qingy1337/graphml-base repository overview

latest: This image is a ready-to-run, CPU-only starting point for graph and general ML work.
It layers the official PyTorch 2.8.0-CPU release with the matching PyTorch-Geometric wheels (torch-scatter, torch-sparse, torch-geometric) built for the 2.8.0+cpu ABI so you can train or inference GNNs without compiling C++/CUDA extensions. Includes Java 17 and uv for package management. Core scientific-Python tools (NumPy, scikit-learn) plus huggingface_hub, datasets, and iv2_utils.

cu128: Essentially the same as latest except includes Python 3.11.13, CUDA 12.8, PyTorch 2.8, uv, huggingface_hub, datasets, etc. plus PyTorch GRL libraries are installed with CUDA support.