AI Glossary

Epoch

What does “Epoch” mean?

Epoch: One complete pass through the entire training dataset. Smaller models are often trained for many epochs. Large language models typically see most of their pre-training data only once or a few times.

Epoch visual guide: One complete pass through the entire training dataset. Smaller models are often trained for many epochs. Large language models typically see most of their pre-training data only once or a few times.
Visual guide Open full size (new tab)

What is an example of “Epoch”?

Training an image classifier for too many epochs on a small dataset can lead to overfitting, so teams watch validation scores after each pass.

Keep learning

See also

Batch sizeTermOverfittingTermCheckpointTermValidation and test setsTerm

Same topic

HyperparameterTermUnderfittingTermFine-tuningTermDirect preference optimizationTermGradient descentTerm