Open ghost opened 3 years ago
Hi,
Yes I've done this in my notebook: https://github.com/NielsRogge/Transformers-Tutorials/blob/master/DETR/Fine_tuning_DetrForObjectDetection_on_custom_dataset_(balloon).ipynb.
I use PyTorch Lightning for early stopping (they provide a callback).
Hi guys, when I trained on my custom dataset, I encountered overfitting. So I want to stop training when val_loss increases. Or Is there a way to save the model when it has the highest mAP? Has anyone ever done this?