Beginner · Fundamentals

Epoch

Visual diagram · (in preparation) · Math · (in preparation) · Worked example · 3 difficulty levels.

TL;DR. One full pass through the entire training dataset during model training.

Technical Definition

One full pass through the entire training dataset during model training.

How it works

Training typically requires many epochs because each pass nudges parameters slightly. Too few epochs underfit; too many overfit. Modern practice tracks validation loss after every epoch and stops early when it stops improving. For huge datasets and LLMs, training may complete in less than one epoch — every example is seen at most once.

Related Concepts

• Overfitting — When a model learns noise and specific patterns in training data too well, causing it to perform poorly on new, unseen data.
• Training — The process of adjusting a model's parameters so it learns patterns from labeled or unlabeled data.
• Batch — A group of training examples processed together in a single forward and backward pass.