Hacker News new | ask | show | jobs
by wongarsu 326 days ago
Exactly. Underfitting would be if the model doesn't pick up on the fact that outro silence is labeled differently from regular silence and transcribes them the same