Ahmadzei's picture
added 3 more tables for large emb model
5fa1a76
Before diving in, we should note
that the metric applies specifically to classical language models (sometimes called autoregressive or causal language
models) and is not well defined for masked language models like BERT (see summary of the models).