32m    799\u001b[0m     response_conn\u001b[39m=\u001b[39;49mresponse_conn,\n\u001b[1;32m    800\u001b[0m     preload_content\u001b[39m=\u001b[39;49mpreload_content,\n\u001b[1;32m    801\u001b[0m     decode_content\u001b[39m=\u001b[39;49mdecode_content,\n\u001b[1;32m    802\u001b[0m     \u001b[39m*\u001b[39;49m\u001b[39m*\u001b[39;49mresponse_kw,\n\u001b[1;32m    803\u001b[0m )\n\u001b[1;32m    805\u001b[0m \u001b[39m# Everything went great!\u001b[39;00m\n\u001b[1;32m    806\u001b[0m clean_exit \u001b[39m=\u001b[39m \u001b[39mTrue\u001b[39;00m\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/urllib3/connectionpool.py:536\u001b[0m, in \u001b[0;36mHTTPConnectionPool._make_request\u001b[0;34m(self, conn, method, url, body, headers, retries, timeout, chunked, response_conn, preload_content, decode_content, enforce_content_length)\u001b[0m\n\u001b[1;32m    534\u001b[0m \u001b[39m# Receive the response from the server\u001b[39;00m\n\u001b[1;32m    535\u001b[0m \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m--> 536\u001b[0m     response \u001b[39m=\u001b[39m conn\u001b[39m.\u001b[39;49mgetresponse()\n\u001b[1;32m    537\u001b[0m \u001b[39mexcept\u001b[39;00m (BaseSSLError, \u001b[39mOSError\u001b[39;00m) \u001b[39mas\u001b[39;00m e:\n\u001b[1;32m    538\u001b[0m     \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_raise_timeout(err\u001b[39m=\u001b[39me, url\u001b[39m=\u001b[39murl, timeout_value\u001b[39m=\u001b[39mread_timeout)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/site-packages/urllib3/connection.py:454\u001b[0m, in \u001b[0;36mHTTPConnection.getresponse\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    451\u001b[0m \u001b[39mfrom\u001b[39;00m \u001b[39m.\u001b[39;00m\u001b[39mresponse\u001b[39;00m \u001b[39mimport\u001b[39;00m HTTPResponse\n\u001b[1;32m    453\u001b[0m \u001b[39m# Get the response from http.client.HTTPConnection\u001b[39;00m\n\u001b[0;32m--> 454\u001b[0m httplib_response \u001b[39m=\u001b[39m \u001b[39msuper\u001b[39;49m()\u001b[39m.\u001b[39;49mgetresponse()\n\u001b[1;32m    456\u001b[0m \u001b[39mtry\u001b[39;00m:\n\u001b[1;32m    457\u001b[0m     assert_header_parsing(httplib_response\u001b[39m.\u001b[39mmsg)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/http/client.py:1375\u001b[0m, in \u001b[0;36mHTTPConnection.getresponse\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m   1373\u001b[0m \u001b[39mtry\u001b[39;00m:\n\u001b[1;32m   1374\u001b[0m     \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m-> 1375\u001b[0m         response\u001b[39m.\u001b[39;49mbegin()\n\u001b[1;32m   1376\u001b[0m     \u001b[39mexcept\u001b[39;00m \u001b[39mConnectionError\u001b[39;00m:\n\u001b[1;32m   1377\u001b[0m         \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mclose()\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/http/client.py:318\u001b[0m, in \u001b[0;36mHTTPResponse.begin\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    316\u001b[0m \u001b[39m# read until we get a non-100 response\u001b[39;00m\n\u001b[1;32m    317\u001b[0m \u001b[39mwhile\u001b[39;00m \u001b[39mTrue\u001b[39;00m:\n\u001b[0;32m--> 318\u001b[0m     version, status, reason \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_read_status()\n\u001b[1;32m    319\u001b[0m     \u001b[39mif\u001b[39;00m status \u001b[39m!=\u001b[39m CONTINUE:\n\u001b[1;32m    320\u001b[0m         \u001b[39mbreak\u001b[39;00m\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/http/client.py:279\u001b[0m, in \u001b[0;36mHTTPResponse._read_status\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    278\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39m_read_status\u001b[39m(\u001b[39mself\u001b[39m):\n\u001b[0;32m--> 279\u001b[0m     line \u001b[39m=\u001b[39m \u001b[39mstr\u001b[39m(\u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mfp\u001b[39m.\u001b[39;49mreadline(_MAXLINE \u001b[39m+\u001b[39;49m \u001b[39m1\u001b[39;49m), \u001b[39m\"\u001b[39m\u001b[39miso-8859-1\u001b[39m\u001b[39m\"\u001b[39m)\n\u001b[1;32m    280\u001b[0m     \u001b[39mif\u001b[39;00m \u001b[39mlen\u001b[39m(line) \u001b[39m>\u001b[39m _MAXLINE:\n\u001b[1;32m    281\u001b[0m         \u001b[39mraise\u001b[39;00m LineTooLong(\u001b[39m\"\u001b[39m\u001b[39mstatus line\u001b[39m\u001b[39m\"\u001b[39m)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/socket.py:705\u001b[0m, in \u001b[0;36mSocketIO.readinto\u001b[0;34m(self, b)\u001b[0m\n\u001b[1;32m    703\u001b[0m \u001b[39mwhile\u001b[39;00m \u001b[39mTrue\u001b[39;00m:\n\u001b[1;32m    704\u001b[0m     \u001b[39mtry\u001b[39;00m:\n\u001b[0;32m--> 705\u001b[0m         \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_sock\u001b[39m.\u001b[39;49mrecv_into(b)\n\u001b[1;32m    706\u001b[0m     \u001b[39mexcept\u001b[39;00m timeout:\n\u001b[1;32m    707\u001b[0m         \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_timeout_occurred \u001b[39m=\u001b[39m \u001b[39mTrue\u001b[39;00m\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/ssl.py:1274\u001b[0m, in \u001b[0;36mSSLSocket.recv_into\u001b[0;34m(self, buffer, nbytes, flags)\u001b[0m\n\u001b[1;32m   1270\u001b[0m     \u001b[39mif\u001b[39;00m flags \u001b[39m!=\u001b[39m \u001b[39m0\u001b[39m:\n\u001b[1;32m   1271\u001b[0m         \u001b[39mraise\u001b[39;00m \u001b[39mValueError\u001b[39;00m(\n\u001b[1;32m   1272\u001b[0m           \u001b[39m\"\u001b[39m\u001b[39mnon-zero flags not allowed in calls to recv_into() on \u001b[39m\u001b[39m%s\u001b[39;00m\u001b[39m\"\u001b[39m \u001b[39m%\u001b[39m\n\u001b[1;32m   1273\u001b[0m           \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m\u001b[39m__class__\u001b[39m)\n\u001b[0;32m-> 1274\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mread(nbytes, buffer)\n\u001b[1;32m   1275\u001b[0m \u001b[39melse\u001b[39;00m:\n\u001b[1;32m   1276\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39msuper\u001b[39m()\u001b[39m.\u001b[39mrecv_into(buffer, nbytes, flags)\n",
      "File \u001b[0;32m~/miniconda3/envs/mactalk/lib/python3.10/ssl.py:1130\u001b[0m, in \u001b[0;36mSSLSocket.read\u001b[0;34m(self, len, buffer)\u001b[0m\n\u001b[1;32m   1128\u001b[0m \u001b[39mtry\u001b[39;00m:\n\u001b[1;32m   1129\u001b[0m     \u001b[39mif\u001b[39;00m buffer \u001b[39mis\u001b[39;00m \u001b[39mnot\u001b[39;00m \u001b[39mNone\u001b[39;00m:\n\u001b[0;32m-> 1130\u001b[0m         \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_sslobj\u001b[39m.\u001b[39;49mread(\u001b[39mlen\u001b[39;49m, buffer)\n\u001b[1;32m   1131\u001b[0m     \u001b[39melse\u001b[39;00m:\n\u001b[1;32m   1132\u001b[0m         \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_sslobj\u001b[39m.\u001b[39mread(\u001b[39mlen\u001b[39m)\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "from openai.embeddings_utils import get_embedding\n",
