Changelog

Overview

This quarter we've moved many of our machines over to Debian, which has allowed us to focus more on CUDA/NVidia upgrades. As such we've updated our scipy-ml-notebook container to use CUDA 12.6, which has allowed us to upgrade Tensorflow and Torch.

Package Upgrades

• Tensorflow 2.15.0 -> 2.18.0
• Torch 2.2.1 -> 2.6.0
• Keras 2.15.0 -> 3.9.0
• Transformers 4.46.3 -> 4.49.0
• CUDA 12.1 -> 12.6
• CUDNN 8.9.2 -> 9.5.1

New Image Manifests:

Base notebook (Datascience): https://github.com/ucsd-ets/datahub-docker-stack/wiki/ghcr-io-ucsd-ets-datascience-notebook-2025.2-stable

Scipy-ML notebook: https://github.com/ucsd-ets/datahub-docker-stack/wiki/ghcr-io-ucsd-ets-scipy-ml-notebook-2025.2-stable

RStudio Notebook: https://github.com/ucsd-ets/datahub-docker-stack/wiki/ghcr-io-ucsd-ets-rstudio-notebook-2025.2-stable