hanyue1648 / reft Goto Github PK

请问什么时候可以开源代码呢~

Thank you for your effort. Any updates on the code?

Hi @hanyue1648 and @lxtGH, thanks for the great work! 😊

Did you tried using RefT without fine-tuning it on the novel categories, as some meta-learners get fairly decent performance anyways? In your experience, how long does it take the fine-tuning to converge in the incremental setting (speaking in terms of epochs or num. gradient steps)? I am interested in a rough estimate if you'd need a GPU for that or if this could technically also work on an embedded device.