📚
UbuntuにDockerをインストールしてGPUを認識させる

2022/09/04 に公開
Docker
NVIDIA
CUDA
tech
Ubuntu20.04にDockerをインストールして、コンテナ内でNVIDIAのGPUを認識させるまでのメモ。ちなみにGPUはNVIDIA RTX3070の環境でやってます。
> sudo add-apt-repository ppa:graphics-drivers/ppa
> sudo apt update
> sudo apt search nvidia-driver-
nvidia-driver-515/focal-security 515.86.01-0ubuntu0.20.04.1 amd64 [upgradable from: 515.65.01-0ubuntu0.20.04.1]
  NVIDIA driver metapackage
> sudo apt install nvidia-driver-515
> sudo reboot
 CUDAインストール
Dockerで使う場合はいらないが、ホスト側でも使いたいのでとりあえずインストールしている。バージョンはフレームワークに合わせる必要があるので、ver11.2をインストールした。

（なので、Dockerでやったほうがいい）
> wget https://developer.download.nvidia.com/compute/cuda/11.2.2/local_installers/cuda_11.2.2_460.32.03_linux.run
> sudo sh cuda_11.2.2_460.32.03_linux.run
.bashrcに追加
> vi ~/.bashrc
export CUDA_HOME=/usr/local/cuda
export PATH=$PATH:$CUDA_HOME/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CUDA_HOME/lib64
> source ~/.bashrc
> nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Built on Sun_Feb_14_21:12:58_PST_2021
Cuda compilation tools, release 11.2, V11.2.152
Build cuda_11.2.r11.2/compiler.29618528_0
コマンドプロンプトで確認
> nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Built on Mon_May__3_19:41:42_Pacific_Daylight_Time_2021
Cuda compilation tools, release 11.3, V11.3.109
Build cuda_11.3.r11.3/compiler.29920130_0
 cuDNNのインストール
cuDNNはCUDAバージョンに対応したものを入れないと動作しない
&lt;/span&gt;&lt;a href="https://qiita.com/konzo_/items/a6f2e8818e5e8fcdb896" style="" target="_blank" rel="nofollow noopener noreferrer"&gt;https://qiita.com/konzo_/items/a6f2e8818e5e8fcdb896&lt;/a&gt;&lt;/p&gt;
&lt;p&gt;NVIDIAアカウントでログインする必要がある。&lt;/p&gt;
&lt;p&gt;&lt;span class="embed-block zenn-embedded zenn-embedded-card"&gt;&lt;iframe id="zenn-embedded__1451c223b2a1d" src="https://embed.zenn.studio/card#zenn-embedded__1451c223b2a1d" data-content="https%3A%2F%2Fdeveloper.nvidia.com%2Fcompute%2Fmachine-learning%2Fcudnn%2Fsecure%2F8.1.1.33%2F11.2_20210301%2Fcudnn-11.2-linux-x64-v8.1.1.33.tgz" frameborder="0" scrolling="no" loading="lazy"/&gt;&lt;/span&gt;&lt;a href="https://developer.nvidia.com/compute/machine-learning/cudnn/secure/8.1.1.33/11.2_20210301/cudnn-11.2-linux-x64-v8.1.1.33.tgz" style="" target="_blank" rel="nofollow noopener noreferrer"&gt;https://developer.nvidia.com/compute/machine-learning/cudnn/secure/8.1.1.33/11.2_20210301/cudnn-11.2-linux-x64-v8.1.1.33.tgz&lt;/a&gt;&lt;/p&gt;
&lt;div class="code-block-container"&gt;&lt;pre class="language-bash"&gt;&lt;code class="language-bash"&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;tar&lt;/span&gt; xzvf cudnn-11.2-linux-x64-v8.1.1.33.tgz
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;cp&lt;/span&gt; cuda/include/cudnn.h /usr/local/cuda/include
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;cp&lt;/span&gt; &lt;span class="token parameter variable"&gt;-P&lt;/span&gt; cuda/lib64/libcudnn /usr/local/cuda/lib64
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;chmod&lt;/span&gt; a+r /usr/local/cuda/include/cudnn.h /usr/local/cuda/lib64/libcudnn
&lt;a class="header-anchor-link" href="#docker%E3%82%A4%E3%83%B3%E3%82%B9%E3%83%88%E3%83%BC%E3%83%AB" aria-hidden="true"/&gt; Dockerインストール&lt;/h3&gt;
&lt;p&gt;NVIDIAドライバ関連がインストールできたらDockerをインストールする&lt;/p&gt;
