Comment by starik36

From the article: Pre-trained on a trillion "tokens"...

Doesn't 7B indicates that it was trained on 7 billion tokens? Or am I misunderstanding the nomenclature?