How to fix this finetune, i do finetune with train data and valid data?

[gptsha]

i don’t know the problem. so, help me to fix it