Baby Llama: knowledge distillation from an ensemble of teachers trained on a small dataset with no performance penalty

Re-assign community