Usage¶

Requirements¶

GPU images: Docker with NVIDIA GPU support and host driver >= 570.x (required for CUDA 12.8)
CPU images: Docker

CUDA version

GPU images use CUDA 12.8, supporting Pascal (GTX 10xx) through Blackwell (RTX 50xx/B100/B200). CUDA 12.x is the last series to support Pascal; CUDA 13.0 removed Pascal, Maxwell, and Volta support entirely.

Devcontainer repositories¶

The recommended way to use these images. Each devcontainer repo wraps an image with VS Code configuration, a standard project layout, common extensions, and an environment verification notebook. To start: click Use this template on the repo, clone it, then open it in VS Code; the container starts automatically.

Repo	Use case
`datascience-devcontainer`	Intro Python and ML
`deeplearning-devcontainer`	TensorFlow + PyTorch
`llms-devcontainer`	LLM application development
`kaggle-devcontainer`	Kaggle competitions

See each repo's README for full setup instructions, hardware requirements, and environment details.

Standalone usage¶

The images can also be used directly with Docker, without VS Code Dev Containers.

JupyterLab¶

All images include JupyterLab. Start a container with the port exposed, then open the printed URL in a browser.

CPU images:

docker run --rm -it \
  -p 8888:8888 \
  -v $(pwd):/workspace \
  gperdrizet/datascience-cpu:latest \
  jupyter lab --ip=0.0.0.0 --no-browser --allow-root

NVIDIA GPU images (add --gpus all):

docker run --rm -it \
  --gpus all \
  -p 8888:8888 \
  -v $(pwd):/workspace \
  gperdrizet/datascience-nvidia:latest \
  jupyter lab --ip=0.0.0.0 --no-browser --allow-root

Copy the http://127.0.0.1:8888/lab?token=... URL from the terminal output and open it in a browser. Replace datascience-cpu / datascience-nvidia with whichever image you need.

Persistent storage

The -v $(pwd):/workspace flag bind-mounts your current directory into /workspace inside the container. Files written there survive container removal (--rm). Run the command from whichever directory you want to work in.

VS Code: attach to running container¶

This gives a full VS Code experience inside the container with no extra software beyond the Dev Containers extension.

1. Start the container (keep it running with an interactive shell):

# CPU
docker run --rm -it \
  -v $(pwd):/workspace \
  gperdrizet/datascience-cpu:latest \
  /bin/bash

# NVIDIA GPU
docker run --rm -it \
  --gpus all \
  -v $(pwd):/workspace \
  gperdrizet/datascience-nvidia:latest \
  /bin/bash

2. Attach VS Code:

Open the Command Palette (Ctrl+Shift+P)
Run Dev Containers: Attach to Running Container...
Select the container from the list

Your mounted volume is available at /workspace. Extensions you install persist for the duration of the session.