Hacker News new | ask | show | jobs
by RC_ITR 206 days ago
When the new pre-trained parameters come out in a new model generation, your old fine tuning doesn't apply to them.