When preparing the data, we ran into a problem: after the first step of training, the model could translate over 90% of the training documents.
Original titleWhen preparing the data, we ran into a problem: after the first step of training, the model could translate over 90% of the training docu...
AISummary
Kocmi explains how we got the most difficult samples to level up North Small Translate's capabilities: (3/6)
Source: Cohere · x.com