Missing LoRARequest in GSM8K evaluation script

[yuxizhe]

Thanks for the great work on this repository!

I found that the evaluation script wasn't applying the LoRARequest properly. This meant we were just evaluating the base model every time, which is why the results looked almost the same as pre-training. #14

I have created a PR to add the missing request and ensure the adapters are loaded. #19

[joeyu930]

That makes sense.

Could you share the evaluation metrics when the LoRA is applied?
I'm curious how different the results are compared to pre-training.