Differences

This shows you the differences between two versions of the page.

--- public:docker [2025/02/22 11:58] – liang
+++ public:docker [2025/02/23 23:03] (current) – [Step 4: Work with PyTorch] liang
@@ Line 1: / Line 1: @@
-Last-updated: 2025/02/22 liangz
+Last-updated: 2025/02/23 liangz
 ====== Docker tutorial ======
@@ Line 5: / Line 5: @@
 ===== Step 1: Understand docker =====
-  * Read [[https://docs.docker.com/get-started/docker-overview/|What is docker]] and understand (docker) image, container, daemon, client, hub (registry), pull, push.
+  * Read https://docs.docker.com/get-started/docker-overview/ and understand (docker) image, container, daemon, client, hub (registry), pull, push.
-  * Install docker (Linux as example. Information on other OSes are welcome too.)
+  * Install docker (Linux Mint as example. Information on other OSes are welcome too.)
     * $ sudo apt install docker.io
-    * Add your username to the "docker" group by editing /etc/group; then reboot your Linux
+    * Add your username to the "docker" group (or by editing /etc/group directly).
+      * $ sudo usermod -aG docker your_username
+      * Logout & login to make the change of group into effect, or simply reboot your Linux.
   * Run a ubuntu container (docker client will automatically pull a ubuntu image and create a local container.). If you get a console with command prompt "root@xxxxxxxxxxxx:/," it worked.
     * $ docker run -i -t ubuntu /bin/bash
   * Exit the container by exit (notice the prompt changed to "#", showing you are the root)
     * # exit
+===== Step 2 (if you have an NVidia GPU): Install Nvidia GPU support =====
+  * Follow https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html
+===== Step 3: Install PyTorch (for development) =====
+  * $ docker pull pytorch/pytorch:latest
+  * $ docker run --rm --gpus all pytorch/pytorch:latest python -c "import torch; print(torch.cuda.is_available())"
+  * If you see "True", it worked.
+===== Step 4: Work with PyTorch =====
+  * $ docker run -it --gpus all --rm pytorch/pytorch:latest bash
+  * # nvidia-smi
+  * If you see the use of your GPU, it worked.
+===== References =====
+  * https://github.com/saikhu/Docker-Guide-for-AI-Model-Development-and-Deployment (Chapter 6 is obsolete. See the above Step 2 for the latest information)