    "\n",
    "df_1k = df.sample(1000, random_state=42)\n",
    "\n",
    "df_1k[\"embedding\"] = df_1k.combined.apply(lambda x : get_embedding(x, engine=embedding_model))\n",
    "df_1k.to_csv(\"data/toutiao_cat_data_10k_with_embeddings.csv\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "@backoff.on_exception(backoff.expo, openai.error.RateLimitError) # type: ignore\n",
    "def get_embedding_with_backoff(**kwargs):\n",
    "    return get_embedding(**kwargs)\n",
    "\n",
    "df_10k = df.sample(10000, random_state=42)\n",
    "\n",
    "df_10k[\"embedding\"] = df_10k.combined.apply(lambda x : get_embedding_with_backoff(text=x, engine=embedding_model))\n",
    "df_10k.to_csv(\"data/toutiao_cat_data_10k_with_embeddings.csv\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [],
   "source": [
    "from openai.embeddings_utils import get_embeddings\n",
    "\n",
    "batch_size = 100\n",
    "\n",
    "@backoff.on_exception(backoff.expo, openai.error.RateLimitError)  # type: ignore\n",
    "def get_embeddings_with_backoff(prompts, engine):\n",
    "    embeddings = []\n",
    "    for i in range(0, len(prompts), batch_size):\n",
    "        batch = prompts[i:i+batch_size]\n",
    "        embeddings += get_embeddings(list_of_text=batch, engine=engine)\n",
    "    return embeddings\n",
    "\n",
    "# group prompts into batches of 100\n",
    "prompts = df_1k.combined.tolist()\n",
    "prompt_batches = [prompts[i:i+batch_size] for i in range(0, len(prompts), batch_size)]\n",
    "\n",
    "embeddings = []\n",
    "for batch in prompt_batches:\n",
    "    batch_embeddings = get_embeddings_with_backoff(prompts=batch, engine=embedding_model)\n",
    "    embeddings += batch_embeddings\n",
    "\n",
    "df_1k[\"embedding\"] = embeddings\n",
    "df_1k.to_csv(\"data/toutiao_cat_data_all_with_embeddings_1k.csv\", index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 通过Tiktoken来统计Token数量"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8\n"
     ]
    }
   ],
   "source": [
    "import tiktoken\n",
    "\n",
    "\n",
    "embedding_encoding = \"cl100k_base\"  # this the encoding for text-embedding-ada-002\n",
    "\n",
    "encoding = tiktoken.get_encoding(embedding_encoding)\n",
    "text = \"今天天气真不错\"\n",
    "print(len(encoding.encode(text)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12\n"
     ]
    }
   ],
   "source": [
    "\n",
    "embedding_encoding = \"p50k_base\"  # this the encoding for text-davinci-003\n",
    "\n",
    "encoding = tiktoken.get_encoding(embedding_encoding)\n",
    "text = \"今天天气真不错\"\n",
    "print(len(encoding.encode(text)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{\n",
      "  \"choices\": [\n",
      "    {\n",
      "      \"finish_reason\": \"stop\",\n",
      "      \"index\": 0,\n",
      "      \"logprobs\": null,\n",
      "      \"text\": \"\\n\\nYes, the weather is really nice today.\"\n",
      "    }\n",
      "  ],\n",
      "  \"created\": 1686992682,\n",
      "  \"id\": \"cmpl-7SM9Kr4ILMfYVwkFR6pI7oXCUjoBr\",\n",
      "  \"model\": \"text-davinci-003\",\n",
      "  \"object\": \"text_completion\",\n",
      "  \"usage\": {\n",
      "    \"completion_tokens\": 11,\n",
      "    \"prompt_tokens\": 12,\n",
      "    \"total_tokens\": 23\n",
      "  }\n",
      "}\n",
      "\n",
      "\n",
      "Yes, the weather is really nice today.\n"
     ]
    }
   ],
   "source": [
    "import openai\n",
    "import os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "COMPLETION_MODEL = \"text-davinci-003\"\n",
    "\n",
    "def get_response(prompt):\n",
    "    completions = openai.Completion.create(\n",
    "        engine=COMPLETION_MODEL,\n",
    "        prompt=prompt,\n",
    "        max_tokens=512,\n",
    "        n=1,\n",
    "        stop=None,\n",
    "        temperature=0.0,        \n",
    "    )\n",
    "    return completions\n",
    "\n",
    "response = get_response(\"今天天气真不错\")\n",
    "\n",
    "print(response)\n",
    "print(response.choices[0].text) # type: ignore"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 语义搜索"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "1.【新品】Apple/苹果 iPhone 11 Pro Max 手机全网通\n",
      "2.【限时特价】华为 Mate 30 Pro 5G 8GB+256GB 全网通\n",
      "3.【新款上市】Vivo iQOO Pro 5G 手机 8GB+128GB 全网通\n",
      "4.【热卖】OPPO Reno 10X Zoom 12GB+256GB 全网通\n",
      "5.【新品】小米CC9 Pro 全网通 6GB+128GB\n",
      "6.【热销】华为MateBook X Pro 2019 i7 16G 512G\n",
      "7.【限时特惠】Apple/苹果 MacBook Pro 13.3英寸 512G\n",
      "8.【新品】三星Galaxy Tab S6 10.5英寸 128G WiFi\n",
      "9.【热销】ThinkPad X1 Carbon 2019 14英寸 512G\n",
      "10.【新款上市】联想ThinkPad X1 Yoga 2019 14英寸 512G\n",
      "11.【热卖】华硕ZenBook Pro 15.6英寸 512G\n",
      "12.【新品】惠普Spectre x360 15.6英寸 512G\n",
      "13.【限时特惠】联想ThinkPad E590 15.6英寸 512G\n",
      "14.【新款上市】华为MateBook 13.9英寸 512G\n",
      "15.【热销】华为MateBook X Pro 13.9英寸 512G\n",
      "16.【新品】华为MateBook D 14英寸 512G\n",
      "17.【热卖】联想ThinkPad T480s 14英寸 512G\n",
      "18.【限时特惠】华为MateBook E 12英寸 512G\n",
      "19.【新款上市】联想ThinkPad X1 Carbon 14英寸 512G\n",
      "20.【热销】华硕ZenBook UX430 14英寸 512G\n",
      "21.【新品】华为MateBook X Pro 13.9英寸 1TB\n",
      "22.【热卖】联想ThinkPad X1 Yoga 14英寸 1TB\n",
      "23.【限时特惠】惠普Spectre x360 15.6英寸 1TB\n",
      "24.【新款上市】华硕ZenBook Pro 15.6英寸 1TB\n",
      "25.【热销】联想ThinkPad T480s 14英寸 1TB\n",
      "26.【新品】联想ThinkPad E590 15.6英寸 1TB\n",
      "27.【热卖】华为MateBook D 14英寸 1TB\n",
      "28.【限时特惠】华为MateBook 13.9英寸 1TB\n",
      "29.【新款上市】联想ThinkPad X1 Carbon 14英寸 1TB\n",
      "30.【热销】华硕ZenBook UX430 14英寸 1TB\n",
      "31.【新品】Apple/苹果 iPad Pro 11英寸 256G\n",
      "32.【热卖】三星Galaxy Tab S6 10.5英寸 256G WiFi\n",
      "33.【限时特惠】Apple/苹果 iPad Pro 12.9英寸 512G\n",
      "34.【新款上市】华为MediaPad M6 10.8英寸 256G WiFi\n",
      "35.【热销】华为MediaPad M5 10.8英寸 128G WiFi\n",
      "36.【新品】三星Galaxy Tab S4 10.5英寸 256G WiFi\n",
      "37.【热卖】联想Yoga Tab 3 Pro 10.1英寸 64G\n",
      "38.【限时特惠】小米平板4 10.1英寸 64G WiFi\n",
      "39.【新款上市】三星Galaxy Tab A 10.1英寸 32G WiFi\n",
      "40.【热销】联想Tab 4 10 10.1英寸 16G WiFi\n",
      "41.【新品】华为MateBook D 14英寸 i5 8G 256G\n",
