You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hello, I have tried to train DNABERT models with early_stop = 1 and early_stop = 5 but both times my training log looks like below. It seems the early_stop condition is not being triggered. While I am only recording loss here, and I understand that the early_stop is triggered by auc, I have tested the model performance at different checkpoints. As expected, auc is low when loss ~0.7 and auc is high when loss ~0.3. Maybe I am not understanding something about the early stop condition: is early_stop triggered after each epoch or training step?
Hello, I have tried to train DNABERT models with early_stop = 1 and early_stop = 5 but both times my training log looks like below. It seems the early_stop condition is not being triggered. While I am only recording loss here, and I understand that the early_stop is triggered by auc, I have tested the model performance at different checkpoints. As expected, auc is low when loss ~0.7 and auc is high when loss ~0.3. Maybe I am not understanding something about the early stop condition: is early_stop triggered after each epoch or training step?
Thanks so much for your help!
Training log
The text was updated successfully, but these errors were encountered: