You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Dockerfile contains all the required installations (including Intel MKL and TBB) above. Only additional requirement is to provide NVIDIA*.run script that will load the same NVIDIA driver of host into the target. Then to build the container, execute:
docker build -t "graphnn:test" .
To run it:
docker run --runtime=nvidia graphnn:test bash
If above command fails for a reason, refer to https://github.com/NVIDIA/nvidia-docker. If no error occurs, you can simply follow the below instructions and execute them in the container without failure.
Build static library
cp make_common.example make_common
modify configurations in make_common file
make -j8
Run example
Run mnist
cd examples/mnist
make
./run.sh
Run graph classification
cd examples/graph_classification
make
./local_run.sh
The 5 datasets under the data/ folder are commonly used in graph kernel.
Reference
@article{dai2016discriminative,
title={Discriminative Embeddings of Latent Variable Models for Structured Data},
author={Dai, Hanjun and Dai, Bo and Song, Le},
journal={arXiv preprint arXiv:1603.05629},
year={2016}
}
About
Training computational graph on top of structured data (string, graph, etc)