Add GMNConv and an example on ZINC

[asarigun]

Adding The Graph Mixer Networks (https://arxiv.org/abs/2301.12493) Conv operator and example on ZINC

[codecov]

Codecov Report

Merging #6775 (0af668c) into master (f71ead8) will increase coverage by 0.01%.
Report is 5 commits behind head on master.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master    #6775      +/-   ##
==========================================
+ Coverage   87.40%   87.42%   +0.01%     
==========================================
  Files         473      474       +1     
  Lines       28570    28631      +61     
==========================================
+ Hits        24972    25030      +58     
- Misses       3598     3601       +3     

Files	Coverage Δ
torch_geometric/nn/conv/__init__.py	100.00% <100.00%> (ø)
torch_geometric/nn/conv/gmn_conv.py	100.00% <100.00%> (ø)

... and 8 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[asarigun]

I hope you find the implementation fully functional for PyG!

[asarigun]

Reviewed the changes for GMNConv

[dufourc1]

Hi! Thanks for the implementation, I'll read the paper linked to make sure I'm aligned, you'll probably just need to add some tests.

[dufourc1]

Hi, sorry for the delay and thanks again for the implementation ! Here are a few notes on the code and some more general comments.

Parts of the method are similar to PNA (PNAConv). I wonder if there is a way of reusing the code already available ? What do you think ?

For the tests you might want to check what is done here https://github.com/pyg-team/pytorch_geometric/blob/master/test/nn/conv/test_pna_conv.py.

[asarigun]

Hi! Thanks for the feedback! I rearranged the code that you mentioned in the comments.

Hi, sorry for the delay and thanks again for the implementation ! Here are a few notes on the code and some more general comments.

Parts of the method are similar to PNA (PNAConv). I wonder if there is a way of reusing the code already available ? What do you think ?

For the tests you might want to check what is done here https://github.com/pyg-team/pytorch_geometric/blob/master/test/nn/conv/test_pna_conv.py.

[dufourc1]

Sorry for the delay, thanks a lot for the requested changes 👍🏽

I have added a few comments, but maybe it's only because I am confused with the similarity with PNAConv.

[asarigun]

Hello @dufourc1,

Thank you for your insights on the code. I have made updates in the sections you highlighted and added comments where you've provided feedback.

Sorry for the delay, thanks a lot for the requested changes 👍🏽

I have added a few comments, but maybe it's only because I am confused with the similarity with PNAConv.

[dufourc1]

Hi ! The main points look good to me. Once you've updated your code to pass the CI tests,
this should be all good.

[asarigun]

Thank you so much!

Hi ! The main points look good to me. Once you've updated your code to pass the CI tests, this should be all good.

[asarigun]

Hi @dufourc1,

The code has successfully passed the CI checks! Thank you once again for your feedback🙌.