&lt;div class="code-block-container"&gt;&lt;pre class="language-bash"&gt;&lt;code class="language-bash"&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;apt&lt;/span&gt; update &lt;span class="token operator"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;apt&lt;/span&gt; &lt;span class="token function"&gt;install&lt;/span&gt; ca-certificates &lt;span class="token function"&gt;curl&lt;/span&gt; gnupg lsb-release
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;mkdir&lt;/span&gt; &lt;span class="token parameter variable"&gt;-m&lt;/span&gt; 0755 &lt;span class="token parameter variable"&gt;-p&lt;/span&gt; /etc/apt/keyrings &lt;span class="token operator"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="token function"&gt;curl&lt;/span&gt; &lt;span class="token parameter variable"&gt;-fsSL&lt;/span&gt; https://download.docker.com/linux/ubuntu/gpg &lt;span class="token operator"&gt;|&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; gpg &lt;span class="token parameter variable"&gt;--dearmor&lt;/span&gt; &lt;span class="token parameter variable"&gt;-o&lt;/span&gt; /etc/apt/keyrings/docker.gpg
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token builtin class-name"&gt;echo&lt;/span&gt; &lt;span class="token punctuation"&gt;\&lt;/span&gt;
  &lt;span class="token string"&gt;"deb [arch=&lt;span class="token variable"&gt;&lt;span class="token variable"&gt;$(&lt;/span&gt;dpkg --print-architecture&lt;span class="token variable"&gt;)&lt;/span&gt;&lt;/span&gt; signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/ubuntu \
  &lt;span class="token variable"&gt;&lt;span class="token variable"&gt;$(&lt;/span&gt;lsb_release &lt;span class="token parameter variable"&gt;-cs&lt;/span&gt;&lt;span class="token variable"&gt;)&lt;/span&gt;&lt;/span&gt; stable"&lt;/span&gt; &lt;span class="token operator"&gt;|&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;tee&lt;/span&gt; /etc/apt/sources.list.d/docker.list &lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; /dev/null
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;apt&lt;/span&gt; update &lt;span class="token operator"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;apt-get&lt;/span&gt; &lt;span class="token function"&gt;install&lt;/span&gt; docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose-plugin
&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;div class="code-block-container"&gt;&lt;pre class="language-bash"&gt;&lt;code class="language-bash"&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;docker&lt;/span&gt; &lt;span class="token parameter variable"&gt;-v&lt;/span&gt;
Docker version &lt;span class="token number"&gt;23.0&lt;/span&gt;.1, build a5ee5b1
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;docker&lt;/span&gt; compose version
Docker Compose version v2.16.0
&lt;a class="header-anchor-link" href="#nvidia-container-toolkit%E3%81%AE%E3%82%A4%E3%83%B3%E3%82%B9%E3%83%88%E3%83%BC%E3%83%AB" aria-hidden="true"/&gt; NVIDIA Container Toolkitのインストール&lt;/h3&gt;
&lt;p&gt;Dockerだけではダメで、これを入れる必要がある。&lt;/p&gt;
