Question about finetuning on small datasets

[HankYe]

Hi, authors! Thanks for your great work! But I have a question about the hyper-parameter setting of fine-tuning on small datasets, which is not mentioned in the paper. I'd appreciate it greatly if you could share some details!