      "42.【热卖】华为MateBook X Pro 13.9英寸 i7 16G 512G\n",
      "43.【限时特惠】联想ThinkPad X1 Carbon 14英寸 i7 16G 512G\n",
      "44.【新款上市】华硕ZenBook UX430 14英寸 i7 16G 512G\n",
      "45.【热销】联想ThinkPad E590 15.6英寸 i7 16G 512G\n",
      "46.【新品】联想ThinkPad T480s 14英寸 i7 16G 512G\n",
      "47.【热卖】惠普Spectre x360 15.6英寸 i7 16G 512G\n",
      "48.【限时特惠】华硕ZenBook Pro 15.6英寸 i7 16G 512G\n",
      "49.【新款上市】联想ThinkPad X1 Yoga 14英寸 i7 16G 512G\n",
      "50.【热销】Apple/苹果 MacBook Pro 13.3英寸 i7 16G 512G\n"
     ]
    }
   ],
   "source": [
    "import openai, os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "\n",
    "COMPLETION_MODEL = \"text-davinci-003\"\n",
    "\n",
    "def generate_data_by_prompt(prompt):\n",
    "    response = openai.Completion.create(\n",
    "        engine=COMPLETION_MODEL,\n",
    "        prompt=prompt,\n",
    "        temperature=0.5,\n",
    "        max_tokens=2048,\n",
    "        top_p=1,\n",
    "    )\n",
    "    return response.choices[0].text # type: ignore\n",
    "\n",
    "prompt = \"\"\"请你生成50条淘宝网里的商品的标题，每条在30个字左右，品类是3C数码产品，标题里往往也会有一些促销类的信息，每行一条。\"\"\"\n",
    "data = generate_data_by_prompt(prompt)\n",
    "print(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>product_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1.【新品】Apple/苹果 iPhone 11 Pro Max 手机全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2.【限时特价】华为 Mate 30 Pro 5G 8GB+256GB 全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3.【新款上市】Vivo iQOO Pro 5G 手机 8GB+128GB 全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4.【热卖】OPPO Reno 10X Zoom 12GB+256GB 全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5.【新品】小米CC9 Pro 全网通 6GB+128GB</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                product_name\n",
       "0     1.【新品】Apple/苹果 iPhone 11 Pro Max 手机全网通\n",
       "1    2.【限时特价】华为 Mate 30 Pro 5G 8GB+256GB 全网通\n",
       "2  3.【新款上市】Vivo iQOO Pro 5G 手机 8GB+128GB 全网通\n",
       "3    4.【热卖】OPPO Reno 10X Zoom 12GB+256GB 全网通\n",
       "4              5.【新品】小米CC9 Pro 全网通 6GB+128GB"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "product_names = data.strip().split('\\n')\n",
    "df = pd.DataFrame({'product_name': product_names})\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>product_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>【新品】Apple/苹果 iPhone 11 Pro Max 手机全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>【限时特价】华为 Mate 30 Pro 5G 8GB+256GB 全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>【新款上市】Vivo iQOO Pro 5G 手机 8GB+128GB 全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>【热卖】OPPO Reno 10X Zoom 12GB+256GB 全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>【新品】小米CC9 Pro 全网通 6GB+128GB</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                              product_name\n",
       "0     【新品】Apple/苹果 iPhone 11 Pro Max 手机全网通\n",
       "1    【限时特价】华为 Mate 30 Pro 5G 8GB+256GB 全网通\n",
       "2  【新款上市】Vivo iQOO Pro 5G 手机 8GB+128GB 全网通\n",
       "3    【热卖】OPPO Reno 10X Zoom 12GB+256GB 全网通\n",
       "4              【新品】小米CC9 Pro 全网通 6GB+128GB"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.product_name = df.product_name.apply(lambda x: x.split('.')[1].strip())\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "clothes_prompt = \"\"\"请你生成50条淘宝网里的商品的标题，每条在30个字左右，品类是女性的服饰箱包等等，标题里往往也会有一些促销类的信息，每行一条。\"\"\"\n",
    "clothes_data = generate_data_by_prompt(clothes_prompt)\n",
    "clothes_product_names = clothes_data.strip().split('\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>product_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>【热卖】新款时尚潮流女士手提包</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>【特价】清新简约轻便双肩包</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>【精选】百搭经典款时尚拼接包</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>【专享】简约大容量真皮挎包</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>【热卖】潮流拼接复古款女士手拿包</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       product_name\n",
       "0   【热卖】新款时尚潮流女士手提包\n",
       "1     【特价】清新简约轻便双肩包\n",
       "2    【精选】百搭经典款时尚拼接包\n",
       "3     【专享】简约大容量真皮挎包\n",
       "4  【热卖】潮流拼接复古款女士手拿包"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clothes_df = pd.DataFrame({'product_name': clothes_product_names})\n",
    "clothes_df.product_name = clothes_df.product_name.apply(lambda x: x.split('.')[1].strip())\n",
    "clothes_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0        【新品】Apple/苹果 iPhone 11 Pro Max 手机全网通\n",
      "1       【限时特价】华为 Mate 30 Pro 5G 8GB+256GB 全网通\n",
      "2     【新款上市】Vivo iQOO Pro 5G 手机 8GB+128GB 全网通\n",
      "3       【热卖】OPPO Reno 10X Zoom 12GB+256GB 全网通\n",
      "4                 【新品】小米CC9 Pro 全网通 6GB+128GB\n",
      "                       ...                   \n",
      "95                             【特价】百搭时尚复古款双肩包\n",
      "96                           【精选】清新拼接复古款女士手拿包\n",
      "97                           【专享】潮流拼接复古款真皮手提包\n",
      "98                           【热卖】简约款时尚拼接女士手拿包\n",
      "99                             【特价】复古款拼接真皮手提包\n",
      "Name: product_name, Length: 100, dtype: object\n"
     ]
    }
   ],
   "source": [
    "df = pd.concat([df, clothes_df], axis=0)\n",
    "df = df.reset_index(drop=True)\n",
    "print(df.product_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>product_name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>【新品】Apple/苹果 iPhone 11 Pro Max 手机全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>【限时特价】华为 Mate 30 Pro 5G 8GB+256GB 全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>【新款上市】Vivo iQOO Pro 5G 手机 8GB+128GB 全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>【热卖】OPPO Reno 10X Zoom 12GB+256GB 全网通</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>【新品】小米CC9 Pro 全网通 6GB+128GB</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>95</th>\n",
       "      <td>【特价】百搭时尚复古款双肩包</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>96</th>\n",
       "      <td>【精选】清新拼接复古款女士手拿包</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>97</th>\n",
       "      <td>【专享】潮流拼接复古款真皮手提包</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98</th>\n",
       "      <td>【热卖】简约款时尚拼接女士手拿包</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>99</th>\n",
       "      <td>【特价】复古款拼接真皮手提包</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>100 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                               product_name\n",
       "0      【新品】Apple/苹果 iPhone 11 Pro Max 手机全网通\n",