&lt;p&gt;&lt;span class="embed-block zenn-embedded zenn-embedded-card"&gt;&lt;iframe id="zenn-embedded__5a56b1d070305" src="https://embed.zenn.studio/card#zenn-embedded__5a56b1d070305" data-content="https%3A%2F%2Fdocs.nvidia.com%2Fdatacenter%2Fcloud-native%2Fcontainer-toolkit%2Finstall-guide.html%23setting-up-nvidia-container-toolkit" frameborder="0" scrolling="no" loading="lazy"/&gt;&lt;/span&gt;&lt;a href="https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#setting-up-nvidia-container-toolkit" style="" target="_blank" rel="nofollow noopener noreferrer"&gt;https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html#setting-up-nvidia-container-toolkit&lt;/a&gt;&lt;/p&gt;
&lt;div class="code-block-container"&gt;&lt;pre class="language-bash"&gt;&lt;code class="language-bash"&gt;&lt;span class="token assign-left variable"&gt;distribution&lt;/span&gt;&lt;span class="token operator"&gt;=&lt;/span&gt;&lt;span class="token variable"&gt;&lt;span class="token variable"&gt;$(&lt;/span&gt;&lt;span class="token builtin class-name"&gt;.&lt;/span&gt; /etc/os-release&lt;span class="token punctuation"&gt;;&lt;/span&gt;&lt;span class="token builtin class-name"&gt;echo&lt;/span&gt; $ID$VERSION_ID&lt;span class="token variable"&gt;)&lt;/span&gt;&lt;/span&gt; &lt;span class="token punctuation"&gt;\&lt;/span&gt;
  &lt;span class="token operator"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="token function"&gt;curl&lt;/span&gt; &lt;span class="token parameter variable"&gt;-fsSL&lt;/span&gt; https://nvidia.github.io/libnvidia-container/gpgkey &lt;span class="token operator"&gt;|&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; gpg &lt;span class="token parameter variable"&gt;--dearmor&lt;/span&gt; &lt;span class="token parameter variable"&gt;-o&lt;/span&gt; /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg &lt;span class="token punctuation"&gt;\&lt;/span&gt;
  &lt;span class="token operator"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="token function"&gt;curl&lt;/span&gt; &lt;span class="token parameter variable"&gt;-s&lt;/span&gt; &lt;span class="token parameter variable"&gt;-L&lt;/span&gt; https://nvidia.github.io/libnvidia-container/&lt;span class="token variable"&gt;$distribution&lt;/span&gt;/libnvidia-container.list &lt;span class="token operator"&gt;|&lt;/span&gt; &lt;span class="token punctuation"&gt;\&lt;/span&gt;
  &lt;span class="token function"&gt;sed&lt;/span&gt; &lt;span class="token string"&gt;'s#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g'&lt;/span&gt; &lt;span class="token operator"&gt;|&lt;/span&gt; &lt;span class="token punctuation"&gt;\&lt;/span&gt;
  &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;tee&lt;/span&gt; /etc/apt/sources.list.d/nvidia-container-toolkit.list
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;apt&lt;/span&gt; update &lt;span class="token operator"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; &lt;span class="token function"&gt;apt&lt;/span&gt; &lt;span class="token function"&gt;install&lt;/span&gt; &lt;span class="token parameter variable"&gt;-y&lt;/span&gt; nvidia-container-toolkit
&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;p&gt;以下の&lt;strong&gt;nvidia-ctk&lt;/strong&gt;コマンドを実行するのが重要で、docker composeで&lt;code&gt;runtime: nvidia&lt;/code&gt;を指定すると「そんなサービス知らない」と言われた記憶がある。&lt;/p&gt;
&lt;p&gt;ここまで来たら、Dockerを再起動してインストールは完了。&lt;/p&gt;
