zafirshi/dev-base repository overview

⁠Image Name Description

This image is a base image for deep learning development environment, primarily addressing the issue of inconsistency between the current system environment and the program testing environment, and the complexity of reconfiguring the environment. The image has the following features:

• Derived from the cuda:12.1.0-cudnn8-devel-ubuntu22.04⁠ image, adapted for the new version of Pytorch 2.0+⁠.
• Installed Miniconda-py39_4.12.0⁠ environment in the /opt directory of the container.
• Contained common commands such as cmake,wget, curl, vim, tmux, htop, etc.
• Used user permissions to prevent container breakdown.

Users can directly pull the image and conveniently install the development environment required for deep learning project within the image.

⁠How to use this image

You can use this image by running it as a container:

docker run -itd \           
  -v <Path in Host>:<Path in Container> \
  -w <Workdir in Container> \
  -p 6006:6006 \
  --name=<Container Name> \
  --runtime=nvidia \
  --restart unless-stopped \
  --gpus all 
  <Image Name:Tag>

-v : Mount dataset volumes or code volumes into the container. If you want to mount multiple paths, repeat the use of the -v syntax for mounting.

-w : The default path after entering the container.

-p : Port mapping between the container and the host machine. 6006 is the default port for TensorBoard. Start TensorBoard in the container and bind it to 0.0.0.0

tensorboard --logdir=<path-to-logs> --host=0.0.0.0

--gpus: Visiable GPUs in container. If you want to specify the GPUs visible to the container, modify the CLI parameters as follows: --gpus "device=0,2".

⁠Reference

The docker is built from Github⁠