search

chao1224 / MoleculeSTM

Multi-modal Molecule Structure-text Model for Text-based Editing and Retrieval, Nat Mach Intell 2023 (https://www.nature.com/articles/s42256-023-00759-6)
https://chao1224.github.io/MoleculeSTM
Other
188 stars 18 forks source link

Hyperparameters for pre-training #4

Closed Humanoid-z closed 5 months ago

Humanoid-z commented 8 months ago

Hi, First of all thank you for your work! I noticed that the default hyperparameters for the pre-training scripts seem to be inconsistent with the paper, or some are not mentioned in the paper. Do you confirm that these are the hyperparameters that were ultimately chosen? If not is it possible to provide the hyperparameters for pre-training? Thank you.

chao1224 commented 8 months ago

Hi @SNIKCHS,

Thank you for your interest in our work. The hyperparameters reported in the paper are what we ran, and we will let you know once we fixed this.

Nipers commented 7 months ago

Dear author, I am a little confused, in the paper the learning rate for the chemical structure branch is {1e-5, 3e-5} for pre-training, but the script in the code sets the value as 1e-4, could you please let me know which one should I choose to reproduce your experiment results?

chao1224 commented 7 months ago

Hi @Nipers ,

Sorry for the confusion. Please follow the paper appendix. Now the configs (under the scripts folder) are only for demo, and I will update them in the next version.