&lt;div class="code-block-container"&gt;&lt;pre class="language-bash"&gt;&lt;code class="language-bash"&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; nvidia-ctk runtime configure &lt;span class="token parameter variable"&gt;--runtime&lt;/span&gt;&lt;span class="token operator"&gt;=&lt;/span&gt;docker
INFO&lt;span class="token punctuation"&gt;[&lt;/span&gt;0000&lt;span class="token punctuation"&gt;]&lt;/span&gt; Loading &lt;span class="token function"&gt;docker&lt;/span&gt; config from /etc/docker/daemon.json 
INFO&lt;span class="token punctuation"&gt;[&lt;/span&gt;0000&lt;span class="token punctuation"&gt;]&lt;/span&gt; Config &lt;span class="token function"&gt;file&lt;/span&gt; does not exist, creating new one 
INFO&lt;span class="token punctuation"&gt;[&lt;/span&gt;0000&lt;span class="token punctuation"&gt;]&lt;/span&gt; Flushing &lt;span class="token function"&gt;docker&lt;/span&gt; config to /etc/docker/daemon.json 
INFO&lt;span class="token punctuation"&gt;[&lt;/span&gt;0000&lt;span class="token punctuation"&gt;]&lt;/span&gt; Successfully flushed config                  
INFO&lt;span class="token punctuation"&gt;[&lt;/span&gt;0000&lt;span class="token punctuation"&gt;]&lt;/span&gt; Wrote updated config to /etc/docker/daemon.json 
INFO&lt;span class="token punctuation"&gt;[&lt;/span&gt;0000&lt;span class="token punctuation"&gt;]&lt;/span&gt; It is recommended that the &lt;span class="token function"&gt;docker&lt;/span&gt; daemon be restarted. 
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;cat&lt;/span&gt; /etc/docker/daemon.json
    &lt;span class="token string"&gt;"runtimes"&lt;/span&gt;&lt;span class="token builtin class-name"&gt;:&lt;/span&gt; &lt;span class="token punctuation"&gt;{&lt;/span&gt;
        &lt;span class="token string"&gt;"nvidia"&lt;/span&gt;&lt;span class="token builtin class-name"&gt;:&lt;/span&gt; &lt;span class="token punctuation"&gt;{&lt;/span&gt;
            &lt;span class="token string"&gt;"args"&lt;/span&gt;&lt;span class="token builtin class-name"&gt;:&lt;/span&gt; &lt;span class="token punctuation"&gt;[&lt;/span&gt;&lt;span class="token punctuation"&gt;]&lt;/span&gt;,
            &lt;span class="token string"&gt;"path"&lt;/span&gt;&lt;span class="token builtin class-name"&gt;:&lt;/span&gt; &lt;span class="token string"&gt;"nvidia-container-runtime"&lt;/span&gt;
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;sudo&lt;/span&gt; systemctl restart &lt;span class="token function"&gt;docker&lt;/span&gt;
&lt;a class="header-anchor-link" href="#%E3%82%B3%E3%83%B3%E3%83%86%E3%83%8A%E5%86%85%E3%81%8B%E3%82%89gpu%E3%83%99%E3%83%B3%E3%83%81%E3%83%9E%E3%83%BC%E3%82%AF%E3%82%92%E8%A8%88%E6%B8%AC%E3%81%97%E3%81%A6%E3%81%BF%E3%82%8B" aria-hidden="true"/&gt; コンテナ内からGPUベンチマークを計測してみる&lt;/h3&gt;
&lt;p&gt;認識できてなければ、計測できない。&lt;/p&gt;
&lt;p&gt;&lt;span class="embed-block zenn-embedded zenn-embedded-card"&gt;&lt;iframe id="zenn-embedded__9cddb3a5b5ce8" src="https://embed.zenn.studio/card#zenn-embedded__9cddb3a5b5ce8" data-content="https%3A%2F%2Fzenn.dev%2Fusagi1975%2Farticles%2F2023-02-18-1400_docker-gpu-bench" frameborder="0" scrolling="no" loading="lazy"/&gt;&lt;/span&gt;&lt;a href="https://zenn.dev/usagi1975/articles/2023-02-18-1400_docker-gpu-bench" style="" target="_blank"&gt;https://zenn.dev/usagi1975/articles/2023-02-18-1400_docker-gpu-bench&lt;/a&gt;&lt;/p&gt;