       "1     【限时特价】华为 Mate 30 Pro 5G 8GB+256GB 全网通\n",
       "2   【新款上市】Vivo iQOO Pro 5G 手机 8GB+128GB 全网通\n",
       "3     【热卖】OPPO Reno 10X Zoom 12GB+256GB 全网通\n",
       "4               【新品】小米CC9 Pro 全网通 6GB+128GB\n",
       "..                                      ...\n",
       "95                           【特价】百搭时尚复古款双肩包\n",
       "96                         【精选】清新拼接复古款女士手拿包\n",
       "97                         【专享】潮流拼接复古款真皮手提包\n",
       "98                         【热卖】简约款时尚拼接女士手拿包\n",
       "99                           【特价】复古款拼接真皮手提包\n",
       "\n",
       "[100 rows x 1 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_csv(\"data/product_names.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "from openai.embeddings_utils import get_embeddings\n",
    "import openai, os, backoff\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "embedding_model = \"text-embedding-ada-002\"\n",
    "\n",
    "batch_size = 100\n",
    "\n",
    "@backoff.on_exception(backoff.expo, openai.error.RateLimitError) # type: ignore\n",
    "def get_embeddings_with_backoff(prompts, engine):\n",
    "    embeddings = []\n",
    "    for i in range(0, len(prompts), batch_size):\n",
    "        batch = prompts[i:i+batch_size]\n",
    "        embeddings += get_embeddings(list_of_text=batch, engine=engine)\n",
    "    return embeddings\n",
    "\n",
    "prompts = df.product_name.tolist()\n",
    "prompt_batches = [prompts[i:i+batch_size] for i in range(0, len(prompts), batch_size)]\n",
    "\n",
    "embeddings = []\n",
    "for batch in prompt_batches:\n",
    "    batch_embeddings = get_embeddings_with_backoff(prompts=batch, engine=embedding_model)\n",
    "    embeddings += batch_embeddings\n",
    "\n",
    "df[\"embedding\"] = embeddings\n",
    "df.to_parquet(\"data/taobao_product_title.parquet\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "【新品】小米CC9 Pro 全网通 6GB+128GB\n",
      "【限时特惠】小米平板4 10\n",
      "【限时特价】华为 Mate 30 Pro 5G 8GB+256GB 全网通\n"
     ]
    }
   ],
   "source": [
    "from openai.embeddings_utils import get_embedding, cosine_similarity\n",
    "\n",
    "# search through the reviews for a specific product\n",
    "def search_product(df, query, n=3, pprint=True):\n",
    "    product_embedding = get_embedding(\n",
    "        query,\n",
    "        engine=embedding_model\n",
    "    )\n",
    "    df[\"similarity\"] = df.embedding.apply(lambda x: cosine_similarity(x, product_embedding))\n",
    "\n",
    "    results = (\n",
    "        df.sort_values(\"similarity\", ascending=False)\n",
    "        .head(n)\n",
    "        .product_name\n",
    "    )\n",
    "    if pprint:\n",
    "        for r in results:\n",
    "            print(r)\n",
    "    return results\n",
    "\n",
    "\n",
    "results = search_product(df, \"小米手机\", n=3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 使用向量数据库加速搜索过程"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "import faiss\n",
    "import numpy as np\n",
    "\n",
    "def load_embeddings_to_faiss(df):\n",
    "    embeddings = np.array(df['embedding'].tolist()).astype('float32')\n",
    "    index = faiss.IndexFlatL2(embeddings.shape[1])\n",
    "    index.add(embeddings) # type: ignore\n",
    "    return index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<bound method NDFrame.head of                          id  code          category   \n",
       "0       6551700932705387022   101      news_culture  \\\n",
       "1       6552368441838272771   101      news_culture   \n",
       "2       6552407965343678723   101      news_culture   \n",
       "3       6552332417753940238   101      news_culture   \n",
       "4       6552475601595269390   101      news_culture   \n",
       "...                     ...   ...               ...   \n",
       "382683  6554578634403741966   109         news_tech   \n",
       "382684  6554623450374209806   110     news_military   \n",
       "382685  6554489948580348424   113        news_world   \n",
       "382686  6554706019040100611   113        news_world   \n",
       "382687  6554360505438306824   115  news_agriculture   \n",
       "\n",
       "                                  title                  keywords   \n",
       "0                      京城最值得你来场文化之旅的博物馆  保利集团,马未都,中国科学技术馆,博物馆,新中国  \\\n",
       "1                     发酵床的垫料种类有哪些？哪种更好？                             \n",
       "2                 上联：黄山黄河黄皮肤黄土高原。怎么对下联？                             \n",
       "3             林徽因什么理由拒绝了徐志摩而选择梁思成为终身伴侣？                             \n",
       "4                              黄杨木是什么树？                             \n",
       "...                                 ...                       ...   \n",
       "382683          A10处理器iPhone SE二代值得期待吗？                             \n",
       "382684  先进战机叛逃将带来重大损失，美军如何防止F22飞行员驾机叛逃？                             \n",
       "382685    又一国领导人放话，只要普京下令，数万大军“碾压”美国白宫！         以色列,普京,俄罗斯,叙利亚,车臣   \n",
       "382686        如何看待美国总统连续撕毁美国签署的国际协议的举动？                             \n",
       "382687           农博会上，公安100余种土特产成了抢手货……    特色农产品,农博会,荆州市,生物科技,公安县   \n",
       "\n",
       "                                                 combined  n_tokens   \n",
       "0       标题: 京城最值得你来场文化之旅的博物馆; 关键字: 保利集团,马未都,中国科学技术馆,博物...        56  \\\n",
       "1                            标题: 发酵床的垫料种类有哪些？哪种更好？; 关键字:         32   \n",
       "2                        标题: 上联：黄山黄河黄皮肤黄土高原。怎么对下联？; 关键字:         37   \n",
       "3                    标题: 林徽因什么理由拒绝了徐志摩而选择梁思成为终身伴侣？; 关键字:         44   \n",
       "4                                     标题: 黄杨木是什么树？; 关键字:         20   \n",
       "...                                                   ...       ...   \n",
       "382683                 标题: A10处理器iPhone SE二代值得期待吗？; 关键字:         23   \n",
       "382684         标题: 先进战机叛逃将带来重大损失，美军如何防止F22飞行员驾机叛逃？; 关键字:         50   \n",
       "382685  标题: 又一国领导人放话，只要普京下令，数万大军“碾压”美国白宫！; 关键字: 以色列,普京...        70   \n",
       "382686               标题: 如何看待美国总统连续撕毁美国签署的国际协议的举动？; 关键字:         40   \n",
       "382687  标题: 农博会上，公安100余种土特产成了抢手货……; 关键字: 特色农产品,农博会,荆州市...        55   \n",
       "\n",
       "                                                embedding  \n",
       "0       [-0.02160128578543663, -0.004615371115505695, ...  \n",
       "1       [0.014646276831626892, -0.0075576393865048885,...  \n",
       "2       [0.012643996626138687, 0.0035733033437281847, ...  \n",
       "3       [-0.0004984583356417716, -0.016934111714363098...  \n",
