{ "cells": [ { "cell_type": "markdown", "id": "5d3957b1", "metadata": { "origin_pos": 0 }, "source": [ "# 使用Jupyter Notebook\n", ":label:`sec_jupyter`\n", "\n", "本节介绍如何使用Jupyter Notebook编辑和运行本书各章中的代码。确保你已按照 :ref:`chap_installation`中的说明安装了Jupyter并下载了代码。如果你想了解更多关于Jupyter的信息,请参阅其[文档](https://jupyter.readthedocs.io/en/latest/)中的优秀教程。 \n", "\n", "## 在本地编辑和运行代码\n", "\n", "假设本书代码的本地路径为`xx/yy/d2l-en/`。使用shell将目录更改为此路径(`cd xx/yy/d2l-en`)并运行命令`jupyter notebook`。如果浏览器未自动打开,请打开http://localhost:8888。此时你将看到Jupyter的界面以及包含本书代码的所有文件夹,如 :numref:`fig_jupyter00`所示\n", "\n", "![包含本书代码的文件夹](../img/jupyter00.png)\n", ":width:`600px`\n", ":label:`fig_jupyter00`\n", "\n", "你可以通过单击网页上显示的文件夹来访问notebook文件。它们通常有后缀“.ipynb”。为了简洁起见,我们创建了一个临时的“test.ipynb”文件。单击后显示的内容如 :numref:`fig_jupyter01`所示。此notebook包括一个标记单元格和一个代码单元格。标记单元格中的内容包括“This Is a Title”和“This is text.”。代码单元包含两行Python代码。 \n", "\n", "![“test.ipynb”文件中的markdown和代码块](../img/jupyter01.png)\n", ":width:`600px`\n", ":label:`fig_jupyter01`\n", "\n", "双击标记单元格以进入编辑模式。在单元格末尾添加一个新的文本字符串“Hello world.”,如 :numref:`fig_jupyter02`所示。 \n", "\n", "![编辑markdown单元格](../img/jupyter02.png)\n", ":width:`600px`\n", ":label:`fig_jupyter02`\n", "\n", "如 :numref:`fig_jupyter03`所示,单击菜单栏中的“Cell” $\\rightarrow$ “Run Cells”以运行编辑后的单元格。 \n", "\n", "![运行单元格](../img/jupyter03.png)\n", ":width:`600px`\n", ":label:`fig_jupyter03`\n", "\n", "运行后,markdown单元格如 :numref:`fig_jupyter04`所示。 \n", "\n", "![编辑后的markdown单元格](../img/jupyter04.png)\n", ":width:`600px`\n", ":label:`fig_jupyter04`\n", "\n", "接下来,单击代码单元。将最后一行代码后的元素乘以2,如 :numref:`fig_jupyter05`所示。 \n", "\n", "![编辑代码单元格](../img/jupyter05.png)\n", ":width:`600px`\n", ":label:`fig_jupyter05`\n", "\n", "你还可以使用快捷键(默认情况下为Ctrl+Enter)运行单元格,并从 :numref:`fig_jupyter06`获取输出结果。 \n", "\n", "![运行代码单元格以获得输出](../img/jupyter06.png)\n", ":width:`600px`\n", ":label:`fig_jupyter06`\n", "\n", "当一个notebook包含更多单元格时,我们可以单击菜单栏中的“Kernel”$\\rightarrow$“Restart & Run All”来运行整个notebook中的所有单元格。通过单击菜单栏中的“Help”$\\rightarrow$“Edit Keyboard Shortcuts”,可以根据你的首选项编辑快捷键。 \n", "\n", "## 高级选项\n", "\n", "除了本地编辑,还有两件事非常重要:以markdown格式编辑notebook和远程运行Jupyter。当我们想要在更快的服务器上运行代码时,后者很重要。前者很重要,因为Jupyter原生的ipynb格式存储了大量辅助数据,这些数据实际上并不特定于notebook中的内容,主要与代码的运行方式和运行位置有关。这让git感到困惑,并且使得合并贡献非常困难。幸运的是,还有另一种选择——在markdown中进行本地编辑。 \n", "\n", "### Jupyter中的Markdown文件\n", "\n", "如果你希望对本书的内容有所贡献,则需要在GitHub上修改源文件(md文件,而不是ipynb文件)。使用notedown插件,我们可以直接在Jupyter中修改md格式的notebook。 \n", "\n", "首先,安装notedown插件,运行Jupyter Notebook并加载插件:\n", "\n", "```\n", "pip install d2l-notedown # 你可能需要卸载原始notedown\n", "jupyter notebook --NotebookApp.contents_manager_class='notedown.NotedownContentsManager'\n", "```\n", "\n", "要在运行Jupyter Notebook时默认打开notedown插件,请执行以下操作:首先,生成一个Jupyter Notebook配置文件(如果已经生成了,可以跳过此步骤)。\n", "\n", "```\n", "jupyter notebook --generate-config\n", "```\n", "\n", "然后,在Jupyter Notebook配置文件的末尾添加以下行(对于Linux/macOS,通常位于`~/.jupyter/jupyter_notebook_config.py`):\n", "\n", "```\n", "c.NotebookApp.contents_manager_class = 'notedown.NotedownContentsManager'\n", "```\n", "\n", "在这之后,你只需要运行`jupyter notebook`命令就可以默认打开notedown插件。 \n", "\n", "### 在远程服务器上运行Jupyter Notebook\n", "\n", "有时,你可能希望在远程服务器上运行Jupyter Notebook,并通过本地计算机上的浏览器访问它。如果本地计算机上安装了Linux或MacOS(Windows也可以通过PuTTY等第三方软件支持此功能),则可以使用端口转发:\n", "\n", "```\n", "ssh myserver -L 8888:localhost:8888\n", "```\n", "\n", "以上是远程服务器`myserver`的地址。然后我们可以使用http://localhost:8888 访问运行Jupyter Notebook的远程服务器`myserver`。下一节将详细介绍如何在AWS实例上运行Jupyter Notebook。 \n", "\n", "### 执行时间\n", "\n", "我们可以使用`ExecuteTime`插件来计算Jupyter Notebook中每个代码单元的执行时间。使用以下命令安装插件:\n", "\n", "```\n", "pip install jupyter_contrib_nbextensions\n", "jupyter contrib nbextension install --user\n", "jupyter nbextension enable execute_time/ExecuteTime\n", "```\n", "\n", "## 小结\n", "\n", "* 使用Jupyter Notebook工具,我们可以编辑、运行和为本书做贡献。\n", "* 使用端口转发在远程服务器上运行Jupyter Notebook。\n", "\n", "## 练习\n", "\n", "1. 在本地计算机上使用Jupyter Notebook编辑并运行本书中的代码。\n", "1. 使用Jupyter Notebook通过端口转发来远程编辑和运行本书中的代码。\n", "1. 对于两个方矩阵,测量$\\mathbf{A}^\\top \\mathbf{B}$与$\\mathbf{A} \\mathbf{B}$在$\\mathbb{R}^{1024 \\times 1024}$中的运行时间。哪一个更快?\n", "\n", "[Discussions](https://discuss.d2l.ai/t/5731)\n" ] } ], "metadata": { "language_info": { "name": "python" }, "required_libs": [] }, "nbformat": 4, "nbformat_minor": 5 }