/graph2tac

graph2tac converts graphs to tactics

Primary LanguageJupyter NotebookApache License 2.0Apache-2.0

graph2tac

graph2tac converts graphs to tactics

prerequisites on Linux

  • standard linux c++ compiler build chain (tested on g++)

  • capnproto https://capnproto.org : need capnp proto compiler, the library and the headers. On Ubuntu you can install capnproto and libcapnp-dev.

  • CUDA/GPU (optional, if you want to use GPU): please follow https://www.tensorflow.org/install to install on your system.

installation

The python graph2tac package can be installed with pip install in a standard way from git repo (we aren't yet on pypi.org). For developers a recommended way is pip install -e . from the cloned source repository which allows to develop directly on the source of the installed package. Our key (recommended/tested) pip dependencies are python>=3.9, tensorflow>=2.8 and tensorflow-gnn>=0.2.0.dev1.

entry-points

  • See g2t-train --help to train the model
  • See g2t-train-tfgnn --help to train the TF-GNN model
  • See g2t-server --help to launch the python server for evaluation