       "4       [-0.0008294825092889369, -0.013883085921406746...  \n",
       "...                                                   ...  \n",
       "382683  [-0.00946888793259859, 0.0019106261897832155, ...  \n",
       "382684  [-0.020892947912216187, -0.022526662796735764,...  \n",
       "382685  [-0.03224752098321915, -0.018219713121652603, ...  \n",
       "382686  [-0.03162332996726036, -0.01479333359748125, 0...  \n",
       "382687  [-0.005016508512198925, -0.0130994264036417, 0...  \n",
       "\n",
       "[382688 rows x 8 columns]>"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "toutiao_df = pd.read_parquet(\"data/toutiao_cat_data_all_with_embeddings.parquet\")\n",
    "toutiao_df.head"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "index = load_embeddings_to_faiss(toutiao_df)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "def search_index(index, df, query_embedding, k=5):\n",
    "    query_vector = np.array(query_embedding).reshape(1, -1).astype('float32')\n",
    "    distances, indexes = index.search(query_vector, k)\n",
    "\n",
    "    results = []\n",
    "    for i in range(len(indexes)):\n",
    "        news = df.iloc[indexes[i]]['title'].values.tolist()\n",
    "        results.append((distances[i], news))    \n",
    "    return results\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.16498899459838867\n",
      "[(array([0.22052437, 0.226196  , 0.22855574], dtype=float32), ['超26亿元 北京七家车企将获新能源国补', '北上资金维持积极布局节奏 本周斥逾65亿元加码18只个股', '北上资金维持积极布局节奏 本周斥65亿加码18只个股'])]\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "\n",
    "query = \"北上资金净买入超百亿元 重点加仓新能源汽车产业\"\n",
    "query_embedding = get_embedding(query, engine=embedding_model)\n",
    "\n",
    "start = time.time()\n",
    "news = search_index(index, toutiao_df, query_embedding, k=3)\n",
    "end = time.time()\n",
    "print(end - start)\n",
    "print(news)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 完全暴力检索的速度"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "from openai.embeddings_utils import get_embedding, cosine_similarity\n",
    "\n",
    "# search through the reviews for a specific product\n",
    "def search_index_brutal_force(df, query_embedding, n=3):\n",
    "    df[\"similarity\"] = df.embedding.apply(lambda x: cosine_similarity(x, query_embedding))\n",
    "\n",
    "    results = (\n",
    "        df.sort_values(\"similarity\", ascending=False)\n",
    "        .head(n)\n",
    "        .title\n",
    "    )\n",
    "    return results\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "58.231921911239624\n",
      "195956             超26亿元 北京七家车企将获新能源国补\n",
      "322070    北上资金维持积极布局节奏 本周斥逾65亿元加码18只个股\n",
      "321835      北上资金维持积极布局节奏 本周斥65亿加码18只个股\n",
      "Name: title, dtype: object\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "\n",
    "start = time.time()\n",
    "news = search_index_brutal_force(toutiao_df, query_embedding, n=3)\n",
    "end = time.time()\n",
    "print(end - start)\n",
    "print(news)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 通过知识库解决幻觉问题"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 藤野先生的例子"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "鲁迅先生在日本学习医学的老师是高村光太郎。\n"
     ]
    }
   ],
   "source": [
    "import openai\n",
    "import os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "COMPLETION_MODEL = \"text-davinci-003\"\n",
    "\n",
    "def get_response(prompt):\n",
    "    completions = openai.Completion.create(\n",
    "        engine=COMPLETION_MODEL,\n",
    "        prompt=prompt,\n",
    "        max_tokens=512,\n",
    "        n=1,\n",
    "        stop=None,\n",
    "        temperature=0.0,        \n",
    "    )\n",
    "    return completions\n",
    "\n",
    "\n",
    "answer = get_response(\"鲁迅先生在日本学习医学的老师是谁？\")\n",
    "print(answer.choices[0].text) #type: ignore\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "鲁迅先生在日本学习医学的老师是藤野先生。\n"
     ]
    }
   ],
   "source": [
    "import openai, os\n",
    "from llama_index import VectorStoreIndex, SimpleDirectoryReader\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "\n",
    "documents = SimpleDirectoryReader('./data/mr_fujino').load_data()\n",
    "index = VectorStoreIndex.from_documents(documents)\n",
    "\n",
    "query_engine = index.as_query_engine()\n",
    "response = query_engine.query(\"鲁迅先生在日本学习医学的老师是谁？\")\n",
    "print(response)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 自问自答进行数据增强"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Q1：什么是卷积神经网络（CNN）？\n",
      "A1：卷积神经网络（CNN）是一种深度学习算法，特别适合于处理具有网格结构（如图像）的数据。CNN 的设计灵感来自于生物的视觉皮层机制，通过对输入数据的局部感知和参数共享的特性，使得网络对图像的平移等变化有较好的鲁棒性。\n",
      "\n",
      "Q2：CNN 的基本构成部分有哪些？\n",
      "A2：CNN 的基本构成部分包括：卷积层、激活层（ReLU）、池化层（Pooling）以及全连接层（Fully Connected Layer）。这些层次交织形成了一个典型的 CNN 结构。\n",
      "\n",
      "Q3：卷积层的作用是什么？\n",
      "A3：卷积层是 CNN 的核心，主要进行特征提取。卷积层通过滤波器（或称为卷积核）在输入数据上进行滑动卷积操作，以获取局部特征。同时，同一卷积层的所有滤波器共享参数，减少了模型复杂度。\n",
      "\n",
      "Q4：激活层的作用是什么？\n",
      "A4：激活层通常在卷积层之后接一个激活层，如 ReLU、tanh 或 sigmoid 等。激活函数的目的是引入非线性因素，使得网络可以拟合更复杂的函数。\n",
      "\n",
      "Q5：池化层的作用是什么？\n",
      "A5：池化层通常在卷积层之后，用来降低特征的维度，防止过拟合。常见的池化方式有最大池化（Max Pooling）、平均池化（Average Pooling）等。\n",
      "\n",
      "Q6：全连接层的作用是什么？\n",
      "A6：全连接层在整个网络的最后一般会有全连接层，用于将前面提取的特征进行融合，并输出结果。在分类问题中，输出层通常会接一个 softmax 层进行概率归一化，使得输出可以解释为概率。\n",
      "\n",
      "Q7：现代的 CNN 架构有哪些？\n",
      "A7：现代的 CNN 架构有残差连接（ResNet）、空洞卷积（Dilated Convolution）、深度可分离卷积（Depthwise Separable Convolution）等等。\n",
      "\n",
      "Q8：CNN 在哪些问题上有显著的表现？\n",
      "A8：CNN 在图像分类、检测、分割等问题上有显著的表现。\n",
      "\n",
      "Q9：有哪些经典的 CNN 架构？\n",
      "A9：经典的 CNN 架构有 LeNet-5、AlexNet、VGG、GoogLeNet、ResNet 等。\n"
     ]
    }
   ],
   "source": [
    "content = \"\"\"卷积神经网络（Convolutional Neural Networks，简称 CNN）是一种深度学习算法，特别适合于处理具有网格结构（如图像）的数据。CNN 的设计灵感来自于生物的视觉皮层机制，通过对输入数据的局部感知和参数共享的特性，使得网络对图像的平移等变化有较好的鲁棒性。\n",
    "\n",
    "CNN 的基本构成部分包括：卷积层、激活层（ReLU）、池化层（Pooling）以及全连接层（Fully Connected Layer）。这些层次交织形成了一个典型的 CNN 结构。\n",
    "\n",
    "卷积层：这是 CNN 的核心，主要进行特征提取。卷积层通过滤波器（或称为卷积核）在输入数据上进行滑动卷积操作，以获取局部特征。同时，同一卷积层的所有滤波器共享参数，减少了模型复杂度。\n",
    "\n",
    "激活层：通常在卷积层之后接一个激活层，如 ReLU、tanh 或 sigmoid 等。激活函数的目的是引入非线性因素，使得网络可以拟合更复杂的函数。\n",
    "\n",
    "池化层：池化层通常在卷积层之后，用来降低特征的维度，防止过拟合。常见的池化方式有最大池化（Max Pooling）、平均池化（Average Pooling）等。\n",
    "\n",
    "全连接层：在整个网络的最后一般会有全连接层，用于将前面提取的特征进行融合，并输出结果。在分类问题中，输出层通常会接一个 softmax 层进行概率归一化，使得输出可以解释为概率。\n",
    "\n",
