Does DGL support hundreds of millions of giant Graph?How is performance?
Which dataset are you using?
We have tried DGL on a synthetic graph with 500M nodes and runs OK. In terms of performance, it very much depends on the model.
We’re currently finalizing the code and will release them soon (probably by the end of this month). To evaluate our implementation, we create a reddit graph whose nodes are posts and edges indicate whether a user comments a post. It’s similar to the one used by GraphSage, but we sparsified it.
Do you have any suggestions what other giant graphs we should use for evaluation?
1 Like
Where will your code be published? I’m still learning and can’t give you effective advice. Sorry
We will release our code in the DGL github repo.