Given 60K training samples, what is 1875 in each epoch?
1875 refers to number of mini batches processed per epoch. This term is also referred to as steps per epoch. When batch size is set to 32 (is the default if one isn’t specified), since there are 60_000 examples in the training set, 60_000 / 32 = 1875.
If not set manually, this value corresponds to ceil(train_dataset_size / batch_size)
Please see this for more on steps_per_epoch.
This helps. Thanks.