    "以上是 CNN 的基本结构，但实际上现代的 CNN 架构已经远超这些基础层次，如残差连接（ResNet）、空洞卷积（Dilated Convolution）、深度可分离卷积（Depthwise Separable Convolution）等等。\n",
    "\n",
    "CNN 在图像分类、检测、分割等问题上有显著的表现。经典的 CNN 架构有 LeNet-5、AlexNet、VGG、GoogLeNet、ResNet 等。\n",
    "\"\"\"\n",
    "\n",
    "prompt = f\"\"\"请针对>>>和<<<之间的内容，设计一系列的FAQ问题，以及对应的答案。\n",
    "<<<{content}>>>\"\"\"\n",
    "\n",
    "\n",
    "import openai\n",
    "import os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "COMPLETION_MODEL = \"text-davinci-003\"\n",
    "\n",
    "\n",
    "def get_response(prompt):\n",
    "    completions = openai.Completion.create (\n",
    "        engine=COMPLETION_MODEL,\n",
    "        prompt=prompt,\n",
    "        max_tokens=2048,\n",
    "        n=1,\n",
    "        stop=None,\n",
    "        temperature=0.0,        \n",
    "    )\n",
    "    message = completions['choices'][0].text # type: ignore\n",
    "    return message\n",
    "\n",
    "answer = get_response(prompt)\n",
    "print(answer)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['\\nQ1：什么是卷积神经网络（CNN）？\\nA1：卷积神经网络（CNN）是一种深度学习算法，特别适合于处理具有网格结构（如图像）的数据。CNN 的设计灵感来自于生物的视觉皮层机制，通过对输入数据的局部感知和参数共享的特性，使得网络对图像的平移等变化有较好的鲁棒性。', 'Q2：CNN 的基本构成部分有哪些？\\nA2：CNN 的基本构成部分包括：卷积层、激活层（ReLU）、池化层（Pooling）以及全连接层（Fully Connected Layer）。这些层次交织形成了一个典型的 CNN 结构。', 'Q3：卷积层的作用是什么？\\nA3：卷积层是 CNN 的核心，主要进行特征提取。卷积层通过滤波器（或称为卷积核）在输入数据上进行滑动卷积操作，以获取局部特征。同时，同一卷积层的所有滤波器共享参数，减少了模型复杂度。', 'Q4：激活层的作用是什么？\\nA4：激活层通常在卷积层之后接一个激活层，如 ReLU、tanh 或 sigmoid 等。激活函数的目的是引入非线性因素，使得网络可以拟合更复杂的函数。', 'Q5：池化层的作用是什么？\\nA5：池化层通常在卷积层之后，用来降低特征的维度，防止过拟合。常见的池化方式有最大池化（Max Pooling）、平均池化（Average Pooling）等。', 'Q6：全连接层的作用是什么？\\nA6：全连接层在整个网络的最后一般会有全连接层，用于将前面提取的特征进行融合，并输出结果。在分类问题中，输出层通常会接一个 softmax 层进行概率归一化，使得输出可以解释为概率。', 'Q7：现代的 CNN 架构有哪些？\\nA7：现代的 CNN 架构有残差连接（ResNet）、空洞卷积（Dilated Convolution）、深度可分离卷积（Depthwise Separable Convolution）等等。', 'Q8：CNN 在哪些问题上有显著的表现？\\nA8：CNN 在图像分类、检测、分割等问题上有显著的表现。', 'Q9：有哪些经典的 CNN 架构？\\nA9：经典的 CNN 架构有 LeNet-5、AlexNet、VGG、GoogLeNet、ResNet 等。']\n",
      "\n",
      "现代的 CNN 架构有残差连接（ResNet）、空洞卷积（Dilated Convolution）、深度可分离卷积（Depthwise Separable Convolution）等等。\n"
     ]
    }
   ],
   "source": [
    "faq = answer.split(\"\\n\\n\")\n",
    "print(faq)\n",
    "\n",
    "\n",
    "import openai, os\n",
    "from llama_index import Document, VectorStoreIndex, SimpleDirectoryReader\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "\n",
    "documents = []\n",
    "for qa in faq:\n",
    "    documents.append(Document(qa))\n",
    "\n",
    "index = VectorStoreIndex.from_documents(documents)\n",
    "\n",
    "query_engine = index.as_query_engine()\n",
    "response = query_engine.query(\"现代CNN架构有哪些？\")\n",
    "print(response)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 作业 or 练习题\n",
    "\n",
    "是否可以通过llama-index来实现Self-Ask的功能？"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 通过多步骤的LLM和外部工具来解决问题"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### LLM做数学题"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "OpenAI API 说答案是: \n",
      "\n",
      "352 x 493 = 174,336\n",
      "Python 说答案是: 173536\n"
     ]
    }
   ],
   "source": [
    "import openai, os\n",
    "\n",
    "openai.api_key = os.environ.get(\"OPENAI_API_KEY\")\n",
    "\n",
    "from langchain.prompts import PromptTemplate\n",
    "from langchain.llms import OpenAI\n",
    "from langchain.chains import LLMChain\n",
    "\n",
    "\n",
    "llm = OpenAI(model_name=\"text-davinci-003\", max_tokens=2048, temperature=0.5) # type: ignore\n",
    "multiply_prompt = PromptTemplate(template=\"请计算一下{question}是多少?\", input_variables=[\"question\"])\n",
    "math_chain = LLMChain(llm=llm, prompt=multiply_prompt, output_key=\"answer\")\n",
    "answer = math_chain.run({\"question\": \"352乘以493\"})\n",
    "print(\"OpenAI API 说答案是:\", answer)\n",
    "\n",
    "python_answer = 352 * 493\n",
    "print(\"Python 说答案是:\", python_answer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "print(352 * 493)\n"
     ]
    }
   ],
   "source": [
    "multiply_by_python_prompt = PromptTemplate(template=\"请写一段Python代码，计算{question}?\", input_variables=[\"question\"])\n",
    "math_chain = LLMChain(llm=llm, prompt=multiply_by_python_prompt, output_key=\"answer\")\n",
    "answer = math_chain.run({\"question\": \"352乘以493\"})\n",
    "print(answer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "173536\n",
      "\n"
     ]
    }
   ],
   "source": [
    "multiply_by_python_prompt = PromptTemplate(template=\"请写一段Python代码，计算{question}?\", input_variables=[\"question\"])\n",
    "math_chain = LLMChain(llm=llm, prompt=multiply_by_python_prompt, output_key=\"answer\")\n",
    "answer_code = math_chain.run({\"question\": \"352乘以493\"})\n",
    "\n",
    "from langchain.utilities import PythonREPL\n",
    "python_repl = PythonREPL()\n",
    "result = python_repl.run(answer_code)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/xuwenhao/miniconda3/envs/mactalk/lib/python3.10/site-packages/langchain/chains/llm_math/base.py:50: UserWarning: Directly instantiating an LLMMathChain with an llm is deprecated. Please instantiate with llm_chain argument or using the from_llm class method.\n",
      "  warnings.warn(\n",
      "Error in on_chain_start callback: 'name'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "请计算一下352乘以493是多少?\u001b[32;1m\u001b[1;3m```text\n",
      "352 * 493\n",
      "```\n",
      "...numexpr.evaluate(\"352 * 493\")...\n",
      "\u001b[0m\n",
      "Answer: \u001b[33;1m\u001b[1;3m173536\u001b[0m\n",
      "\u001b[1m> Finished chain.\u001b[0m\n",
      "Answer: 173536\n"
     ]
    }
   ],
   "source": [
    "from langchain import LLMMathChain\n",
    "\n",
    "llm_math = LLMMathChain(llm=llm, verbose=True) #type: ignore\n",
    "result = llm_math.run(\"请计算一下352乘以493是多少?\")\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### LLM调用搜索引擎"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain import LLMMathChain, OpenAI, SerpAPIWrapper\n",
    "from langchain.agents import initialize_agent, Tool\n",
    "from langchain.agents import AgentType\n",
    "\n",
    "llm = OpenAI(temperature=0) # type: ignore\n",
    "search = SerpAPIWrapper() # type: ignore\n",
    "llm_math_chain = LLMMathChain.from_llm(llm=llm, verbose=True)\n",
    "tools = [\n",
    "    Tool(\n",
    "        name=\"Search\",\n",
    "        func=search.run,\n",
    "        description=\"useful for when you need to answer questions about current events. You should ask targeted questions\"\n",
