4 ms·
Got any links to explanations of why fine tuning open models isn’t a productive solution? Besides renting the GPU time, what other downsides exist on today’s SO
by bradfa 11mo ago
Got any links to explanations of why fine tuning open models isn’t a productive solution?
Besides renting the GPU time, what other downsides exist on today’s SOTA open models for doing this?
- RC_ITR 11mo agoWhen the new pre-trained parameters come out in a new model generation, your old fine tuning doesn't apply to them.