Adding validation set evaluation for reranker finetuning

[jackkwok]

To avoid overfitting, I would like to add validation (eval) dataset to the finetuning. I see the BiTrainer is a subclass of Huggingface's Trainer class which accepts a eval_dataset argument. I plan to add the eval dataset within the initialization of BiTrainer (line).

Let me know if there are other considerations.

[EEEnterURname]

@jackkwok Did you successfully add the validation dataset?If successful, can you share how to achieve it?