3 ms·
(2021), still very interesting. Especially the "post-overfitting" training strategy is unexpected.
by xg15 1y ago
(2021), still very interesting. Especially the "post-overfitting" training strategy is unexpected.
- luckystarr 1y agoI remember vaguely that this was observed when training GPT-3 (probably?) as well. Just trained on and on, and the error went up and then down again. Like a phase transition in the model.
- dev_hugepages 1y agoThis is talking about the double descent phenomenon (https://en.wikipedia.org/wiki/Double_descent https://en.wikipedia.org/wiki/Double_descent)