In machine learning, an epoch is one complete pass of a model through the entire training dataset. Training data is usually processed in batches, and one epoch is completed when every batch has been processed once. The number of epochs is a standard hyperparameter alongside the learning rate and batch size.
During training, loss and accuracy are typically recorded at each epoch to check whether the model is converging or overfitting the training data. Too few epochs leave the model undertrained, while too many can lead to overfitting. In general usage the word refers to an era or a reference point in time.