Open 12138xs opened 1 week ago
Hi,
I looked into this. The 'coefficients', 'exponents', and 'norms' have the correct shape. It turns out the data sets were missing information from the first two atoms. I've uploaded the corrected data sets as version 2 in the repository:
https://zenodo.org/records/13932625
Thanks for catching this, and good luck with your research!
On Mon, Oct 14, 2024 at 8:43 PM CenJH @.***> wrote:
I am currently trying to reproduce the results of the ML-DNA section of your model. During this process, I encountered a discrepancy between the documented and actual data shapes for the 'coefficients', 'exponents', and 'norms' keys. The documentation specifies that these keys should have a shape of (n_atom, ...). However, the actual shape in the training data, as found at Zenodo link https://zenodo.org/records/7072319, is (n_atom + 2, ...). Could you please help identify the reason for this shape discrepancy? Any insights or clarifications would be greatly appreciated.
Thank you.
— Reply to this email directly, view it on GitHub https://github.com/JoshRackers/equivariant_electron_density/issues/4, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADRD5SI6NU4YENE7QSX6B5LZ3R6MDAVCNFSM6AAAAABP6D3ZCSVHI2DSMVQWIX3LMV43ASLTON2WKOZSGU4DONBWGE4DMOA . You are receiving this because you are subscribed to this thread.Message ID: @.***>
I am currently trying to reproduce the results of the ML-DNA section of your model. During this process, I encountered a discrepancy between the documented and actual data shapes for the 'coefficients', 'exponents', and 'norms' keys. The documentation specifies that these keys should have a shape of (n_atom, ...). However, the actual shape in the training data, as found at Zenodo link, is (n_atom + 2, ...). Could you please help identify the reason for this shape discrepancy? Any insights or clarifications would be greatly appreciated.
Thank you.