D-FINE icon indicating copy to clipboard operation
D-FINE copied to clipboard

Early Stopping/Patience Epochs help needed

Open FarhAnonymous opened this issue 9 months ago • 1 comments

Any idea how to automatically stop training if your validation metric (AP) doesn't improve for patience epochs, saving compute resources and preventing overfitting?

FarhAnonymous avatar Mar 05 '25 08:03 FarhAnonymous

You can use the 🤗 model implementation (#276) and then use early stopping with either the 🤗 Trainer API, PyTorch Lightning or native PyTorch.

NielsRogge avatar May 05 '25 15:05 NielsRogge