ccsi-toolset / mgn Goto Github PK

Hi! First of all, thanks a lot for sharing this repository, it has a lot of interesting code.

I would like to ask you whether you implemented some dynamic remeshing strategy as well: I couldn't find it in your code so far, but perhaps I didn't look carefully enough. If not, are you planning to do it?

Thanks in advance :)

Is it possible to get versioning on the requirements listed in the README? It's been difficult to setup the environment.
Thanks!

Hi,

I noticed your code does not use the message passing base class from pyg due to not being able to update edges. I see an edge_update method on the base class so I was wondering if this was posted before the added functionality or if there is another reason why you chose not to use it.

Thanks for sharing this.

May I can why the train and test use the same dataset? (I am not sure if this is intended or if I ran the example incorrectly. Please correct me if I am wrong.)

If that's the case, is it possible to split the dataset into train and test from the already created "CylinderFlowDatasetNP" object? Or they just have to be loaded separately from scratch.

Hi,

Thank you so much for your efforts on this project. I have a question about GNS or MCN. I found DeepMind team had another paper (Graph Networks as Learnable Physics Engines for Inference and Control), which was similar to GNS or MCN. But they added recurrent to GNS or MCN. May I ask whether you have any idea about how to add recurrent to GNS or MCN based on PyG?

Thank you so much!

Hi, great project! it's unclear to me the difference between meshgraphnets and multimeshgraphnets.

Can someone add comments on the diff?