Further pretrain on molecular datasets

[luispintoc]

I've been impressed with TabPFN's performance and am interested in launching an open-source project to finetune (further pretrain) it on large-scale molecular datasets (>100k datasets). I'd like to discuss this idea with the paper's authors. I'd appreciate it if someone could connect me with the authors. Thank you in advance!