Epoch
Also known as: epochs
An epoch is one complete pass through the whole training dataset. After one epoch, the model has seen every example once.
Training usually runs for several epochs — but not too many, or the model starts memorizing (overfitting) instead of learning.
- 1 epoch: full pass, once.
- Many epochs: repeated passes, refining the same data.
Watch the loss curve: it drops during training, then flattens. If it starts rising on the validation set, you've overfit — stop there.