tgml 0.1.12

ML Workbench

TigerGraph ML Workbench

tgml provides a python toolkit for machine learning practitioners to integrate TigerGraph into their existing workflow. The core component of tgml is the graph loader, which behaves like a data loader for typical machine learning tasks. Putting differently, users can write their model training code as before but only replace the previous data loader with our graph loader; they will get batches of graph data for training as if the data is read from their local disk. tgml also provides syntactic sugar to the graph data processing APIs, so users can run algorithms such as PageRank on their graphs in TG as calling a normal Python function. Under the hood, tgml takes care of all the communications with the Graph Data Processing Service and convert the final output to a format that users need (dataframes and PyG graphs for now).

See the tutorial notebooks in the docs/examples folder on how to use the package. For tgml to work, the Graph Data Processing Service has to be running on the TigerGraph server.

Getting Started

Install from pypi

pip install tgml

Install from github for the hottest changes:

pip install git+https://github.com/TigerGraph-DevLabs/tgml.git -f https://data.pyg.org/whl/torch-1.10.0+cpu.html