&lt;div class="code-block-container"&gt;&lt;pre class="language-bash"&gt;&lt;code class="language-bash"&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;docker&lt;/span&gt; run &lt;span class="token parameter variable"&gt;--rm&lt;/span&gt; &lt;span class="token parameter variable"&gt;--gpus&lt;/span&gt; all nvcr.io/nvidia/k8s/cuda-sample:nbody nbody &lt;span class="token parameter variable"&gt;-gpu&lt;/span&gt; &lt;span class="token parameter variable"&gt;-benchmark&lt;/span&gt; &lt;span class="token parameter variable"&gt;-numbodies&lt;/span&gt;&lt;span class="token operator"&gt;=&lt;/span&gt;&lt;span class="token number"&gt;640000&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;&lt;/div&gt;&lt;div class="code-block-container"&gt;&lt;pre class="language-bash"&gt;&lt;code class="language-bash"&gt;GPU Device &lt;span class="token number"&gt;0&lt;/span&gt;: &lt;span class="token string"&gt;"Ampere"&lt;/span&gt; with compute capability &lt;span class="token number"&gt;8.6&lt;/span&gt;
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; Compute &lt;span class="token number"&gt;8.6&lt;/span&gt; CUDA device: &lt;span class="token punctuation"&gt;[&lt;/span&gt;NVIDIA GeForce RTX &lt;span class="token number"&gt;3070&lt;/span&gt;&lt;span class="token punctuation"&gt;]&lt;/span&gt;
number of bodies &lt;span class="token operator"&gt;=&lt;/span&gt; &lt;span class="token number"&gt;640000&lt;/span&gt;
&lt;span class="token number"&gt;640000&lt;/span&gt; bodies, total &lt;span class="token function"&gt;time&lt;/span&gt; &lt;span class="token keyword"&gt;for&lt;/span&gt; &lt;span class="token number"&gt;10&lt;/span&gt; iterations: &lt;span class="token number"&gt;7012.648&lt;/span&gt; ms
&lt;span class="token operator"&gt;=&lt;/span&gt; &lt;span class="token number"&gt;584.087&lt;/span&gt; billion interactions per second
&lt;span class="token operator"&gt;=&lt;/span&gt; &lt;span class="token number"&gt;11681.750&lt;/span&gt; single-precision GFLOP/s at &lt;span class="token number"&gt;20&lt;/span&gt; flops per interaction
&lt;a class="header-anchor-link" href="#(%E3%81%9D%E3%81%AE%E4%BB%96%EF%BC%89tensorflow%E3%81%AE%E3%82%B3%E3%83%B3%E3%83%86%E3%83%8A%E3%81%A7%E7%A2%BA%E8%AA%8D" aria-hidden="true"/&gt; (その他）Tensorflowのコンテナで確認&lt;/h2&gt;
&lt;div class="code-block-container"&gt;&lt;pre class="language-bash"&gt;&lt;code class="language-bash"&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;docker&lt;/span&gt; run &lt;span class="token parameter variable"&gt;--gpus&lt;/span&gt; all &lt;span class="token parameter variable"&gt;-it&lt;/span&gt; &lt;span class="token parameter variable"&gt;--rm&lt;/span&gt; tensorflow/tensorflow:2.8.1-gpu &lt;span class="token function"&gt;bash&lt;/span&gt;
________                               _______________
___  __/__________________________________  ____/__  /________      __
__  /  _  _ &lt;span class="token punctuation"&gt;\&lt;/span&gt;_  __ &lt;span class="token punctuation"&gt;\&lt;/span&gt;_  ___/  __ &lt;span class="token punctuation"&gt;\&lt;/span&gt;_  ___/_  /_   __  /_  __ &lt;span class="token punctuation"&gt;\&lt;/span&gt;_ &lt;span class="token operator"&gt;|&lt;/span&gt; /&lt;span class="token operator"&gt;|&lt;/span&gt; / /
_  /   /  __/  / / /&lt;span class="token punctuation"&gt;(&lt;/span&gt;__  &lt;span class="token punctuation"&gt;)&lt;/span&gt;/ /_/ /  /   _  __/   _  / / /_/ /_ &lt;span class="token operator"&gt;|&lt;/span&gt;/ &lt;span class="token operator"&gt;|&lt;/span&gt;/ /