    "    ),\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "agent = initialize_agent(tools, llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Error in on_chain_start callback: 'name'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[32;1m\u001b[1;3m I need to find out who will win the NBA championship in 2023\n",
      "Action: Search\n",
      "Action Input: 2023年NBA总冠军\u001b[0m\n",
      "Observation: \u001b[36;1m\u001b[1;3m2023年NBA总决赛（英语：2023 NBA Finals）是2022–23 NBA赛季的冠军系列赛，将由2023年6月1日至6月12日进行，由西部第一种子丹佛掘金对战东部第八种子迈阿密热火，比赛采用七场四胜制2-2-1-1-1的形式，常规赛战绩较佳的丹佛掘金有多一场主场优势。\u001b[0m\n",
      "Thought:\u001b[32;1m\u001b[1;3m It looks like the Denver Nuggets will have home court advantage in the 2023 NBA Finals\n",
      "Final Answer: 2023年NBA总冠军可能是丹佛掘金队。\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished chain.\u001b[0m\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "'2023年NBA总冠军可能是丹佛掘金队。'"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "agent.run(\"2023年的NBA总冠军是谁？请用中文回答\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 做选择题的方式来解决问题"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Langchain的工具接口"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### FAQ"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.embeddings.openai import OpenAIEmbeddings\n",
    "from langchain.vectorstores import FAISS\n",
    "from langchain.text_splitter import SpacyTextSplitter\n",
    "from langchain import OpenAI\n",
    "from langchain.chains import RetrievalQA\n",
    "from langchain.document_loaders import TextLoader\n",
    "from langchain.agents import tool\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "loader = TextLoader('./data/ecommerce_faq.txt')\n",
    "documents = loader.load()\n",
    "text_splitter = SpacyTextSplitter(chunk_size=256, pipeline=\"zh_core_web_sm\")\n",
    "texts = text_splitter.split_documents(documents)\n",
    "\n",
    "embeddings = OpenAIEmbeddings() # type: ignore\n",
    "docsearch = FAISS.from_documents(texts, embeddings)\n",
    "faq_chain = RetrievalQA.from_chain_type(llm=OpenAI(temperature=0), retriever=docsearch.as_retriever(), verbose=True) # type: ignore\n",
    "\n",
    "@tool(\"FAQ\")\n",
    "def faq(intput: str) -> str:\n",
    "    \"\"\"\"useful for when you need to answer questions about shopping policies, like return policy, shipping policy, etc.\"\"\"\n",
    "    return faq_chain.run(intput)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Error in on_chain_start callback: 'name'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\u001b[1m> Finished chain.\u001b[0m\n",
      " 我们支持全国大部分省份的配送，包括三亚。一般情况下，大部分城市的订单在2-3个工作日内送达，偏远地区可能需要5-7个工作日。\n"
     ]
    }
   ],
   "source": [
    "question = \"请问你们的货，能送到三亚吗？大概需要几天？\"\n",
    "result = faq_chain.run(question)\n",
    "print(result)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 商品信息"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "from langchain.text_splitter import CharacterTextSplitter\n",
    "from langchain.document_loaders import CSVLoader\n",
    "\n",
    "product_loader = CSVLoader('./data/ecommerce_products.csv')\n",
    "product_documents = product_loader.load()\n",
    "product_text_splitter = CharacterTextSplitter(chunk_size=1024, separator=\"\\n\")\n",
    "product_texts = product_text_splitter.split_documents(product_documents)\n",
    "product_search = FAISS.from_documents(product_texts, embeddings)\n",
    "product_chain = RetrievalQA.from_chain_type(llm=OpenAI(temperature=0), retriever=product_search.as_retriever(), verbose=True)# type: ignore\n",
    "\n",
    "@tool(\"Recommend Product\")\n",
    "def recommend_product(input: str) -> str:\n",
    "    \"\"\"\"useful for when you need to search and recommend products and recommend it to the user\"\"\"\n",
    "    return product_chain.run(input)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Error in on_chain_start callback: 'name'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\u001b[1m> Finished chain.\u001b[0m\n",
      " 根据您的要求，我推荐您可以买一件毛衣裙，它采用保暖的面料，颜色为米白色，适合春秋季节穿着，搭配一双黑色长靴和小包包非常优雅大方，非常适合您去公园穿着。\n"
     ]
    }
   ],
   "source": [
    "question = \"我想买一件衣服，想要在春天去公园穿，但是不知道哪个款式好看，你能帮我推荐一下吗？\"\n",
    "answer = product_chain.run(question)\n",
    "print(answer)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 订单查询"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "import re\n",
    "import json\n",
    "\n",
    "ORDER_1 = \"20230101ABC\"\n",
    "ORDER_2 = \"20230101EFG\"\n",
    "\n",
    "ORDER_1_DETAIL = {\n",
    "    \"order_number\": ORDER_1,\n",
    "    \"status\": \"已发货\",\n",
    "    \"shipping_date\" : \"2023-01-03\",\n",
    "    \"estimated_delivered_date\": \"2023-01-05\",\n",
    "} \n",
    "\n",
    "ORDER_2_DETAIL = {\n",
    "    \"order_number\": ORDER_2,\n",
    "    \"status\": \"未发货\",\n",
    "    \"shipping_date\" : None,\n",
    "    \"estimated_delivered_date\": None,\n",
    "}\n",
    "\n",
    "answer_order_info = PromptTemplate(\n",
    "    template=\"请把下面的订单信息回复给用户： \\n\\n {order}?\", input_variables=[\"order\"]\n",
    ")\n",
    "answer_order_llm = LLMChain(llm=OpenAI(temperature=0),  prompt=answer_order_info) # type: ignore\n",
    "\n",
    "\n",
    "@tool(\"Search Order\", return_direct=True)\n",
    "def search_order(input:str)->str:\n",
    "    \"\"\"useful for when you need to answer questions about customers orders\"\"\"\n",
    "    pattern = r\"\\d+[A-Z]+\"\n",
    "    match = re.search(pattern, input)\n",
    "\n",
    "    order_number = input\n",
    "    if match:\n",
    "        order_number = match.group(0)\n",
    "    else:\n",
    "        return \"请问您的订单号是多少？\"\n",
    "    if order_number == ORDER_1:        \n",
    "        return answer_order_llm.run(json.dumps(ORDER_1_DETAIL))\n",
    "    elif order_number == ORDER_2:\n",
    "        return answer_order_llm.run(json.dumps(ORDER_2_DETAIL))\n",
    "    else:\n",
    "        return f\"对不起，根据{input}没有找到您的订单\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "tools = {\"search_order\" : search_order, \"recommend_product\": recommend_product, \"faq\": faq}\n",
    "\n",
    "\n",
    "choicellm = OpenAI(max_tokens=2048, temperature=0.5) # type: ignore\n",
    "multiple_choice = \"\"\"\n",
    "请针对用户问题，选择一个合适的工具去回答她的问题。只要用A、B、C的选项字母告诉我答案。\n",
    "\n",
    "我们有的工具包括：\n",
    "A. recommend_product[$keywords], 一个能够查询商品信息，为用户进行商品导购的工具。\n",
    "B. search_order[$order_number], 一个能够查询订单信息，获得最新的订单情况的工具\n",
