A farmers-market inspection desk watches training loss fall for many extra epochs while the validation score has already peaked and begun to worsen. What should stop the run?
Select an answer to reveal the explanation.
Short Explanation
Training loss can keep falling after the useful run is over. Stop on the validation curve, early stopping or any similar val-based halt. Driving loss to zero, reading GPU heat, or writing a kernel is not that signal.
Full Explanation
The validation metric is the stop and select signal for an adaptation run. Training loss can keep falling after the model has started to overfit the training notes, which is exactly when the validation score peaks and then worsens. Early stopping — or any similar val-based halt — ends the run at that useful point. Driving loss to zero, reading GPU heat, or writing a kernel does not replace the validation curve.