Sign inSign up

gperdrizet/deeplearning-nvidia

By gperdrizet

•Updated 22 days ago

Image
0

2.5K

gperdrizet/deeplearning-nvidia repository overview

⁠Deep learning GPU development environment

A ready-to-use deep learning environment with NVIDIA GPU support for VS Code. Includes both PyTorch and TensorFlow frameworks. Designed for cross-platform support and wide GPU compatibility.

Base image: NVIDIA TensorFlow 25.02⁠

Note: This container is designed to be used as part of a VS Code Dev Container environment, not run directly. See the full environment setup on GitHub⁠ for devcontainer configuration and usage instructions.

CUDA version note: This image uses CUDA 12.8, which supports GPU architectures from 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. CUDA 12.8 requires host driver ≥ 570.x on Linux. Note: 25.02 is the final NVIDIA TensorFlow NGC container release.

⁠1. Features

  • Dual framework support: Both TensorFlow 2.17 and PyTorch 2.11.0 pre-installed and GPU-ready
  • GPU-accelerated arrays: CuPy 13.6.0 (custom wheel, sm_60-sm_100) for GPU NumPy-compatible computation
  • NVIDIA GPU acceleration: Built on NVIDIA's official TensorFlow container with CUDA 12.8 and Blackwell support
  • VS Code Dev Container ready: Pre-configured vscode user for seamless devcontainer integration

⁠2. Included software

ComponentVersion
Base Imagenvcr.io/nvidia/tensorflow:25.02-tf2-py3
TensorFlow2.17
Keras3.x
PyTorch2.11.0 (custom build, CUDA 12.8)
CuPy13.6.0 (custom build, sm_60-sm_100)
Python3.12
CUDA12.8
GPU SupportPascal through Blackwell (sm_60 - sm_100)
⁠2.1. Additional Python packages
  • jupyterlab - Notebook interface
  • pandas - Data manipulation
  • scikit-learn - Machine learning utilities
  • scipy - Scientific computing
  • matplotlib - Data visualization
  • seaborn - Statistical data visualization
  • tensorboard - Training visualization
  • keras_tuner - Hyperparameter tuning for Keras
  • optuna - Hyperparameter optimization framework
  • ipywidgets - Interactive widgets for Jupyter

⁠3. Usage with VS Code Dev Containers

This image is designed for use with the deeplearning-devcontainer⁠ repository, which provides ready-to-use VS Code Dev Container configurations. The vscode user (UID 1000) is pre-configured with sudo access for a seamless development experience.

⁠4. Running directly with Docker

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

⁠5. Requirements

  • Docker with NVIDIA GPU support and host driver >= 570.x
  • NVIDIA Container Toolkit

⁠6. License

See the GitHub repository⁠ for license information.

Tag summary

Content type

Image

Digest

sha256:47c8ad71d…

Size

8.6 GB

Last updated

22 days ago

docker pull gperdrizet/deeplearning-nvidia