Epoch
One full pass of the model through the entire training dataset.
Think of it like
Like reading a whole textbook cover to cover once — an epoch is one complete read-through.
Example
Training for three epochs means the model sees every example in the dataset three times.
How it actually works
An epoch is one complete sweep over all training data, made up of many iterations (one per batch). Too few epochs and the model underfits; too many and it starts memorizing and overfitting. Large-scale pretraining often runs less than a single epoch over a giant corpus, since seeing that much data even once is plenty — the concept matters most for smaller fine-tuning runs.
For product teams
A rough dial on how long to train; more isn’t always better, since extra passes invite overfitting.
For engineers
One full pass over the dataset (many iterations); count trades underfitting vs. overfitting.
Related
- Iteration — Composed of many iterations.
- Overfitting — Too many risks overfitting.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome