Tokenization and Encoding

Unigram Language Model Tokenization

Tokenization method that initializes a large vocabulary and then iteratively reduces it by removing subwords with the least impact on the unigram model's likelihood, producing an optimal vocabulary.

← Wstecz