/_/    &lt;span class="token punctuation"&gt;\&lt;/span&gt;___//_/ /_//____/ &lt;span class="token punctuation"&gt;\&lt;/span&gt;____//_/    /_/      /_/  &lt;span class="token punctuation"&gt;\&lt;/span&gt;____/____/&lt;span class="token operator"&gt;|&lt;/span&gt;__/
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; python
&lt;span class="token operator"&gt;&amp;gt;&amp;gt;&lt;/span&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; from tensorflow.python.client &lt;span class="token function"&gt;import&lt;/span&gt; device_lib
&lt;span class="token operator"&gt;&amp;gt;&amp;gt;&lt;/span&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; print&lt;span class="token punctuation"&gt;(&lt;/span&gt;device_lib.list_local_devices&lt;span class="token punctuation"&gt;(&lt;/span&gt;&lt;span class="token punctuation"&gt;))&lt;/span&gt;
incarnation: &lt;span class="token number"&gt;6470249754415725170&lt;/span&gt;
physical_device_desc: &lt;span class="token string"&gt;"device: 0, name: NVIDIA GeForce RTX 3070, pci bus id: 0000:07:00.0, compute capability: 8.6"&lt;/span&gt;
xla_global_id: &lt;span class="token number"&gt;416903419&lt;/span&gt;
&lt;a class="header-anchor-link" href="#(%E3%81%9D%E3%81%AE%E4%BB%96%EF%BC%89pytorch%E3%81%AE%E3%82%B3%E3%83%B3%E3%83%86%E3%83%8A%E3%81%A7%E7%A2%BA%E8%AA%8D" aria-hidden="true"/&gt; (その他）PyTorchのコンテナで確認&lt;/h2&gt;
&lt;div class="code-block-container"&gt;&lt;pre class="language-bash"&gt;&lt;code class="language-bash"&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;docker&lt;/span&gt; pull pytorch/pytorch:1.10.0-cuda11.3-cudnn8-devel
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;docker&lt;/span&gt; run &lt;span class="token parameter variable"&gt;-it&lt;/span&gt; &lt;span class="token parameter variable"&gt;--rm&lt;/span&gt; &lt;span class="token parameter variable"&gt;--gpus&lt;/span&gt; all pytorch/pytorch:1.10.0-cuda11.3-cudnn8-devel &lt;span class="token function"&gt;bash&lt;/span&gt;
&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; python
&lt;span class="token operator"&gt;&amp;gt;&amp;gt;&lt;/span&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; &lt;span class="token function"&gt;import&lt;/span&gt; torch
&lt;span class="token operator"&gt;&amp;gt;&amp;gt;&lt;/span&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; torch.cuda.is_available&lt;span class="token punctuation"&gt;(&lt;/span&gt;&lt;span class="token punctuation"&gt;)&lt;/span&gt;
&lt;span class="token operator"&gt;&amp;gt;&amp;gt;&lt;/span&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; torch.cuda.device_count&lt;span class="token punctuation"&gt;(&lt;/span&gt;&lt;span class="token punctuation"&gt;)&lt;/span&gt;
&lt;span class="token operator"&gt;&amp;gt;&amp;gt;&lt;/span&gt;&lt;span class="token operator"&gt;&amp;gt;&lt;/span&gt; torch.cuda.current_device&lt;span class="token punctuation"&gt;(&lt;/span&gt;&lt;span class="token punctuation"&gt;)&lt;/span&gt;