跳转至

安装⚓︎

:label:chap_installation

In order to get up and running, we will need an environment for running Python, the Jupyter Notebook, the relevant libraries, and the code needed to run the book itself.

Installing Miniconda⚓︎

Your simplest option is to install Miniconda. Note that the Python 3.x version is required. You can skip the following steps if your machine already has conda installed.

Visit the Miniconda website and determine the appropriate version for your system based on your Python 3.x version and machine architecture. Suppose that your Python version is 3.9 (our tested version). If you are using macOS, you would download the bash script whose name contains the strings "MacOSX", navigate to the download location, and execute the installation as follows (taking Intel Macs as an example):

# The file name is subject to changes
sh Miniconda3-py39_4.12.0-MacOSX-x86_64.sh -b

A Linux user would download the file whose name contains the strings "Linux" and execute the following at the download location:

# The file name is subject to changes
sh Miniconda3-py39_4.12.0-Linux-x86_64.sh -b

A Windows user would download and install Miniconda by following its online instructions. On Windows, you may search for cmd to open the Command Prompt (command-line interpreter) for running commands.

Next, initialize the shell so we can run conda directly.

~/miniconda3/bin/conda init

Then close and reopen your current shell. You should be able to create a new environment as follows:

conda create --name d2l python=3.9 -y

Now we can activate the d2l environment:

conda activate d2l

Installing the Deep Learning Framework and the d2l Package⚓︎

Before installing any deep learning framework, please first check whether or not you have proper GPUs on your machine (the GPUs that power the display on a standard laptop are not relevant for our purposes). For example, if your computer has NVIDIA GPUs and has installed CUDA, then you are all set. If your machine does not house any GPU, there is no need to worry just yet. Your CPU provides more than enough horsepower to get you through the first few chapters. Just remember that you will want to access GPUs before running larger models.

:begin_tab:mxnet

To install a GPU-enabled version of MXNet, we need to find out what version of CUDA you have installed. You can check this by running nvcc --version or cat /usr/local/cuda/version.txt. Assume that you have installed CUDA 11.2, then execute the following command:

# For macOS and Linux users
pip install mxnet-cu112==1.9.1

# For Windows users
pip install mxnet-cu112==1.9.1 -f https://dist.mxnet.io/python

You may change the last digits according to your CUDA version, e.g., cu101 for CUDA 10.1 and cu90 for CUDA 9.0.

If your machine has no NVIDIA GPUs or CUDA, you can install the CPU version as follows:

pip install mxnet==1.9.1

:end_tab:

:begin_tab:pytorch

You can install PyTorch (the specified versions are tested at the time of writing) with either CPU or GPU support as follows:

pip install torch==2.0.0 torchvision==0.15.1

:end_tab:

:begin_tab:tensorflow You can install TensorFlow with either CPU or GPU support as follows:

pip install tensorflow==2.12.0 tensorflow-probability==0.20.0

:end_tab:

:begin_tab:jax You can install JAX and Flax with either CPU or GPU support as follows:

# GPU
pip install "jax[cuda11_pip]==0.4.13" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html flax==0.7.0

If your machine has no NVIDIA GPUs or CUDA, you can install the CPU version as follows:

# CPU
pip install "jax[cpu]==0.4.13" flax==0.7.0

:end_tab:

Our next step is to install the d2l package that we developed in order to encapsulate frequently used functions and classes found throughout this book:

pip install d2l==1.0.3

Downloading and Running the Code⚓︎

Next, you will want to download the notebooks so that you can run each of the book's code blocks. Simply click on the "Notebooks" tab at the top of any HTML page on the D2L.ai website to download the code and then unzip it. Alternatively, you can fetch the notebooks from the command line as follows:

:begin_tab:mxnet

mkdir d2l-en && cd d2l-en
curl https://d2l.ai/d2l-en-1.0.3.zip -o d2l-en.zip
unzip d2l-en.zip && rm d2l-en.zip
cd mxnet

:end_tab:

:begin_tab:pytorch

mkdir d2l-en && cd d2l-en
curl https://d2l.ai/d2l-en-1.0.3.zip -o d2l-en.zip
unzip d2l-en.zip && rm d2l-en.zip
cd pytorch

:end_tab:

:begin_tab:tensorflow

mkdir d2l-en && cd d2l-en
curl https://d2l.ai/d2l-en-1.0.3.zip -o d2l-en.zip
unzip d2l-en.zip && rm d2l-en.zip
cd tensorflow

:end_tab:

:begin_tab:jax

mkdir d2l-en && cd d2l-en
curl https://d2l.ai/d2l-en-1.0.3.zip -o d2l-en.zip
unzip d2l-en.zip && rm d2l-en.zip
cd jax

:end_tab:

If you do not already have unzip installed, first run sudo apt-get install unzip. Now we can start the Jupyter Notebook server by running:

jupyter notebook

At this point, you can open http://localhost:8888 (it may have already opened automatically) in your web browser. Then we can run the code for each section of the book. Whenever you open a new command line window, you will need to execute conda activate d2l to activate the runtime environment before running the D2L notebooks, or updating your packages (either the deep learning framework or the d2l package). To exit the environment, run conda deactivate.

:begin_tab:mxnet Discussions :end_tab:

:begin_tab:pytorch Discussions :end_tab:

:begin_tab:tensorflow Discussions :end_tab:

:begin_tab:jax Discussions :end_tab:

最后更新: November 25, 2023
创建日期: November 25, 2023