dmlc / gnnlens2 Goto Github PK

Hi,
I follow the instructions in tutorial 1 and try to show the graph structure of ogbn-products. There is no error showed when I run the code, however, when I open the gnnlens website and select the ogbn-products graph in the graph selector, it shows nothing. I wonders what may cause this happen? Is it because the graph too large, or because the graph does not include the label? Hope I describe it clearly, if needed I could show my code.

Hi there,

I'm wondering if the limit of graph size has been tested, such as the maximal number of nodes and edges in the visualized graph? And the correspondence between the memory overhead/simulation time and the graph size also interests me. For example, how big the graph is that can be processed by the GNNLens2 within 10secs on a machine with 16 CPUs and 64GB memory?

Hello! Thank you so much for the wonderful tool and the extensive tutorials! I have a question regarding the nodes and edges colors. Is there a way I can specify the nodes color based on a color of choice? And for nodes, is it possible to change the color from grey to a color of choice? Thank you

Since 2.1.0, flask deprecates safe_join, as elaborated in its release note here. For now, a workaround is to degrade flask to an older version like pip install Flask==2.0.3. This should be fixed in the future release of GNNLens2 by either restricting Flask version or follow the latest recommended practice.

Credit to @SherylHYX for reporting the issue.

Thanks for this wonderful library. In writer.add_graph(), I saw that it only accepts homogeneous graphs, do you have any plans to support heterogeneous graphs?

I'd like to ask tutoria_graph placed in its own directory or in the library installation directory of Anaconda

If not, how can we adapt PyG data and networks to this framework? Thank you!