    "C. faq[$question], 一个能够搜索商家的退换货政策、运费、物流时长、支付渠道、覆盖国家的工具\n",
    "\n",
    "\n",
    "返回的结果需要包含函数名称，以及对应的参数，比如：\n",
    "\n",
    "用户问题：我的订单 20220101ABC 的物流信息是什么？\n",
    "返回结果：search_order[20220101ABC]\n",
    "\n",
    "用户问题：{question}\n",
    "返回结果：\n",
    "\"\"\"\n",
    "multiple_choice_prompt = PromptTemplate(template=multiple_choice, input_variables=[\"question\"])\n",
    "choice_chain = LLMChain(llm=choicellm, prompt=multiple_choice_prompt, output_key=\"answer\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "search_order[2022ABCDE]\n"
     ]
    }
   ],
   "source": [
    "question = \"我有一张订单，订单号是 2022ABCDE，一直没有收到，能麻烦帮我查一下吗？\"\n",
    "result = choice_chain.run(question)\n",
    "print(result) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('search_order', '2022ABCDE')\n"
     ]
    }
   ],
   "source": [
    "from typing import Tuple\n",
    "\n",
    "def parse_result(result: str) -> Tuple[str, str]:\n",
    "    pattern = r\"(\\w+)\\[(.+)\\]\"\n",
    "    match = re.search(pattern, result)\n",
    "    if match:\n",
    "        return match.group(1), match.group(2)\n",
    "    else:\n",
    "        return \"\", \"\"\n",
    "    \n",
    "\n",
    "print(parse_result(result))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "对不起，根据2022ABCDE没有找到您的订单\n"
     ]
    }
   ],
   "source": [
    "parsed_result = parse_result(result)\n",
    "method = tools.get(parsed_result[0])\n",
    "params = parsed_result[1]\n",
    "\n",
    "\n",
    "answer = method(params) # type: ignore\n",
    "print(answer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "\n",
      "您的订单号为20230101ABC，状态为已发货，发货日期为2023-01-03，预计到货日期为2023-01-05。\n"
     ]
    }
   ],
   "source": [
    "question = \"我有一张订单，订单号是 20230101ABC，一直没有收到，能麻烦帮我查一下吗？\"\n",
    "\n",
    "result = choice_chain.run(question)\n",
    "parsed_result = parse_result(result)\n",
    "method = tools.get(parsed_result[0])\n",
    "params = parsed_result[1]\n",
    "\n",
    "\n",
    "answer = method(params) # type: ignore\n",
    "print(answer)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 作业，通过Langchain实现上面的功能"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Langchain的Agent功能"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [],
   "source": [
    "agent_tools = [search_order,recommend_product, faq]\n",
    "agent = initialize_agent(agent_tools, OpenAI(temperature=0), agent=AgentType.CHAT_ZERO_SHOT_REACT_DESCRIPTION, verbose=True) #type: ignore"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Error in on_chain_start callback: 'name'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[32;1m\u001b[1;3m\n",
      "Thought: I need to find out the status of the order\n",
      "Action:\n",
      "```\n",
      "{\n",
      "  \"action\": \"Search Order\",\n",
      "  \"action_input\": \"2022ABCDE\"\n",
      "}\n",
      "```\n",
      "\u001b[0m\n",
      "Observation: \u001b[36;1m\u001b[1;3m对不起，根据2022ABCDE没有找到您的订单\u001b[0m\n",
      "\u001b[32;1m\u001b[1;3m\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished chain.\u001b[0m\n",
      "对不起，根据2022ABCDE没有找到您的订单\n"
     ]
    }
   ],
   "source": [
    "question = \"我有一张订单，订单号是 2022ABCDE，一直没有收到，能麻烦帮我查一下吗？\"\n",
    "answer = agent.run(question)\n",
    "print(answer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Error in on_chain_start callback: 'name'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[32;1m\u001b[1;3m\n",
      "Thought: The customer is asking about an order they have not received.\n",
      "\n",
      "Action:\n",
      "```\n",
      "{\n",
      "  \"action\": \"Search Order\",\n",
      "  \"action_input\": \"20230101ABC\"\n",
      "}\n",
      "```\n",
      "\n",
      "\u001b[0m\n",
      "Observation: \u001b[36;1m\u001b[1;3m\n",
      "\n",
      "您的订单号为20230101ABC，状态为已发货，发货日期为2023-01-03，预计到货日期为2023-01-05。\u001b[0m\n",
      "\u001b[32;1m\u001b[1;3m\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished chain.\u001b[0m\n",
      "\n",
      "\n",
      "您的订单号为20230101ABC，状态为已发货，发货日期为2023-01-03，预计到货日期为2023-01-05。\n"
     ]
    }
   ],
   "source": [
    "question2 = \"我有一张订单，订单号是 20230101ABC 一直没有收到，能麻烦帮我查一下吗？\"\n",
    "answer2 = agent.run(question2)\n",
    "print(answer2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Error in on_chain_start callback: 'name'\n",
      "Error in on_chain_start callback: 'name'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[32;1m\u001b[1;3m\n",
      "Thought: I need to answer a question about the return policy\n",
      "Action:\n",
      "```\n",
      "{\n",
      "  \"action\": \"FAQ\",\n",
      "  \"action_input\": \"return policy\"\n",
      "}\n",
      "```\n",
      "\u001b[0m\n",
      "\u001b[1m> Finished chain.\u001b[0m\n",
      "\n",
      "Observation: \u001b[38;5;200m\u001b[1;3m 自收到商品之日起7天内，如产品未使用、包装完好，您可以申请退货。某些特殊商品可能不支持退货，请在购买前查看商品详情页面的退货政策。\u001b[0m\n",
      "Thought:\u001b[32;1m\u001b[1;3m I now know the final answer\n",
      "Final Answer: 自收到商品之日起7天内，如产品未使用、包装完好，您可以申请退货。某些特殊商品可能不支持退货，请在购买前查看商品详情页面的退货政策。\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished chain.\u001b[0m\n",
      "自收到商品之日起7天内，如产品未使用、包装完好，您可以申请退货。某些特殊商品可能不支持退货，请在购买前查看商品详情页面的退货政策。\n"
     ]
    }
   ],
   "source": [
    "question3 = \"你们的退货政策是怎么样的？\"\n",
    "answer3 = agent.run(question3)\n",
    "print(answer3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Error in on_chain_start callback: 'name'\n",
      "Error in on_chain_start callback: 'name'\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[32;1m\u001b[1;3m\n",
      "Thought: I need to recommend a product\n",
      "Action:\n",
      "```\n",
      "{\n",
      "  \"action\": \"Recommend Product\",\n",
      "  \"action_input\": \"Clothing for Spring\"\n",
      "}\n",
      "```\n",
      "\u001b[0m\n",
      "\u001b[1m> Finished chain.\u001b[0m\n",
      "\n",
      "Observation: \u001b[33;1m\u001b[1;3m 毛衣裙、卫衣连衣裙、雪纺连衣裙、长款风衣\u001b[0m\n",
      "Thought:\u001b[32;1m\u001b[1;3m I now know the final answer\n",
      "Final Answer: 毛衣裙、卫衣连衣裙、雪纺连衣裙、长款风衣\u001b[0m\n",
      "\n",
      "\u001b[1m> Finished chain.\u001b[0m\n",
      "毛衣裙、卫衣连衣裙、雪纺连衣裙、长款风衣\n"
     ]
    }
   ],
   "source": [
    "question4 = \"能推荐一件现在适合春天穿的衣服给我么？\"\n",
    "answer4 = agent.run(question4)\n",
    "print(answer4)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "mactalk",
   "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": "3.10.11"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}