Do you support learning to distill knowledge from larger reranker models to smaller reranker models?

[QuangTQV]

[545999961]

We currently do not support distillation for reranker; however, you can label the data and modify the fine-tuning code to include a KL loss, thereby implementing this function.