Inquiry regarding hyperparameters and training configurations for the official weights

[Dondorodon]

Dear authors,

First of all, thank you for sharing this excellent work and the codebase. It has been very helpful for my research on protein-ligand interactions.

I am writing to ask about the specific training details used to generate the official pre-trained weights provided in this repository. While I have reviewed the paper and the code, I would like to know the exact hyperparameters and training options used for the final model to ensure I can reproduce the results accurately or use them as a proper baseline for fine-tuning.

Could you please share the specific configuration details? For example:

Hyperparameters: Learning rate, batch size, weight decay, dropout rate, etc.

Training Options: Number of epochs, scheduler settings, or any specific random seeds used.

Command Line Arguments: If possible, sharing the exact command line or the config file used to train the official model would be greatly appreciated.

Thank you for your time and assistance.

Best regards,