Dear scGPT Team,

I want to express my admiration for the remarkable work you have done with scGPT! I am particularly interested in the implementation details of the scGPT pretraining process.

I have a few questions regarding the architecture of the model. Specifically, I noticed that the cell embedding function is defined in both scGPT/scGPT/model/model.py and scGPT/scGPT/tasks/cell_emb.py. This overlap has left me uncertain about how these functions interact or differ.

Could you please provide a brief explanation of the pretraining code logic? Additionally, if possible, I would greatly appreciate it if you could share an example of the pretraining code.

Thank you very much for your time and help!