system 第二章加入一些文字描述

content/Building Systems with the ChatGPT API/2.Language Models, the Chat Format and Tokens.ipynb

@@ -1,6 +1,7 @@
 {
  "cells": [
   {
+   "attachments": {},
    "cell_type": "markdown",
    "id": "3e71eee8",
    "metadata": {},
@@ -9,6 +10,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "id": "c7e800b0",
    "metadata": {},
@@ -92,7 +94,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": 1,
    "id": "b8b8f1c7",
    "metadata": {},
    "outputs": [],
@@ -104,10 +106,11 @@
     "# from dotenv import load_dotenv, find_dotenv\n",
     "# _ = load_dotenv(find_dotenv()) # 读取本地的.env环境文件\n",
     "\n",
-    "openai.api_key  = 'sk-xxxxxxxxxxxx' #更换成你自己的key"
+    "openai.api_key  = '***' #更换成你自己的key"
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "id": "19cf84a8",
    "metadata": {},
@@ -118,7 +121,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 15,
+   "execution_count": 2,
    "id": "3fda8d03",
    "metadata": {},
    "outputs": [],
@@ -136,6 +139,7 @@
    ]
   },
   {
+   "attachments": {},
    "cell_type": "markdown",
    "id": "536ad0c2",
    "metadata": {},
@@ -236,6 +240,19 @@
     "![image-2.png](attachment:image-2.png)"
    ]
   },
+  {
+   "attachments": {},
+   "cell_type": "markdown",
+   "id": "c06944b4",
+   "metadata": {},
+   "source": [
+    "对于英文输入，一个 token 一般对应4个字符或者四分之三个单词；对于中文输入，一个 token 一般对应一个或半个词。\n",
+    "\n",
+    "不同模型有不同的 token 限制，需要注意的是，这里的 token 限制是输入的 prompt 和输出的 completion 的 token 数之和，因此输入的 prompt 越长，能输出的 completion 的上限就越低。\n",
+    "\n",
+    "ChatGPT3.5 的 token 上限是 4096。"
+   ]
+  },
   {
    "attachments": {
     "image.png": {
@@ -252,13 +269,23 @@
     "![image.png](attachment:image.png)"
    ]
   },
+  {
+   "attachments": {},
+   "cell_type": "markdown",
+   "id": "ffafb0f1",
+   "metadata": {},
+   "source": [
+    "System 信息用于制定模型的规则，例如设定、回答准则一类的，而 assistant 信息就是具体让模型完成的指令"
+   ]
+  },
   {
    "cell_type": "code",
-   "execution_count": 9,
+   "execution_count": 3,
    "id": "bd009d67",
    "metadata": {},
    "outputs": [],
    "source": [
+    "# 支持更多参数自定义的封装函数\n",
     "def get_completion_from_messages(messages, \n",
     "                                 model=\"gpt-3.5-turbo\", \n",
     "                                 temperature=0, \n",
@@ -274,7 +301,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": 4,
    "id": "e09def17",
    "metadata": {},
    "outputs": [
@@ -282,20 +309,25 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "在海洋深处有一个小鲸鱼，\n",
-      "它快乐的游啊游，\n",
-      "它的身体白晃晃，\n",
-      "好像穿着白雪的衣裳。\n",
+      "在蓝色的海洋深处，\n",
+      "有一只小鲸鱼嬉闹着。\n",
+      "它游啊游，欢笑着，\n",
+      "尽情享受自由的感受。\n",
       "\n",
-      "它游来游去，玩得很开心，\n",
-      "偶尔还秀一下游泳技能，\n",
-      "小鲸鱼是个聪明的孩子，\n",
-      "它让波涛跟着它摇晃。\n",
+      "它跳啊跃，翻滚着，\n",
+      "好像永远都停不下来。\n",
+      "它的身体，是那么轻盈，\n",
+      "好像没有任何重量。\n",
       "\n",
-      "快乐的小鲸鱼，\n",
-      "带给海洋欢乐的音符，\n",
-      "我们也跟着它笑起来，\n",
-      "真是一个美好的夏日午后。\n"
+      "它的笑声，犹如海浪，\n",
+      "响彻于大海深处。\n",
+      "它的快乐，也如海浪，\n",
+      "绵延不绝，没有尽头。\n",
+      "\n",
+      "啊，这可爱的小鲸鱼，\n",
+      "让人忍不住想嬉闹。\n",
+      "让我们跟它一起游，\n",
+      "在蓝色的海洋里畅游。\n"
      ]
     }
    ],
@@ -364,7 +396,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 5,
    "id": "b0907cac",
    "metadata": {},
    "outputs": [],
@@ -385,8 +417,11 @@
     "    content = response.choices[0].message[\"content\"]\n",
     "    \n",
     "    token_dict = {\n",
+    "# prompt 对应的 token 数\n",
     "'prompt_tokens':response['usage']['prompt_tokens'],\n",
+    "# completion 对应的 token 数\n",
     "'completion_tokens':response['usage']['completion_tokens'],\n",
+    "# 总共的 token 数\n",
     "'total_tokens':response['usage']['total_tokens'],\n",
     "    }\n",
     "\n",
@@ -475,7 +510,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.9"
+   "version": "3.10.11"
   }
  },
  "nbformat": 4,