Replies: 2 comments
-
Thanks for the pointer. This is very interesting and a cool contribution to PyG. We currently do not have plans to implement this on our own, so I label this as "Help wanted" if anyone is interested to contribute this. |
Beta Was this translation helpful? Give feedback.
-
Hi, a more recent method from the paper Geometry-enhanced molecular representation learning for property prediction seems to outperform GROVER. They also have their own official implementation here. Just in case someone wanted to prioritize the SOTA. Edit: I've realized that there's an update to GEM, GEM-2, and neither of these are SOTA on some of the benchmarks they were superior on before (e.g. the OGB Quantum Chemistry Leaderboard). Uni-Mol+ seems to be the the top method at the moment. |
Beta Was this translation helpful? Give feedback.
-
There is a GTransformer architecture from the paper "Self-Supervised Graph Transformer on Large-Scale Molecular Data". The github link is https://github.com/tencent-ailab/grover. Has this been implemented with the pyG api? If not, can this be considered for future implementations, as this model preforms better than AttentiveFP (existing molecular implementation with pyG).
Beta Was this translation helpful? Give feedback.
All reactions