3 ms·
Perplexity scores are pretty common, which I think involves taking a text corpus like wikipedia and seeing how well the model predicts the next word (token) of
by QuadrupleA 4y ago
Perplexity scores are pretty common, which I think involves taking a text corpus like wikipedia and seeing how well the model predicts the next word (token) of it.