> pre-trained on Google News Collection like most Transformer models are BERT for example was trained on Wikipedia and Book corpora. Do you have a source for t

> pre-trained on Google News Collection like most Transformer models are BERT for example was trained on Wikipedia and Book corpora. Do you have a source for this claim?