Open resnant opened 1 year ago
Thanks. This is a great addition. However, I would like to suggest for development to be postponed until the new reimplementation in the https://github.com/materialsvirtuallab/matgl repository to be completed. The implementation is actually already done but we need to port over the existing MEGNet models over. The new implementation is replace this TF implementation and the current repo will be deprecated.
Thanks for your response. I agree to postpone additional development of the current repo until the new implementation is completed. As a PyTorch user, I welcome the transition from TensorFlow to PyTorch and DGL. I'll check out the PyTorch implementation of MEGNet and M3GNet. Thank you again for the update, and please let me know if there are any other ways I can assist during the transition.
Hi, I have been exploring the application of MEGNet to structures with site mixing. I have written a simple example using
CrystalGraphDisordered
class to handle these structures, and I would appreciate it if you could review it to confirm that my understanding is correct and there are no processing errors.If this code is useful to others, I would like to submit a pull request as an example of processing disordered structures. I think that MEGNet's capability to handle disordered structures is a powerful advantage, especially for researchers dealing with experimentally determined crystal structures.
Thank you in advance.
Related:
317
https://github.com/materialsvirtuallab/megnet/blob/master/notebooks/model_reconstruct.ipynb